Subscribe now

A few months ago I realized that I was getting an increasing volume of cold email and LinkedIn outreach. Not normally a problem, but the quality of this outreach was abysmal and getting worse. Comically bad. You can see that evidence here and here. The thing is… I know it can work. It has worked incredibly well for me as an early scaling strategy at many, many companies over a 20+ year career. Cold outreach is one of the initial strategies that propelled Patreon’s growth and started its flywheel. Same with sparking initial Lyft demand.

Still, I’m bothered by the fact that most outreach seems to be heading in the wrong direction and it’s going to absolutely ruin it for everybody. I’m fearful that AI is only going to make it worse and make the senders even lazier. To try and fix this I asked my network who the absolute best and most effective people working on cold email outreach are and then reached out to the two experts who were consistently recommended above everyone else. There’s an irony here related to cold emailing… but let’s move past it. I’m thankful that Kellen Casebeer and Noah Adelstein graciously accepted my request to collaborate with me on this post.

Kellen Casebeer is the founder of The Deal Lab and an outbound marketing expert. He works with companies who are looking to build outbound marketing programs or give their existing ones a kick in the pants. Previously he was a VP of Sales and Head of GTM. You can find him on LinkedIn and TheDealLab.com

Noah Adelstein ran Rippling’s cold email efforts in 2023 when they first setup Clay for deeper personalization. He is now leading their entire automated growth function. He also hosts a newsletter and podcast called the GTM Engineer. He operates on the cutting edge of outbound GTM and you can find him on LinkedIn.

Read today’s newsletter to learn:

• Why cold email is broken and getting worse

• A step-by-step guide to fixing your cold email strategy

• Surprising urvey results across in-house and agency cold email senders

Subscribe now

In the beginning…

A long, long time ago there was a company called DiscoverOrg that was founded in 2007 by two gentlemen named Henry Schuck and Kirk Brown. Their mission was to “unlock growth potential for professionals and businesses through better data.” They started by gathering actionable business contact information: emails, phone numbers, important contacts, etc. to help sales, marketing and ops teams engage with prospects.

In 2019 DiscoverOrg bought a little company called Zoom Information which had started as Eliyon Technologies in 2000. They rebranded the entire thing (DiscoverOrg + Zoom Information + All their other acquisitions) into ZoomInfo and IPO’d as ZI in 2020. They have since changed their stock ticker to GTM and have a market cap of ~$3B as of the writing of this newsletter.

So why does this matter?

ZoomInfo sells business contact information to companies and has for a very, very long time. This contact information was incredibly valuable in the beginning when no one was emailing people. Reply rates were through the roof in an unsaturated market. But as the Law of Shitty Clickthroughs dictates, eventually people flood the zone and the effectiveness of the tactic inevitably declines. This is what happened with cold email outreach (the irony that the acronym for this is CEO is not lost on me). It is easier than ever to find emails and build the architecture to get them to deliver. So naturally, we get a lot more of them and tune more of them out. For example, I probably get 10-20 emails per day just related to the Startup Dad podcast… and it’s not even that big of a show!

This is unsurprising if you think about the natural progression of commerce and demand. As cave people, we went out and hunted to provide. As civilization settled, trading drove commerce and we had more communal (i.e. scaled) ways of trading our time and effort for needs. Eventually that led to stores and towns and innovation came from door to door sales. The digital age meant we could decouple reach from physical proximity and cold email was born. Cold email scales impressions like billboards along a highway. As the cost of effort has decreased to nearly 0 even with only a 0.1% chance of a win, people will send more emails.

The Problem

As you might imagine from this brief history of cold email, more emails are being sent than ever before. The average office worker receives ~15 cold emails per week.

Today it is the easiest time in history to:

• Find someone’s email address (or many emails at scale).

• Get them to deliver. Despite deliverability getting harder, the people who send emails have much more to lose or gain than Google does. So they continuously outpace the delivery channels.

• Write fake-but-personalized-emails with AI and tools like Clay.

• Start a business. This means there are more companies vying for attention in the first place.

• Find an agency to help you with cold email. The proliferation of agencies that do this work means you’re compounding the sending because businesses who might have never sent cold emails historically are now sending them (e.g. plant sellers).

On top of this you have two major problems:

1. Practitioners are getting lazy and

2. The audience is getting fed up

Problem 1: Laziness

Success in outbound comes down to Message-Market-Fit (MMF). The formula is:

The throughput potency of messaging against an audience 

x 

Volume of send 

x

Channels (i.e. the number of different channels like email, phone, LinkedIn DM, direct mail, etc.)

x

Frequency (do more of this over time and you’ll win with people who previously weren’t interested by wearing them down)

People use the levers they can to win and most people only understand adding volume and channels because it seems like a quick and inexpensive way to drive growth.

They’re all used to throwing more bodies at the problem with SDRs and tools but this has opened the door to people like me (Kellen) outcompeting entire teams via sending 1:many.

What do I mean by this?

Previously, a company might look at a “coin-op” sales process, do some math and assume for every SDR they hire they will get 10-20 meetings per month (on average). They then add incremental reps until they’re reaching their TAM coverage and see diminishing returns. Only one out of ten of those reps is “the best” which means that at most only 10% of the market hears from that person. This leads to a cycle of increasing headcount, people leaders and coordination.

Problem 2: Fed Up Audience

The market’s perception and “how bad it really is” are not synonymous. The irony here is that people say things like “all email is spam” and then are fine with Ramp billboards on buses that apply to <1% of the people who see them [Editor note: guilty].

This doesn’t hurt the advertisers all that much because people still tolerate advertising in the hope that occasionally it’s entertaining or interesting. Bad emails, of which there is an almost infinite supply, make people hate ALL email… even the good ones!

Sending any amount of email will draw some complaints. But don’t equate complaining with “doesn’t” or “can’t work.” It only makes it harder to be successful.

So, woe is me as the cold email sender, right? Not so fast.

It’s not ALL bad in email land…

The good news is that it does still work, otherwise allllll of these people wouldn’t be doing it. It’s also easier than ever to find relevant data AND personalize it against your true ICP without writing horrible and cringeworthy emails.

Let’s take an example:

For less than $1,000 you could personalize emails to 10,000 people in ~1 day. Your only blocker would be warming up your sender address across two-three weeks. Let’s look at how this adds up:

• Pay $99 for 10K email credits with LeadMagic

• Spend $200 on Mailforge for domain warming, addresses and sending capabilities

• Pay $350 for 10K Clay credits to use for website scraping

• Pay ~$100 in OpenAI API credits

…that’s $749.

To execute you’d scrape the websites of the companies using Claygent, layer on AI to tokenize how you can help their business with OpenAI and then send the personalized emails via Mailforge (or another inexpensive tool like Instantly or Smartlead).

A Step-by-Step Guide to Fixing Your Cold Email Strategy

Cold email doesn’t have to be broken. Instead of dwelling on what’s failing [editor note: guilty again, sorry], let’s focus on what actually works. Here’s how to rebuild your approach from the ground up.

Step 1: Develop Deep Audience Understanding

Your market exists whether or not your startup does. Before you write a single email, you need to answer these questions:

• Where does your audience consume content?

• What exact words do they use to describe their problems?

• What is the most acute pain that your product solves?

• What “land mines” will immediately signal you’re not worth their time?

This requires nuanced language and cultural empathy. You can’t AI your way to genuine understanding, even with ChatGPT 5.2. You need to do the research yourself.

A few suggestions:

1. If you have anyone internal who works with customers on a daily basis, go talk to the best reps. Ask:

   1. What pain points do your customers have?

   2. What seems to be trending or top of mind for your customers recently?

   3. What indicates a customer is ready for your product; what sorts of triggers or observable signals might exist?

   4. Do they have to adapt how they communicate to resonate with your specific base of users?

2. Listen to recorded sales calls or go talk to prospective customers directly. Nothing replaces gathering your own understanding and following your own curiosity.

3. Spend time in the watering holes where your customers exist.

   1. Brandon Camhi, COO of Netic, who ran marketing at Rippling, also previously worked for a business selling software to home improvement contractors. He would go browse their Facebook groups to see how they communicated and what they found engaging.

   2. Find top influencer accounts that create content for your prospects - be curious about why those people have drawn so much attention. AI can help analyze these accounts and conduct research on your behalf.

4. Test your hypotheses.

   1. As you get to know your customers there will be a few fundamentally important hypotheses you develop around what makes them tick and why they like your product. Using our home improvement example: maybe you believe they have a lot of FOMO about the competition in their area.

   2. Isolate 2-3 of your strongest held beliefs and go validate them with your sellers or with the customers themselves.

Step 2: Rethink Your Offer (Demos Are Dead)

People are increasingly averse to traditional demo requests. No one has time for it anymore. You need to give them something valuable in exchange for their attention.

Why the demo aversion? Most people don’t like being sold to; especially by salespeople who don’t understand their world. They want to quickly determine if your product solves their specific needs. The rise of product-led growth, self-service tools and free trials have created even more “ick” around demos.

The best offers are:

• Valuable to the user

• In line and unique to your offering

• A strong filter mechanism for who is and is not a good customer fit

Eric from Growth Engine X (an outbound email agency) has nailed this. They offer a free email campaign to prospects before they work together.

• Valuable to the user - if it works, they get leads!

• In line with their offering - it’s exactly what Growth Engine X will do if they end up working with the client.

• A strong filter mechanism - if they can find success with a test campaign up front just imagine how much they’ll accomplish if given more time to ramp on the product, the signals and the users.

For enterprise software businesses that require complex implementations, “trying it out” isn’t as simple. But - there are many creative workarounds:

• Do you have proprietary data you can share in an offer?

• Can you vibe code an app that’s adjacent to the real product?

• Can you pull in one of your partners to provide a faster-to-value solution?

Almost anything is better than a demo; especially for someone not actively looking for one. Taking the time to ideate and test new ways of providing up-front value is a well worthwhile exercise.

A few other examples:

• Company that decorates offices offers to send prospective clients a unique item specific to that company’s vibe

• An accounting software business offers a free audit

• An SEO firm does a comprehensive assessment of your opportunities

Step 3: Think Full-Funnel

Email is much improved with a multi-touch approach. Here are a few additional touches you can leverage:

• Strategic gifting

• LinkedIn connection requests

• Valuable content sharing

• Other touchpoints that build relationships before the ask [editor note: generic flattery about my podcast content doesn’t meet the bar here].

Step 4: Do the Math Before You Burn Your List

Before sending a single email, calculate whether this channel can actually work for your business. Even though cold email has the potential to be a great channel there are some businesses where it just doesn’t make sense.

Calculate your monthly email capacity:

1. Determine how many people are in your total addressable market (TAM)

2. Decide how many times you can sequence them per year (3 is a reasonable baseline)

3. Divide by 12 to get monthly volume

Example:

10,000 people in TAM × 3 sequences per year ÷ 12 months = 2,500 emails per month

Not sure about your TAM size? Figure out how many domains exist in your market, run a sample of 5,000 through enrichment providers, then apply that ratio across the remaining domains. Just be careful to choose a representative sample, not just the best domains.

Calculate your revenue upside:

How much revenue would this channel drive if you got 1 successful engagement out of every 50 sends? Every 100? 500? 1,000? Use these benchmarks to determine what success would need to look like for the revenue upside to justify the investment.

Many businesses try email with TAMs or deal sizes that are simply too small for it to be viable. Or, they get started on cold email and don’t have the proper bar they need to reach for the channel to be successful. Understanding the full picture helps you decide whether the potential return is worth the time, money, and brand reputation cost.

Step 5: Decide What You’re Willing to Risk

Every email you send is a trade. You’re exchanging potential short-term gains for long-term brand equity. Opt-outs never come back and a tarnished brand almost never recovers.

Be honest about what level of success you’re okay with in exchange for potentially burning your email list and/or damaging your brand. If you’re not sure about response rates you can do a small test to get a statistically significant sample size and tighten your estimates.

Step 6: Think Long-Term

If you want to maximize revenue over 5 years, don’t email everyone in your database in month one.

Be strategic with your timing: if your product will improve over time, or if your downstream unit economics will get better, you might want to hold certain segments for later. There’s calculus involved in deciding when to go harder on email and when to hold back.

The businesses that win at cold email aren’t the ones sending the most emails today. They’re the ones building sustainable, self-reinforcing, systems that generate revenue year after year without destroying their market position.

Step 7: Write Punchy, Compelling Copy

You can nail the audience, email warming, offer and strategy… but if your copy is weak then it’s all for naught.

Here are a few tips to writing great copy:

Make it real

If you’re going to mention your company’s value props in an email, get specific. Instead of giving a blanket statement, bring it back to a real world example that identifies and acknowledges the pain you uncovered in your research.

Don’t do this: “We can help you automate HR, payroll and IT”

Do this instead: “Onboard employees in 90 seconds across laptop access, tax documents, 3rd party apps (Slack, Office, etc.), and more”

Stick to ONE simple CTA

Don’t ask the user to do too much. Provide a simple single action point for the user to take.

Do this: “Reply yes if you’d like to try out our calculator.”

Test unique styles of writing

How can you make your emails stand out? It could be a two-liner, adding a GIF (sometimes it doesn’t hurt deliverability as much as you’d think), using unique language/tone, etc.

Anything you can do to disrupt the pattern of someone going through their email and hitting “delete” will dramatically improve your replies.

Editor note: Don’t get overly cute on this one though. Nothing makes me want to spike an email into the junk folder faster than someone giving the appearance of trying too hard.

Cold Outreach Survey Results and Benchmarking

If you’re looking for some benchmark data around cold outreach, specifically email, look no further than the survey we conducted to in-house and agency cold email senders.

Collectively the participants are responsible for ~3-4 million monthly email sends. Here are some highlights:

• Average sequence length: 2.97 emails per sequence

• 84% never or only occasionally use LinkedIn touches in their email sequences (we see this as a huge opportunity for multi-touch campaigns still)

• AI usage: 75% of senders are using AI to write snippets of their emails BUT only 7% are using AI to write the entire email

• Reply rates: average reply rate is 3.2%

• 71% used more than one lead vendor with Apollo as the #1

• 31% used more than one validation vendor with LeadMagic as the top provider
  

Conclusion and Future Outlook

This post represents what the best practices are today but, as with most tactics, we’re in fast-moving waters right now and this could change in six or twelve months. Ideally, we’ve captured enough that is long-lasting and sustainable through changes in the landscape.

As we look to the future, we have some questions and thoughts about what “cutting edge” will look like:

• More and more we will have core architects of these email systems and fewer, low-tier practitioners collecting a paycheck to copy paste failing messaging. The market should be excited about how much this is accelerating - “survival of the fittest.”

• There will always be new entrants that try to win, send awful messages, and make others look bad.

• B2B companies need to grow, it is valuable to test and refine messaging, in our opinion email is “Lindy” and here to stay.

Some questions to ask yourself:

• Is there a future where AI sits on top of the inbox evaluating responses and we end up with two AIs talking to each other and deciding what to filter through to the human? If you look at the number of “AI Inbox Assistants” the answer is: most certainly.

• How do you build a model that improves over time based on the learnings of what you send? The self-analyzing feedback loops that an email agent should be able to execute will make a big difference here.

• How do you build a brand and opt-in email list to nurture over time as email quality and copy eventually commoditizes?

This last question is most interesting and probably the one that cold outreach senders will have to wrestle with the most in the coming years. The solution may lie in Step 2 from above: rethinking your offer. Those who strike a good ratio of value creation to value capture will be more likely to hold people’s attention and interest over a longer period of time.

Subscribe now

Questions you’d like to see me answer? Ask them here.

About 18 months ago I had the pleasure of attending a dinner for a select group of growth leaders in enterprise companies. Attendees were from companies like OpenAI, Atlassian, Postman, Glean, and Workato amongst others. The conversation topic turned to hiring executives at many of these fast-growing companies. That’s when I met Sarah Charlton. She had a specific point of view and a framework for hiring executives at “startups” (broadly defined as any fast-growing, scrappy company - including hers at the time, Asana). She expertly captured and articulated the essence of what I had experienced as an interviewer in the executive hiring process. I was so intrigued by her style and this no-bullshit approach to interviewing that I invited her to write a guest post for this newsletter.

Sarah Charlton is currently an AI company growth advisor. She was previously the Chief Business Officer for a generative video AI company, GM of SMB and Self Serve at Asana, Director of Marketing at Dropbox, has been a GTM advisor to several startups and has led strategy teams around the world.

Today’s newsletter is for founders, executives and all interviewers who are hiring outside of their primary area of expertise.

Sarah covers:

• How she developed her framework with an assist from a NASA leader

• A step-by-step guide to the Depth Method for interviewing executives

• Real examples of this approach applied to her interview experience

• Interviewing pitfalls and areas to watch out for

Becoming a CEO, department head, or a manager of managers is a tough transition because it’s typically the first time you’re running functions that you aren’t excellent in yourself.

This is triply true for founders, who are often in their most senior role ever, running the largest team they ever have, and face large systemic consequences for hiring and building entire functions badly. But it also applies to anyone making this jump. You’ve been promoted because you crushed it in your domain. Now suddenly you’re also responsible for sales when you trained as a marketer (a transition I’ve had to make!), or operations when you’re a product person (editor: been there).

The most overwhelming step is hiring someone to do something you probably couldn’t do even with 20-hour days and some killer AI agents. How do you identify people who have the skills you lack and can get the results you want when you can’t even evaluate their expertise?

This is where most people get burned. Experienced executives are really good at interviewing but might actually be really bad at the job you need them to do. They excel at presentation and storytelling, but may not excel at building. They’ll seem unflappable and all-knowing. They’re impressive and inspiring. At the same time, they might not know the limitations of their own knowledge, where they need support to be successful, or whether they’re a good fit for your company and role.

Most interviewers don’t know how to uncover this because they don’t go particularly deep. A talented storyteller can weave and dodge their way through an interview and the interviewer walks away without a deep sense of whether they can roll up their sleeves and do what is required.

I’ve found a solution to this problem that I learned from an unlikely source: NASA. I call it the Depth Method.

I developed the Depth Method after listening to an interview with Dr. Thomas Zurbuchen, the longest running Associate Administrator for NASA’s Science Mission Directorate. His job was to oversee the launch of NASA missions and make a “go” or “no-go” decision to bring a mission to the launch pad. If a launch goes way off budget and off track he’s the guy who got hauled in front of Congress for questioning (see: James Webb Space Telescope budget overruns 😬). An important job with big technical scope and real-world consequences.

Dr. Zurbuchen had to evaluate whether a mission was ready and whether he could trust what people were saying about their projects, but it wasn’t possible to be an expert in every relevant technical domain. His elegant solution: create a “secret” checklist of the key areas that needed to be rock solid, then listen to mission teams discuss their projects while specifically noting areas they weren’t covering from his checklist. He would ask questions in those gaps and continuously drill down question after question in those areas to determine if they really knew what was going on in their missions and had surfaced and addressed the most critical problems.

When I listened to this podcast, I realized he wasn’t just sharing a good project management technique. Zurbuchen had developed a systematic framework for gathering reliable information in an environment without complete knowledge or deep personal trust relationships. I saw that this framework could help navigate any interpersonal uncertainty and decision-making so I began testing and using it in my day-to-day work.

The most surprising place it proved useful, and where it’s now my go-to approach, is interviewing and hiring senior leaders and executives.

Step-by-Step Guide to the Depth Method of Interviewing

There are three phases to this interview method: Map the Territory (pre-interview prep), Dive Below the Surface (interview), and Surface (evaluate).

Step 1: Map the Territory

If you’re completely naive to a function (engineers hiring for marketing is the most common pain point I hear about from founders), you have to start by acknowledging you don’t know the domain expertise areas for this function, but you do know something about it.

You can start with skill maps common to all executives: are they good speakers who can rep your company (Communications), do they seem like good team adds (Culture), can they manage people well (Management). These areas are critical in every function, though the expectations for each role are different.

But for functional expertise, your first step is to gain an understanding of the function. I like to talk to people who have “graduated” out of it, because they tend to have a higher level and more philosophical understanding of it. A former CMO who is now COO, a former sales AE who now runs all of enterprise, a former customer success leader who is now CCO. Friends who do the job and are not trying to impress you are also good resources as are VC talent partners.

Create a checklist that maps the different areas that someone has to be great at in this particular role.

If you’re interviewing for a Growth leader, your checklist might include

• Growth Loops

• Forecasting

• Experimentation

• User Understanding

• Data Strengths & Limitations

[Editor’s note: I wrote about the Growth Competency model here].

While If you’re interviewing a Product executive, you can include:

• Data Fluency

• Product Strategy

• Product Vision & Roadmapping

• Delivering Business Outcomes

• Customer Insight Development

• Product Execution (specification, delivery, quality)

• Influencing

To support the creation of the checklist you’re going to want to ensure that the entire interview team is collaborating on its development. Because executives need to excel across multiple competencies, each individual interviewer should have their own set of ~3-5 areas that they’re going to go deep on.

[Editor’s note: For an in-depth guide to interview preparation you can read this].

Step 2: Go Below the Surface

The key to conducting the interview is getting below the surface. As the name implies you might only get a surface-level view of the executive unless you dig deeper. To do this you need to continue digging down with each subsequent question.

Let’s say that you’re interviewing a Head of Growth and one of the areas you want to explore is building growth models. You might start with a question like:

“Tell me about your experience building a new growth model at [company X].”

Then they might say something like:

“Well, it’s really important that you start with a solid foundation of data.”

You’d follow-up with:

“Oh, that’s great – how do you go about establishing that foundation?”

They’d say:

“We had to instrument our entire product and send that data into our data warehouse in addition to the product analytics tool.”

You’d follow up with:

“Ahh, getting instrumentation right can be really challenging. How did you do it? Which analytics tool did you choose, why did you choose it?”

…and so on.

The key is to keep drilling deeper with more specific questions until you reach a place where the candidate stumbles or can’t go any deeper.

Alternatively, you might also find that a candidate will launch into a detailed explanation from the first question. Especially in a long answer, listen carefully to what they’re saying and not saying. For example, if they leave out the process of instrumentation, you can choose to ignore the rest of their answer and dig into this area after they finish their initial explanation. Be conversational, but don’t let conversational distraction dissuade you from what your checklist requires you to evaluate.

What you’re looking for here is twofold:

1. Knowledge boundaries

2. Expertise signals

Knowledge boundaries are where the sparkle starts to come off the diamond. Through your continued, specific questions you run into the edge of their expertise. Here you’re evaluating not just their knowledge but also honesty about their own limitations. Do they acknowledge their limit, and if so do they act defensive? What happens at this critical point will show you how they perform at any knowledge limit in the future.

Expertise signals are areas where they light up and are able to continue elaborating for days. As Dr. Zurbuchen says,

“You know, there’s some people that have never missed. I’ve never got to the end of my questioning. There’s some people that were two questions deep and I got them. And it’s like, OK, they are a two question kind of person. You know, there’s the other one that I’ve never found the end. After six questions, he’s still going. And I’m like, OK, don’t worry about him.”

This might look something like:

You: “Ahh, getting instrumentation right can be really challenging. How did you do it?”

Them: “It was challenging, but we figured it out. First we spec’d out all of the events we wanted to fire based on the critical paths through the product. We built it out in a giant table and specified the syntax for the event call so we could make it super easy and templatized for our engineers…”

You’ll notice in the above that this person clearly knows what they’re talking about and has done the actual work. First, they’re excited to talk about it and second, they can articulate with great detail the process they went through. And with every follow up, you get more information and depth. After a few more questions you can just stop: they’ve got what you need in that area.

A Real Example: Customer Success Leader

I was once on the interview panel for a Customer Success exec. I was the GM and Head of Growth for the self-service and SMB customers. Given my role, my checklist included:

• Management: standard in any functional leader

• Customer love: obviously expected in this role, I wanted to see someone who knew and named their customers, their roles, and pain points

• White-glove large customer management: what’s their strategy and techniques for handling the whales?

• Long tail, small dollar value customer management: did they also have strategies and plans for handling my book of more than 100k customers?

• Technical stack knowledge: did they know what systems and tools they and their teams needed to be successful

In one interview, I asked a candidate how they handled the long tail. Here’s the exchange:

Me: “How do you handle long-tail customers, making sure they’re successful without spending too much money?”

Them: “It’s all about A/B testing.”

Me: Oooh, I’m a growth person, this is catnip for me… “I love A/B testing. Tell me about a specific A/B test you ran and what you’ve learned about small customers?”

Them: “We ran some trials on email that showed it helped change behavior at scale.”

Me: pushing another level deeper… “Wow, I love email. It’s always been really mixed for me. Tell me about an email test you ran that was really impactful?”

Them: Long pause… followed by a very small experiment from ~4 years prior.

What I uncovered was that this candidate knew that A/B testing was an important thing to do, they knew it worked at scale… but they actually hadn’t done it. Maybe they’d overseen a team doing it, but clearly without paying that much attention to what was going on. In this particular area they would need a lot of help.

It happens that I know a lot about A/B testing given my background in growth so this area was easy to dig into. Let’s take an area I knew less about but was equally important: white-glove customer management. Make sure to probe especially in areas where you don’t know the answer: digging deeper not only assesses the candidate, it also gives you an opportunity to learn something meaningful and tangible from the conversation.

Here’s how that line of questioning went with another candidate:

Me: “Beyond our huge book of small customers we have some very large customers that are very important to us. What does it look like to do a great job with customers like this?”

Them: “You need to build the support they require to be super successful into the package so that they pay for it at the beginning of the contract, and you can afford to provide them with the higher-level of engagement they need to be successful and happy with you as a vendor.”

Me: “That’s really interesting; how do you make sure they’re willing to pay enough?”

Them: “Well, most enterprises expect there to be onboarding and installation fees. But sometimes they aren’t willing to pay the whole value of the package. In this case, you can take some of the charges you impose on mid-market customers to help subsidize operations at larger customers. But you can also…”

Me: “What happens at the sale if they’re not willing to pay for support but you know that they actually need it?”

Them: (Super detailed answer about what to do in this situation and how to make it work.)

So here I am not knowing much about implementation of a white glove customer engagement program but I’m following my own curiosity and staying on that specific topic. I kept going deeper and deeper, down to as many levels as possible. At the end of that part of the interview I had learned something about how to run a white-glove team with big name customers. After three of four of these, I had a good sense of the top candidates even in this area where I had less experience, and I had learned an incredible amount about building a great white-glove customer success organization.

The best part about this approach is that even though the areas you’re probing are things someone should know to do the job well, the specific line of questioning looks different every time. They can’t rehearse for it.

You need at least 45-60 minutes for this method, because otherwise you can’t cover your entire checklist while going really deep in a few areas. There will be pressure to do shorter interviews for candidate experience. Push back on this. Hiring an executive is one of the most important things you can do and also most fraught with false positives. If ~50% of exec hiring fails you owe it to yourself to spend the time up front.

Step 3: Surface

After the interview, you need to do three things: assess how deep their knowledge actually goes, identify any gaps where they’ll need support, and compare notes with other interviewers.

During and right after your interview, you can note how deep the candidate could go in each of your critical areas. Make sure you leverage your actual conversation content, not their resume and supposed experience here.

Remember that the goal isn’t to find someone who knows everything, but to identify their gaps where they’ll need strong support and to have a sense of how they react when coming up against their own limitations. A sales leader might be brilliant at hiring and training but weak on sales operations. If they are aware of it and can work around it, there’s no problem. If they’re sensitive or arrogant around their limitations this will be a harder person to take on.

Most importantly, compare insights across your interview panel. Other interviewers might have been impressed by the candidate’s presence and missed the shallow spots you uncovered. They may have gone deep and gotten signals in areas that conflict with your insights. Compare the checklist, share specifics, and then make a call.

Common Pitfalls and Areas to Watch Out For

There are four primary traps you can fall into when interviewing executive hires:

1. Being swayed by (over)confident delivery or resume

2. Conversation redirects

3. Time

4. Preparation

Avoid being swayed by (over)confident delivery

You can’t let someone’s smooth delivery or paper resume trick you into thinking that they know something. Confidence is unfortunately largely uncoupled from actually knowing things at the executive level. Leveraging the Depth Method helps you push through this because you’re increasing the required specificity and depth until you have a personal assessment of their knowledge and capability boundaries.

Being aware of executives who redirect the conversation

If you come across an interviewing pro - especially true of people who have done thorough press training - they can try to turn the interview questions around to you when they hit their limits. It’s a jujitsu move: you ask something and they’ll flip it and say “Yeah, that is really important, I’m fascinated by [you, your company’s] approach to that?”

Watch out for this; it’s not a huge red flag, but you’re letting them dictate the terms of the conversation and you’re losing the thread of what you’re trying to assess and learn. Stay friendly and interested, but save your opinions and the conversational back-and-forth for after you’ve finished assessing capabilities.

Giving yourself enough time

As I mentioned above, you need more than a standard 30-minute conversation to run this interview method. Without at least 45-60 minutes you won’t have time to go really deep and cover the 3-5 competencies that you need to get to. But sometimes you’ll hear arguments for shorter interviews for the sake of candidate experience. Too bad. Push back on this when it’s a senior critical hire.

Preparation

You really have to thoughtfully create your checklist and interview plan. Winging it leads to the common strategic pitfall I call “SQUIRREL!”

Have you ever taken a dog for a walk, and then a squirrel crosses your path? If the dog’s on a leash you’re often ok after a brief struggle, but off-leash the dog is gone chasing the squirrel up trees and around the park a long time, while you’re left standing there wondering where things went wrong.

Prepare for squirrels by organizing your checklist and interview plan so you can stay on track and avoid exciting distractions.

Final Thoughts

The Depth Method has become my default approach for any senior hire, but it’s particularly crucial when you’re scaling into new functions. I’ve used it to hire growth leaders, partnerships execs, and even technical roles where I needed to evaluate architectural thinking despite not being an engineer myself.

What I’ve learned is that the best leaders actually enjoy this process. They light up when you ask them to go deep on something they genuinely understand and love. The ones who get defensive or try to redirect? That tells you something important about how they’ll be on the job as well.

The method also scales your own learning as a leader. Every time I’ve used it, I’ve walked away understanding the function better than before. That customer success interview where I learned about white-glove operations? That knowledge directly improved how we later structured our enterprise support packages six months later.

The most surprising aspect of this approach is it has made me better at assessing expertise across the board. Once you default to wide-ranging questions with strategic deep-diving, you start gaining insights in every conversation, from vendor pitches to board meetings. You develop an instinct for when someone actually knows their stuff versus when they’re performing expertise, and you can choose your own reactions accordingly.

The stakes are too high to wing executive hiring. A bad senior hire doesn’t just fail: they take six months to fail while burning cash and potentially damaging your team’s confidence in both you and the company. The Depth Method gives you a systematic way to cut through the presentation skills and get to the substance.

Most importantly, it gives you confidence to hire outside your domain expertise. You don’t need to become a marketing expert to hire a great CMO. You just need to map the territory, go deep, and trust the process.

Subscribe now

Questions you’d like to see me answer? Ask them here.

The explosion of AI prototyping tools has created a problem: there’s no comprehensive guide to help you figure out which one fits your needs. So I approached Sachin Rekhi, who has experimented deeply with all of them and asked for his help with this newsletter. This is a complete reimagining of how product teams should work. If you’re involved in product development in any capacity and you don’t build the prototyping skill… you’ll be left behind.

Sachin is a renowned product leader and entrepreneur, best known as the co-founder and CEO of Notejoy, a fast and focused notes app for individuals and teams. Previously, he served as a Head of Product at LinkedIn, where he incubated and scaled LinkedIn Sales Navigator to over $200M ARR. Sachin’s twenty-year career has spanned founding startups (Connected and Anywhere.FM), leading game-changing products at Microsoft and imeem, and advising a wide array of high-growth companies. He’s also widely recognized for sharing actionable product frameworks through essays, courses, and videos, helping thousands master the craft of product management and product-market fit.

Check out his most popular course on Reforge, AI Productivity, which teaches how leading product managers use AI to become faster, smarter, and gain super powers beyond their traditional role. It has by far the deepest and most actionable curriculum I’ve seen to date on AI prototyping for product teams.

🎉 SPONSOR ALERT 🎉

Today’s newsletter is sponsored by Reforge Build! Are you ready to stop wasting time building prototypes from scratch? If you’re part of a product team with established designs and real customers, Reforge Build is the AI-powered prototyping tool you’ve been waiting for. It instantly understands your product’s design guidelines, your customer feedback, and your strategy so you get real, context-aware prototypes that actually look and feel like your brand.

Collaborate faster, validate ideas in hours (not weeks), and enjoy powerful features built for high-velocity product teams. Try Reforge Build today—get 1 month free with code BUILD!

Try Reforge Build

In today’s article Sachin goes deep on the future of AI-driven product management: prototyping. He has quietly become one of the most sophisticated practitioners of AI prototyping in the product world.

What makes Sachin’s approach different is that he’s moved far beyond basic “prompt-to-mockup” workflows. He’s developed techniques for maintaining design consistency across prototypes, strategies for divergent exploration that generate multiple solution paths, and methods for creating functional prototypes that customers can actually use and provide behavioral feedback on. Most importantly, he’s figured out how to integrate AI prototyping into product development workflows in ways that fundamentally change how teams prioritize and build features.

Read today’s newsletter to learn:

• Why AI prototyping represents a revolutionary shift in how we build products

• The specific tools we recommend (and why), including detailed breakdowns of 14 tools and who and what they are best suited for

• How to choose the right prototyping tool for you and your organization

If you enjoy this newsletter, let us know in the comments and we’ll do a follow-up primer on how to get started with AI prototyping.

I remember starting my career at Microsoft writing 30+ page specifications, complete with detailed user scenarios, error conditions, and screen copy. Design tools matured and we moved to high-fidelity design mockups as the primary medium of communication amongst our product team and with stakeholders. As they say, a picture is worth a thousand words, and so these design mockups allowed us to simultaneously convey far greater detail while enabling us to significantly shorten our specs.

AI prototypes continue that evolution, enabling us to convey our desired user experience at an even higher fidelity than static mockups. By adding interactions, real-world data, and even functionality, our prototypes come to life, removing even more of the lossy-ness between our intended experiences and what we can convey.

Now, prototypes of course aren’t new, but what AI prototyping did was make it far easier and faster than ever before to put together a highly interactive prototype. No longer is prototyping a time-consuming task limited to highly trained designers. It’s no wonder that product teams are already starting to embrace prototypes as their new primary design artifact.

Need evidence? Here it is:

What gets me most excited about AI prototyping is its ability to fundamentally reshape how teams prioritize what to build.

If you look at the way most traditional product teams prioritize their roadmap they start by prioritizing specific customer problems that need to be addressed. Once those problems are on the roadmap they then get into designing a potential solution to those problems and then build and launch it.

But the very best companies like Apple, Stripe, and Amazon don’t actually work this way. They instead leverage a process known as product shaping. The idea here is they come up with a bunch of potential customer problems to solve and then immediately start prototyping solutions to those problems. They then prioritize what to put on their roadmap based on the best problem/solution pairs. This way they know that they are not only solving an acute problem for a customer, but also solving it with a solution that is likely to resonate strongly as well.

Take, for example, Apple. Jony Ive had been prototyping a tablet capacitive touch interface for years. When he finally showed it to Steve Jobs, Steve loved it but thought it would be far more useful for a phone than a tablet. So they shelved the tablet project, built a phone instead, and went on to change the world.

The question then becomes why haven’t all companies historically taken a prototype-driven approach like Apple? It turns out it’s incredibly expensive to run a prototyping lab that throws out and never ships most of its prototypes, making it very difficult for traditional organizations to justify.

But AI prototyping completely changes that. Now AI prototypes are incredibly cheap to build, enabling any team to leverage the product shaping techniques of the very best companies.

This prototype-driven prioritization approach is already taking hold at companies like Anthropic, where they use a prototype -> dogfood -> prioritize -> launch loop for deciding what to build next for Claude Code.

I’m convinced AI prototypes and prototype-driven prioritization will soon become the default way that teams build products going forward.

The AI Prototyping Tools Landscape

Now that you are convinced that AI prototyping is the future [editor note: you are convinced, right?!] the first step to getting started is picking an appropriate AI prototyping tool. This has become a lot harder because the market has recently exploded into a dizzying array of tools. This makes tool selection far from a straightforward process. So I wanted to give you a lay of the land to help you navigate your search.

Let’s start with a high-level look at the prototyping tools spectrum. There are three main categories of tools that people often consider when looking to build an AI prototype:

1. AI App Builders

2. AI Prototyping Tools

3. AI Coding Tools

Let’s walk through each category and my take on it.

AI App Builders — Recommended

AI app builders are the most mature category of tools, including popular tools like Replit, Bolt, v0, and Lovable. To give you a sense of their scale, these tools collectively are already earning over $500M in revenue. What’s powerful about this category of tools is they allow you to develop entire applications purely from prompts. While they are certainly used for prototyping, they are also widely used for building internal applications and even launching fully vibe coded apps.

The maturity of these tools is what makes this category an instant recommendation for prototyping. They have robust capabilities, user-friendly interfaces, and scale to prove they aren’t going anywhere. The only downside of this category is that they aren’t purpose-built for the prototyping use case, like the next category we’ll explore.

AI Prototyping Tools — Recommended

AI prototyping tools are an emerging category of tools focused squarely on product teams that are looking to prototype & validate their product ideas before handing it off to engineering. Popular tools include Reforge Build (sponsor), Figma Make, Magic Patterns, Reforge Build, and Alloy. Since these tools are purpose-built for the prototyping scenario instead of broader scenarios like fully vibe coded apps, they introduce unique capabilities tailored for prototyping that I’ve found to be extremely valuable for the use case.

Product teams are going to love these unique prototyping capabilities, including automatically importing your existing design, designing multiple variants for a single experience, and simplifying validation of prototypes with customers.

While I think all product teams will ultimately move to tools in this category, the main downside is most of these tools are still early and don’t yet have the same robust capabilities and proven track record of the AI app builders. But I’m betting on them getting there.

AI Coding Tools — Upgrade Pick

Engineering teams have fully embraced a broad set of AI coding tools to significantly accelerate writing production code. Popular tools include Cursor, Claude Code, and OpenAI Codex. While this category of tools has existed for awhile, they have more recently introduced agentic capabilities that now enable you to develop full applications from just a prompt. It’s these agentic capabilities that led to teams using them for prototyping.

I don’t recommend product teams starting with this category of tools because they are far more technical and can be intimidating for non-technical team members.

However, there are cases where this becomes the right category of tools to upgrade into. As you look to put together really robust functional prototypes you might start to run into the limitations of the other categories of tools. AI coding tools can handle the most sophisticated requirements as well as handle the largest code bases.

These tools also integrate with your engineering team’s code base, which can be helpful when you are trying to prototype on top of that existing code base or take advantage of design components your engineering team has already built.

Recommended AI App Builders

Let’s now walk through each category of tools and I’ll provide you my opinionated guide on what each tool is best suited for.

Let’s start with the well established AI app builders category.

Bolt

Best for: Rapid iteration and speed

Bolt is a full-featured app builder that can build both front-end experiences as well as back-end functionality. What sets Bolt apart is its unique WebContainer technology that enables it to efficiently run front-end and back-end code directly in your browser. The upside of this technology is that Bolt is generally faster than the other app builders and prototyping tools.

When using these tools, you’ll find yourself spending a lot of time waiting between prompts. As an impatient person, I relish the speed Bolt offers. And it’s why Bolt is one of my daily preferred prototyping tools.

Here’s a live demo of me building a prototype with Bolt:

V0

Best for: Front-end UI experiences

v0 is another popular app builder that has been on the market for awhile. What sets v0 apart is its focus on building beautiful front-end UI experiences quickly and easily.

If you are looking to quickly iterate on some UI concepts, v0 does that fast and elegantly. The reality of our work as product builders is that we are often trying to figure out what the right front-end experience is and v0 is a fantastic tool for doing just that.

Editor Note - here’s an example of a new onboarding experience for Firefox that I prototyped with v0.

Replit

Best for: Robust full-stack experiences

Replit offers by far the most robust front-end and back-end app building capabilities. Many people use Replit to build fairly rich internal applications if not full external applications as well. Replit also offers you a lot of control over your technology choices when building out your application.

While these robust capabilities can be valuable, they do come with a few downsides. The first is speed. Replit is one of the slower tools since it by default builds out both front-end and back-end capabilities, which you may not need if you are just building a front-end prototype. The code complexity is also higher, which increases the chance of introducing bugs as you iterate on your prototype.

So if you go with Replit, make sure you need the richer capabilities, otherwise they may be overkill.

Editor note: here’s a podcast transcript to tweet-thread generator I built in a couple of turns with Replit:

Lovable

Best for: Non-technical users

Lovable made a name for itself being the best app builder tool for non-technical users. In their recent ads they show off how even kids now can vibe code their own apps with Lovable. I myself have seen business professionals create pretty compelling experiences without any technical understanding.

They accomplish this by making smart technical choices on your behalf and hiding the code by default. So if you and your team find yourself intimidated by the other tools, Lovable is worth checking out.

Our pal Elena Verna built her personal website with Lovable:

Base44

Best for: Non-technical users

Base44 is one of the first AI app builders to be acquired, in this case by Wix, a traditional website builder. The Wix team is aggressively marketing Base44 through YouTube ads and more.

Their positioning is quite similar to Lovable in that they are squarely focused on non-technical audiences, accomplishing this with a full suite of built-in and default back-end services, including database, authentication, email sending, and more.

What I also like is they offer up-front styling options, helping to ensure your app by default looks different from the generic output you often get from these tools.

Here’s a <5 minute vibe-coded prototype for a SMB metrics dashboard. A nice touch is that they give you product suggestions at the end for adding new features:

Google Firebase Studio

Best for: Full-stack experiences for Google customers

Google Firebase Studio is one of Google’s app builders. It’s similar in scope to Replit in that it enables you to build robust full-stack front-end and back-end applications, relying on Firebase for backend support. It also has the same trade-offs previously mentioned with Replit.

If you are already a Google or Firebase customer, you’ll find these tools incredibly easy to adopt and I expect that’s where they’ll see the most demand.

Here’s a personal finance tracker - notice how it gives you the components so you can edit them prior to developing the prototype:

Google AI Studio

Best for: Building Google AI apps

Google AI Studio is a newer app builder from Google that differentiates based on its focus on helping you build AI apps using Google’s AI technologies, including Gemini for text, Nano Banana for image, and Veo for video generation. Though it’s less focused on helping you build full-stack applications.

If you have been looking to leverage Google’s AI models, this is going to be the fastest platform for you to start experimenting with them.

Here’s a prototype of a tool that builds email marketing campaigns from a single problem; built in <2 minutes:

Microsoft GitHub Spark

Best for: Microsoft customers

GitHub Spark is Microsoft’s recently introduced app builder. It’s fairly new so it’s still difficult to assess its full capabilities. But its appeal will certainly be amongst Microsoft and GitHub customers who are already deeply embedded in the Microsoft stack.

Sadly, Spark is only available to Copilot Pro+ and Enterprise users. Womp womp.

Recommended AI Prototyping Tools

Let’s now move on to the AI prototyping tools, which as I mentioned, are a newer generation of tools that I’m very excited about because they are purpose-built for the product prototyping use case as opposed to broader app building use cases.

Reforge Build [newsletter sponsor!]

Best for: Product teams focused on full-stack experiences

Reforge Build is a recent entrant into the prototyping tools space, offering a prototyping solution squarely focused on product teams.

[Editor note: it just launched out of beta today November 12, 2025 and we will say lots of nice things about it here because 1) It’s the sponsor of this newsletter, 2) Brian Balfour is our friend and 3) It’s objectively awesome for prototyping].

They offer many of the same product-focused prototyping capabilities as other prototyping tools we’ll discuss below: including a browser extension to import your existing product design, custom instructions for conveying your design system, as well as generating multiple variations for a given design.

One of the unique capabilities and the true differentiator for Reforge Build is Projects and their ability to uniquely incorporate context into your prototype. You can easily add any project-related files to the project, including product strategy docs, product specs, meeting transcripts related to the feature, design guidelines and more. This context is automatically fed into every prompt you then subsequently write to construct your prototype. So many of these prototyping tools exist in a silo separate from the product-related context on the feature you are building, so it’s great to see Reforge Build bring that together into the fold.

As an example of how this comes to life: once it knows everything about your upcoming product via the docs mentioned above it will then get small, important details correct like writing copy, talking about features/benefits, etc. because it has the context.

I also like that Reforge Build has embraced full-stack applications, incorporating the latest Claude Code coding agents to let you build fairly sophisticated prototypes.

The Variants functionality helps you generate, explore and compare multiple design directions to find the best path (which isn’t always the fastest one).

In addition to Reforge Build, Reforge offers Reforge Research and Reforge Insights, which make it easy for you to run surveys and customer interviews from the prototype you’ve built in Build. When you get that feedback you can put it back into Reforge Build for improved prototypes. It’s a wonderful loop! While still early, Reforge Build offers a really rich set of capabilities truly built for product team prototypes.

I used the Variants feature to generate 4 alternative setup experiences for SplashSports contest creation:

Magic Patterns

Best for: Product teams focused on front-end experiences

Magic Patterns is one of the earliest tools in this category, pioneering many of the unique capabilities purpose-built for product teams.

First, they were early to the idea of offering a browser extension so that you can easily extract your existing product design from your live app or website. This significantly reduces the effort to establish a baseline prototype from which you can then build novel features on top of.

They were also early to the idea of enabling you to build a component library directly in Magic Patterns that would then get used to assemble your next prototype, ensuring you have very consistent designs. Along the same lines, they also let you define custom instructions that let you convey additional elements of your design system, like typography, colors, and more.

They also offer an Inspiration command (similar to Reforge Build’s Variants), which will automatically come up with four completely different design directions for your prototype. I find this incredibly helpful as a way to leverage AI as a brainstorming partner when I’m early in the prototyping phase.

Another prototyping-oriented feature offered is Templates, which are pre-built prototypes that can be shared and serve as a baseline for exploring new features or redesigns.

You can ultimately even export your prototype to Figma to polish the designs before handing them off to engineering.

Since Magic Patterns is squarely focused on the prototyping scenario, they are light on back-end support, which can be a limitation. But their distinctive capabilities optimized for product prototyping make them uniquely well suited for the task.

Here’s a calorie tracker prototype it built in <1 minute:

Figma Make

Best for: Figma Designers

Figma Make is Figma’s answer to AI prototyping. While late to the market, Figma’s market dominance amongst designers gives them a natural audience to optimize for along with a built-in user base.

Their initial value proposition for designers was flawless import of their existing Figma design that serves as the starting point for your interactive prototype. While many tools are now offering Figma import, Figma’s own importer will always be perceived as the best with the tightest integration.

Beyond Figma import, Figma just announced upcoming support for leveraging your existing Figma design system. Figma Make will take your design system components, import them into Figma Make as re-usable React components, and then leverage those components when putting together new prototypes. One of the biggest challenges in prototyping is ensuring your prototypes actually look like your product and design system import is Figma’s elegant solution to this problem.

If your team is already heavily using Figma as well as heavily invested in an existing Figma design system, you’ll find yourself right at home with Figma Make.

Here’s a generic, 3-step onboarding flow in less than <2 minutes:

Alloy

Best for: Matching your existing product design

Alloy is the most recent entrant in the prototyping tool space. Also optimized for product teams, their core differentiation is their ability to leverage a browser extension to import your existing app or site.

While others also offer this capability, they’ve done a really nice job of trying to match your existing product exactly by leveraging a custom algorithm for extracting and recreating your product’s HTML and CSS. This can be a big timesaver since so often we want to first recreate our product experience as a prototype and then start building new features or redesigns on top of it.

Given Alloy’s new-ness, it isn’t nearly as robust as the other prototyping tools, but I suspect they are working on closing those gaps.

I had it redo the SplashSports contest homepage to emphasize a particular product type. It took awhile but gave me something that matched the design system perfectly:

Recommended AI Coding Tools

Let’s now touch on the final category of tools, which are AI coding tools. As I mentioned, these are rarely my first pick for prototyping. However, as your prototype gets large in scale, you might find it helpful to upgrade to these tools since they do a far better job of handling large codebases. If you are also looking to integrate with your existing code base, to leverage existing designed components, then these tools make it easy for you to leverage that existing code base.

While there are dozens of tools in the AI coding tools space, I wanted to just briefly mention the two most popular ones I see being used for prototyping.

Cursor

Best for: Technical users

Cursor is an AI-powered Integrated Development Environment (IDE) that is popular amongst software engineering teams. It enables full-stack development of any kind of application.

In some ways, it will look similar to the other prototyping tools in that you’ll have a chat box where you can prompt whatever you’d like to build. The main view will show the code that’s being generated by the coding agent.

There are however a couple of key differences. First, you won’t see a preview of your application within Cursor. Instead you’ll open your application in your browser tab and refresh it whenever you update your code. The second big difference is there is not built-in support for publishing your prototype. So you’ll need to use Netlify, Vercel, or another deployment solution to launch your prototype publicly.

These limitations do make this tool far more technical to use than the other prototyping tools. But since it’s the same tool your engineers are using for writing production code, there is no ceiling on what you could potentially prototype with it.

Here’s the shell of a meal planning application that I built with Cursor - it took longer and I needed to install some files locally on my machine - but this is a fully functional prototype:

Claude Code

Best for: The most technical users

Claude Code is a command-line AI coding tool that gives you the fullest power of frontier coding models, enabling you to easily handle millions of lines of code.

However, the interface is a command-line tool that you run in terminal to provide your prompts. There are ways to integrate Claude Code into other IDE’s like VS Code, which is what I do, to make the interface more friendly.

Once again, if you are technical, need to integrate with your existing code base, or have an incredibly sophisticated prototype, you may need to upgrade to Claude Code. But it’s not where I recommend most people start.

Editor’s note: here’s a fully functioning application I built for the Startup Dad podcast to help me automate some post-production tasks. It started with a prototype! This took a awhile but is really robust:

How to Choose Your Prototyping Tool

Now that we have comprehensively explored the full spectrum of prototyping tools available on the market, from AI app builders to prototyping tools to AI coding tools, I wanted to provide some additional guidance on how your team can go about narrowing the options to the ones that will likely be a fit for your organization.

The first aspect you should assess is your team’s technical ability. If you expect most of the folks who will be building prototypes to be non-technical, pick the solutions optimized for that audience, including all the prototyping tools and the AI app builders focused on non-technical users. However, if your team is more technical, you’ll have the full range of tools at your disposal.

The second consideration is how much you expect your prototypes to be front-end focused vs full-stack applications. If you’re focused on prototyping front-end experiences, you’ll have the full suite of options available to you. But if you are focused on full-stack prototypes, it’ll narrow the selection of tools available to you.

The third consideration is whether you have an existing heavy investment in design system components that your team is looking to leverage in your prototyping workflow. If those design system components are built in Figma, then Figma Make will likely be your best tool. If those design components are built as React components, then the AI coding tools will offer the easiest ability to leverage them.

Once you’ve narrowed the field down, pick up to 3 tools that you want to experiment with. Come up with a prototyping use case and then try building that exact same prototype in all three tools. You’ll quickly observe the differences and be able to determine what feels like a better fit for you from there.

Wrapping Up

I hope you now understand the important benefits of AI prototyping, why it’s the future of product development, and how to get started by taking the first step of deciding what AI prototyping tool will best suit your needs.

While AI prototyping is an exciting new world it’s important to end with a few caveats and myths that you should be aware of:

Myth #1: Prototyping Speeds Up Delivery

It’s easy to see how you might start to believe this: we can now build a prototype super fast, hand the context over to engineering, and they can crank out a production-ready version.

That’s not the right way to use or think about prototypes. The code produced in a “vibe-coding” prototype won’t be production ready. It’s not a delivery tool, it’s a product discovery tool. We can learn and validate a lot faster, but this won’t necessarily impact product delivery speed.

Myth #2: Prototyping Visualizes My Idea Faster

While you certainly can go from idea to workable version of that idea incredibly fast you’ll be missing one of the best benefits of these new tools: divergence and variety. You can use this to come up with multiple ideas and directions and then choose which one makes the most sense.

Myth #3: Prototypes Replace PRDs & Mockups

I don’t think this is the way that we’re going to go. The prototype doesn’t actually answer the question of whether the product or feature fits into our product strategy. You can’t answer that from a prototype alone. What will likely happen here is that PRDs will get smaller and more focused; they’ll get replaced by Product Briefs.

When it comes to mockups, I don’t think they’ll replace them either. Prototypes are in a mid-fidelity place (they have interactions, but they’re not pixel-perfect). If we need pixel-perfect mockups we’ll have to build them. You can export these prototypes and then clean them up in your tool of choice (like Figma).

Thanks for joining us on our journey today! If you liked what you read and found it helpful please let us know in the comments. If there’s enough interest we’ll prioritize a follow-up to this that walks through a 5-step process to getting started AI prototyping at your company.

Thanks for reading,

Sachin (and Adam)

Collaborate faster, validate ideas in hours (not weeks), and enjoy powerful features built for high-velocity product teams. Try Reforge Build today—get 1 month free with code BUILD!

Try Reforge Build

Questions? Ask them here.

Share

🎃 Happy Halloween everyone 🎃

For the last two years I’ve written about how a group of volunteer Dads applied operational rigor and growth strategies to a local pumpkin patch to transform our earnings. Last year I ended with this:

When we started this year’s patch I was convinced that all of the optimizations we had done in years past would make it hard to grow again like we had the past two years.

Going into this year I was equally convinced that we’d reached our maxima. How could we possibly top the numbers from last year? There’s only so much (pumpkin) juice to squeeze out of a small-market, school-fundraiser. Combine that with Halloween landing on a Friday which meant we’d be closing days ahead of time (our historical plan has always been to do teardown on a weekend for volunteer availability). Topping last year’s $54k seemed incredibly daunting if not impossible.

Once again, I was wrong. We did it again.

You can read about 2023 here and 2024 here.

This year the real story isn’t about new tactics—it’s about what happens behind the scenes.

But first, what is the Wildwood Dads Club?

The Wildwood Dads Club non-profit organization is a storied institution that dates back 101 years to the founding of Wildwood Elementary School.

Our mission is to drive fundraising and community building for Wildwood alongside the Wildwood Parents Club. In our district, ~37% of school funding comes from local funding sources: parcel taxes, our education foundation nonprofit, and parent / community support. Throughout the year the Dad’s Club raises ~$15-30k+ for the school which is then spent on capital improvements, infrastructure projects, and classroom support. Things like sports equipment for PE, classroom technology, new computer lab equipment, AV system upgrades, and more.

One of our largest fundraisers is the Pumpkin Patch.

For more than 13 years our seasonal, non-profit pumpkin patch was stuck at around $30,000 in revenue and $15-20k in net income. Then starting in 2022 revenue has looked like this:

2022: $38,000

2023: $42,000

2024: $54,000

And this year, in 2025, our patch revenue was $69,100 a 27% YoY increase!

Let’s start with some big achievements:

• Total revenue: $69,100

• AOV: $45.17 (We reversed the 2024 decline in AOV and once again did not change prices for a 3rd year in a row)

• Total Purchases: 1,530

• Pumpkins Sold: ~6,891

• Highest single-day sales ever: $8,002 (Sunday before Halloween) despite heavy rain

What We Kept Doing (And Why It Still Works)

Before we jump into this year’s lessons, it’s important to recognize that the last several years involved measuring, refining and optimizing our tactics. We’re building off of a very solid foundation.

Signage and Local Visibility

We maintained the refreshed signage from 2024 that emphasized a “Support Our Schools” message along with the side-of-the-road Giant Jack signage.

This year we were able to secure a win by getting special permission from the city to place signs back on public right-of-ways. We also put the signs out ahead of opening weekend to start to generate interest. Both were successful! It turns out the city decided they had better things to do than take down the signage of a volunteer fundraiser hanging on some trees. Victory!

Why does this work?

Local signage remains a top-3 traffic driver to the patch. In our in-person checkout “HDYHAU” (i.e. asking customers how they found out about the patch) a large percentage of people cited that they were driving by and saw a sign.

Many also added that they wanted to see their money stay locally in the community which underscores the importance of the “Support Our Schools” messaging.

Giant Jack remains the main attraction and top photo-taking opportunity. Our staged “photo shoot” location is a hit with the crowds. We even saw people setup their own mini shoot and bring some props:

Continued SEO and AEO Dominance

We maintained our Google Business Listing with updated hours, photos, etc. and continued our review generation strategy. We now have 34, 5-star reviews on Google (+11 YoY) and 11 on Yelp (+1 YoY).

Even better, when you search for “top pumpkin patches near me” we are in the #1 position in Google and Yelp.

Our AEO listings within ChatGPT and Perplexity got a bit more crowded this year though we were still listed in the top 3. I asked Perplexity why we weren’t in the top spot and this is what it told me:

Apparently offering fewer experiences takes us down a couple notches. Perhaps we’ll see if we can get a goat for next year.

Why does this work?

We all know how powerful Google is in capturing high-intent searches and it remains a top traffic driver (along with WoM and signage). This year we added a clear CTA at the patch to scan a QR code and leave us a review. This helped increase the number of reviews YoY by 36% and they’re all 5 stars!

Those reviews serve a double purpose in helping drive discovery and social proof. We had one family who drove 20 miles after hearing about our patch from friends and neighbors and then looking us up on Google.

As a comparison: last year we had ~700 “direction requests” from our Google Business page. This year it was 1100!!!

Local School Distribution and Door Busters

We continued our tradition of distributing coloring sheets to all the local elementary schools with an offer of a free pumpkin if you brought the sheet to the patch. We also distributed it to local churches for youth groups, posted it on our Google Business page and new for this year we had takeaway coloring sheets at the patch itself. If people showed up without a sheet they could take one and come back. This drove higher repeat and retention rates as we got deeper into the season.

Why does this work?

In addition to driving repeat visits it creates an important word-of-mouth opportunity within the school community. Kids talk about it, they bring them home for their parents, and it costs us basically nothing to produce them. This in turn makes the patch a “school community event” rather than just a retail stop.

We continued our display of the coloring sheets along the fencing at the patch. The rain wasn’t exactly our friend sometimes but displaying hundreds of coloring sheets drove social proof and engagement. We’d also have one of our kid volunteers go out and tape up the sheet with the (usually younger) kid who colored it. It instills a sense of pride and community for those kids.

Merchandising and the Psychology of Abundance

Once again we piled the pumpkins high instead of single stacking them. No one wants to shop at a thinly-stocked, depressing-looking patch. And they also don’t want to take the “last pumpkin.” We never gave them the chance! By merchandising pumpkins literally spilling into the walkways and leveraging multiple stacks and displays throughout we created a psychology of abundance for our guests.

In the next section I’ll elaborate on why we really committed to this approach for this year.

Why does this work?

As mentioned above, no one wants to take the last piece of cake (or the last pumpkin). Having displays piled high gives people the confidence to browse, touch the merchandise and buy. If the displays are sparse as they were in prior years we saw hesitant customers and lower spend. The overflowing displays also give people an excuse to linger just a litttttle bit longer and choose carefully.

Volunteer-Led Sales Approach

As with last year everyone who volunteered at the patch was always wearing a bright purple lanyard and volunteer badge. It was a great conversation starter and allowed us to introduce ourselves as a 100% volunteer-run organization and a 501(c)(3) nonprofit.*

We had far more kids joining their Dads this time around: emphasizing the local school support, helping carry pumpkins out to cars, working the register and wandering around the patch helping customers with carrying and photo-taking.

Why does this work?

This is a clear differentiator from commercial patches. No one is helping you pick out the perfect pumpkin at a grocery store, or offering to carry them to your car. Most of the employees at local mega-patches couldn’t be less interested in helping you (I know, I’ve seen the apathy firsthand). The human touch matters; it’s like walking into the Apple Store of pumpkin patches. Supporting schools resonates deeply with visitors and kid volunteers are especially effective. Who says no to a kid?!?!?

*Note: we’re not having any OpenAI-level debates about our non-profit status btw. That’s rock solid.

Our Secret Weapon: Inventory Management

We continued our successful 2024 process of daily inventory tracking and sales forecasting based on historical data. Knowing what pumpkin varietals were popular at different times of the month (gourds in the beginning, carvers in the end) makes a huge difference especially in the psychology of abundance mentioned above.

We also continue to have a firm grasp on our per-unit margin calculations which helps us manage our discounting strategy. The good news is that costs to us didn’t really increase all that much this year either!

Why does this work?

You can’t sell what you don’t have, period. By understanding the cadence and demand for all 6,000+ pumpkins sold throughout the month we were able to prevent both over-ordering (waste, margin hit) and under-ordering (lost-sales, revenue hit).

What We Did Differently in 2025

I ended last year’s post with this:

I am not pleased to say we did very little of this in ‘25, hah! But we did do some new tactics, leaned heavily into other successful ones, and have a lot of learning from this year.

Weather Is a Gift or a Disaster

Last year we had an incredible heat wave in the earlier part of October, with temperatures in the upper 90’s and over 100 on a few days. This completely decimated sales. No one wants to shop for Fall when it feels like summer.

You could see the opposite effect this year: cooler weather meant sales jumped right from the beginning and trended above last year the entire time. This year that same weekend was ~30-35 degrees cooler. Wow.

The Keep It Full Philosophy

The one thing holding us back in prior years from fully stocking the patch all the way until shutdown was that we were worried about spoilage and waste. So our ordering, while pegged to historical trends, was always a little more conservative to preserve our margins. Selling every single pumpkin felt like success.

It turns out that this isn’t true. Selling every single pumpkin means we had to heavily discount at the end and left money on the table. Visual emptiness costs us more than waste and an empty patch discourages purchasing; especially in the final hours.

Our new approach for this year was to “take the gloves off.” We kept the patch fully stocked through the final day. We committed to abundance despite the fear of waste (more on why that was possible in a second). Our goal shifted from “sell everything” to “never disappoint a customer.”

So what made it possible for us to play so much offense?!? A few things:

1. In working with our supplier we were able to shift to a just-in-time inventory model.

2. We moved to a daily ordering and delivery system in the final week so that we could top up every day.

3. We negotiated heavily with our supplier to take inventory back that we didn’t sell. This was key! We were able to proceed knowing pumpkins wouldn’t be wasted.

4. A 4th year of forecasting has really helped us dial in our rates of selling.

Patch Papa Emeritus Matt Derrigo put it best:

“You can’t sell product you don’t have. We should be buying even more and the goal should not be to sell everything, but have some pumpkins left on the shelf; no different than selling the newspaper.”

The results of this strategy were immediately clear:

• The Sunday before Halloween produced our highest single-day sales (both gross and net) ever.

• This happened despite heavy rain and overcast weather.

• Because of the rain the parking opened up significantly and open parking spaces means people show up to shop.

• We saw our average order value and sales numbers climb in the last few weeks of the month, especially for carving pumpkins.

The positive feedback loop works!

Focus On Your People

With our tactics working and inventory strategy well-optimized we expected growth would be automatic. We learned something new though: our biggest constraint became people.

Getting Dads to volunteer this year was surprisingly difficult and different from previous years when participation came easily. We were baffled; why was this suddenly so hard?!?

Two separate conversations happened on two different occasions that clued us in to what the issue was:

Several Dads mentioned that they hadn’t been hearing about how the patch was going. I noticed in our Dads Club mailing list we were asking for volunteers but not celebrating the successes. These dads were both curious and concerned; they felt out of the loop.

One quote from a Dad who talked to Matt:

“I assumed you weren’t involved this year or if you were, things must not be going very well because the usual sense of excitement and progress was missing.”

Sharing updates is great, but they also need to be personal to connect with our volunteers.

What broke?

Historically after every volunteer shift we’d sent a personal note to the volunteers. That note thanked them specifically, shared how much money they helped raise, connected their efforts to the bigger picture and generally closed the loop on their volunteerism. This makes people feel valued, see their tangible impact, connect individual effort to the group’s success, and perhaps most importantly: encourages return volunteering. This is no different than running a company. How many times have you heard as a leader that employees feel disconnected from the strategy or what their impact is?!? It’s one of the most common pieces of feedback.

We missed this in 2025 when a new Dad took over scheduling, we hadn’t codified this lesson from prior years and it made volunteer recruitment harder.

Matt summed it up best:

“Feedback isn’t just a courtesy; it’s fuel. Timely, thoughtful feedback sustains engagement and reinforces a sense of progress and purpose. It’s essential to keep any volunteer effort thriving.”

Many Hands Make Light Work, But Is There A Limit?

As a way to include more people in the planning of the patch (hey, it’s Northern California; we value our inclusivity) we extended the work of the “Patch Papas” to ~6 people. In prior years it had been concentrated to 2. This had some upsides and downsides.

It lightened individual workload and gave everyone a specific, vertical task: social media, inventory, volunteerism, etc. This was generally good. But unintentionally it weakened cohesion and introduced a lot more new people into the system who didn’t benefit from historical knowledge. This was less good.

In the division of labor a few things got lost:

• The connection between volunteer shifts and sales results as mentioned above.

• Coordinating updates with the local newspaper to keep the momentum going.

• Alignment of social media with on-site activities

• A through-line that made it feel like we were speaking to the community with one voice.

Ultimately, including more people was probably a good thing but we have to ensure that the unconscious work gets documented and distributed so that we don’t drop any balls (or pumpkins).

New Social Media and Hello SORA!

We introduced two new tactics this year that helped create more engagement and awareness for the patch within the community.

First, even though we’re all old we started to realize that having a presence on Instagram mattered. Our ability to share visually appealing pictures from the patch, memes, and real-time announcements in the forms of reels and stories was super important. It takes time to build a social media following and it’s wild to think we didn’t start Instagram until this year. But our posts were highly engaging relative to our follower counts. We expect to build on this in subsequent years.

The second change for 2025 was in the form of a gift from OpenAI: SORA. Suddenly we were able to create highly engaging, highly shareable social media videos to introduce the patch, the Giant Jack, and build enthusiasm.

Take a look at some of these posts:

Clearly (tiny) Sergio (our social media guru) and I had a lot of fun with this. It turns out the community really got into it too!

Conclusion and Lessons Learned

As with prior years the lessons here aren’t just suitable for a volunteer, non-profit Pumpkin Patch. Here’s what I’d recommend for any team-based organization:

Personal, Timely Feedback

Team member (volunteer) engagement was down this year because we forgot the small, specific pat-on-the-back and thank you we’d sent to each volunteer after their shift in prior years. These really simple touchpoints build pride, momentum and retention. You want to make people feel the impact of their specific contribution.

Re-establish One, Clear Communication Thread

Distributing the work is helpful, but maintaining a single, central connector also matters. This helps to ensure consistency across team member updates, sales stats, social media, and press. There was some confusion this year on who made the final call on things. A central connector guarantees that everyone still contribute but communication feels unified.

Regular Status Updates

Just like you’d send out updates to your company each week a “State of the Patch” summary could have been really beneficial to everyone. It would provide concise snapshots of our progress and could include sales highlights, volunteer shout-outs, needs for the next week and photos. The objective would be to keep energy high and make everyone feel like they’re part of the story.

Lead with Calm

There was some uncertainty that hit this year: should we put signs out early, should we order more or less, should we open early and stay open later. It’s important to pause and look at the data then move forward confidently based on what it tells us. To keep pushing past our previous ceilings we’re going to have to take some bigger swings and that requires a steady hand.

Over the last four years we went through four different stages for our patch. Capturing the learning and paying it forward to future volunteers and Patch Papas was critical and something we’d never really done before.

2022: Discovery

• Learn what works and doesn’t

• Focus on customer acquisition and conversion

2023: Optimization

• Refine existing tactics and improve execution

• Focus on operational efficiency

2024: Scale

• Dramatically improved performance by leaning into proven tactics

• Focus on doing more of what works

2025: Sustainability

• Discovered our people constraint and learned leadership was our growth lever

• Focus on culture and confidence

For three years we optimized all of our customer-facing elements; then we discovered that further growth came from optimizing the team. The constraint wasn’t marketing, pricing or inventory - we pushed that to the limit this year - rather it was culture and leadership. Unless we figure out SaaS pricing next year this will probably remain true!

So how are these lessons universally applicable?

In the beginning of a company you’re focusing on finding your PMF, growing your business through customer acquisition and retention, and leveraging a ton of experimentation and data to drive improvement. You can really optimize your way to meaningful gains.

Beyond the early days, sustainable growth requires investing in your people. The invisible work of leadership compounds over time. Culture becomes the thing that makes the tactics work.

Tactics: You can’t sell from an empty patch.

Culture: You can’t run a patch without engaged volunteers.

Whether you’re running a pumpkin patch, a startup, or a scaled organization the lesson is the same: tactics matter, but culture compounds.

Happy Halloween and thanks for reading. We’ll see you next year at the Patch.

The Wildwood Dads Club is a registered 501c3 nonprofit. If you were inspired by this newsletter and want to contribute to local schools you can make a donation to the club here. It’s for the kids!

Subscribe now

Questions you’d like to see me answer? Ask them here.

Marina Miller spent nearly 12 years of her career building on the front lines with Palantir. At Palantir, she worked as an Echo: a hybrid of product manager, operator, and strategist embedded with customers in the field. She grew to the Head of Internal Data, Head of Cloud, VP of Product Ops and SVP of Product, Operations and Delivery before leaving for Etsy where she led Product for ML, AI, Data, DevX and Experimentation.

I met Marina about a year ago and was so impressed I’ve been bothering her to write a guest post for my newsletter ever since. I finally won!

In today’s article Marina shares some of her personal notes, stories and lessons from working shoulder-to-shoulder with different engineering partners at Palantir in a forward-deployed environment. She’s never shared these stories of high-stakes, high-stress and fast pace anywhere else and I learned a lot from her throughout this writing process. Like my last guest post with Dan Birdwhistell it’s rare to get such unique insights out of the head of a practitioner who did the work in the trenches.

Read on to learn:

• Why some of the most defining lessons of her career didn’t come from product frameworks or management books, but from her engineering partners.

• How misalignment between execution and empathy breaks great ideas, and how shared accountability builds them back.

• How speed depends on trust, not process, and how to move fast without losing people along the way.

• How true partnership becomes culture, and how empathy and feedback scale an organization.

I especially appreciate some of Marina’s stories that didn’t make this newsletter, like her writing about being a working Mom while at Palantir.

The quality of your engineering partnership can make or break your ability to ship great products. Yet for all the talk about collaboration between product and engineering, there’s surprisingly little written about what those partnerships actually look like in practice - how they form, where they break, and what they teach you when you get them right.

At Palantir, I served as what we called an Echo. This is a hybrid of product manager, operator, and strategist embedded with customers in the field. Every Echo had a Delta, their engineering counterpart. Together, we were forward-deployed into government and commercial environments to build software side by side with the people who used it. We had badges, logins, and direct access to analysts, investigators, and decision-makers. It was product development at point blank range.

Some of the most defining lessons of my career didn’t come from product frameworks or management books, they came from my engineering partners. Each one shaped not just how I build products, but how I lead.

This essay shares three case studies of those partnerships and what they taught me:

• James: How misalignment between execution and empathy breaks great ideas, and how shared accountability builds them back.

• Krasi: How speed depends on trust, not process, and how to move fast without losing people along the way.

• Brian: How true partnership becomes culture, and how empathy and feedback scale an organization.

Each partnership was different. One slow and methodical, one chaotic and high-stakes, one long-term and deeply trusted. Together they taught me that trust matters, how to forge it under pressure, and how to scale it into culture that outlasts any one relationship. They became the blueprint.

Learning to Execute with James

Where I learned that execution matters more than ideas.

James was the first real engineering partner I ever had at Palantir. It was 2012 and we were working on a federal law enforcement agency deployment together - what we called customer implementations. It was my first time building a real product for a customer. Following the forward-deployed model, we had badges, hands-on-keyboard accounts, and spent time with agents and analysts at the field offices. The intent was to truly understand their pain points before building features for Gotham, Palantir’s platform for investigative work and data analysis.

I was working with a financial crimes unit in NY where the entire division spent countless hours doing complex math on messy data for thousands of cases. James and I shadowed agents, analysts, and forensic accountants to observe their manual workflows when building cases. My product idea was an automatic calculator widget. A user would pull all of their relevant objects into Palantir Gotham, select them, select a few filters and hit “run.” In the background, the calculator would run all of the math formulas and produce the end result numbers that were previously calculated manually. When we launched it: crickets. No one used it. Here’s why and what I learned:

Lesson #1: Accountability is Everything

Work with your engineering partners to set realistic, specific deadlines, agree on priorities, and hold the line together on execution. Shared clarity on scope and timing prevents slippage and builds mutual trust. This was where I first learned that trust in product-engineering partnerships isn’t assumed - it’s earned through accountability.

When I proposed the calculator widget, I had full buy-in from the entire division - including the division head - but it still took weeks to ship the MVP. James was talented but moved slowly and I hadn’t pushed hard enough to define what the actual MVP was or when it needed to land. That lack of specificity eroded momentum and credibility. I learned that product and engineering have to align early on what’s realistic, write it down, and hold each other to it. Otherwise, “slow” becomes the default.

Lesson #2: Field Work is Product Work

Spend time embedded with customers to see how they actually work. What users say they need and what they truly need often diverge; direct observation exposes the real problems worth solving.

At the start, James and I thought we understood our users perfectly. We’d heard endless complaints from agents and forensic accountants about how tedious their math calculations were, so we built a helper to automate it. But we hadn’t realized that seeing the math was part of how they verified cases. By hiding those calculations behind automation, we broke their trust. If we’d spent more time watching their real process before building, we would’ve recognized that visibility was as critical as speed.

Lesson #3: Bring Engineers to the Field

Instead of filtering insights through specs or documentation, bring engineers to see users firsthand. When they experience the workflow themselves, they design smarter, more empathetic solutions.

Secondhand understanding is almost always incomplete. When engineers only read tickets or summaries, they miss the context that shapes a user’s real-world decisions. The best breakthroughs happened when James and I were both shoulder-to-shoulder with the agents, watching how they navigated Gotham. Those shared observations gave James the intuition to build tools that fit their environment instead of fighting it.

Lesson #4: Assumptions are Expensive

Validate your understanding before you build. Use quick prototypes to confirm that users’ mental models match yours. Catching misalignments early saves months of wasted development.

When we finally launched our “helper,” it technically worked. The math ran flawlessly in the background. But the forensic accountants immediately lost confidence when they couldn’t see how the calculations were done. Their entire verification process depended on visibility. If we’d shared a mock or walked through an early demo, we would’ve caught that assumption before building. The result was a perfect storm: a fragile UI, users who no longer trusted the product, and analysts unable to do their jobs.

Lesson #5: Depth before Breadth

Focus on the single user type whose success matters most. Broad consensus upfront often dilutes the outcome; depth for one audience creates a strong foundation others can adopt later.

Our team tried to build something that pleased everyone - agents, analysts, and forensic accountants - at the same time. In doing so, we served no one well. The widget worked for general math but not for the specialized needs of those who relied on traceable calculations. Had we focused on one primary user group, we could’ve built depth before breadth and earned organic adoption instead of forced consensus.

Lesson #6: Optimize for Superpowers

James and I continued working together on and off throughout my time at Palantir. What I’d initially seen as a frustrating weakness - his deliberate, methodical pace - eventually revealed itself as a strength when matched to the right problems. Years later, when he was assigned to complex, high-stakes search challenges that required painstaking investigation and careful analysis, that same pace became invaluable. The calculator widget didn’t fail because James wasn’t talented. It failed because we’d put a careful investigator on a project that needed rapid iteration.

Staff teams so individual superpowers fit the setup and ensure those strengths target business-critical outcomes. The sweet spot lies where capability, environment, and company priorities overlap.

Learning to Move Without Breaking Things with Krasi

Where speed, trust, and survival all blurred together at 30,000 feet.

James taught me the foundations of execution and product-market fit. My next partnership would test them in a very different context: operating at breakneck speed across messy, interdependent systems.

Lesson #7: Design the Escape Hatch

When doing foot-in-the-door custom projects for near-term revenue, keep scalability in mind. At the very least maintain excellent documentation and small, reversible pull requests. Otherwise you’ll end up maintaining unsustainable one-offs that stall progress toward a real product business.

When I worked with Krasi, it was my first time leading an entire portfolio with an engineering partner. Palantir had grown through a strategy of signing contracts with anyone who could pay and delivering massive amounts of custom work to make the product “sticky.” It worked. Our Gotham deployments were now mission-critical across government agencies, but the result was fragile. Each instance had evolved into a unique, highly customized system that the customer’s limited budget couldn’t sustain.

My job was to evaluate which deployments warranted continued investment, quantify the true cost of each relationship, plan graceful exits for those that didn’t, and preserve trust with our partners. Krasi and I were like a team of doctors: diagnose the code, the business relationship, and the operational health; decide what was worth saving; prescribe a sustainable treatment plan. I’d evaluate how each deployment was being used, the cost to support it, and the political stakes of exiting or scaling back. Krasi would dig into the technical implementation details to identify the exact work required to reach a stable, maintainable state. Then I’d take the findings back to the customers. Multiple difficult conversations explaining the transition plan, why it was necessary, and how they could continue using what we’d built by funding third-party contractors to maintain it.

This approach let us preserve the relationships and the impact of Gotham while freeing Palantir from the trap of endless custom maintenance. It also showed me the bigger risk: if you don’t build with scale in mind (or a clear path back to it) you stop being a software company and become a consultancy.

Lesson #8: Speed Comes from Trust, Not Frameworks

Working with James taught me that trust matters and how to build it through accountability. With Krasi, I learned how trust operates under entirely different conditions - at extreme speed with no time for frameworks or governance. Instead, build a rhythm grounded in:

• Assuming positive intent. No time for defensiveness or second-guessing motives.

• Asking clarifying questions without ego. No such thing as a stupid question when moving fast.

• Focusing solely on the shared problem. Forget who gets credit, just make it happen

We were moving at hyperspeed. Flying constantly from Salt Lake City to Louisville, living out of suitcases, and waking up unsure what city we were in. Government shutdowns halted work without warning. Regulatory changes forced mid-project pivots. Technical solutions that looked elegant in theory collapsed under real-world constraints.

Through all of it, what made the partnership work was complete interdependence. Neither of us could succeed without the other executing flawlessly. The stakes were enormous: politically sensitive work, limited budgets, and a brand reputation on the line. We developed a simple operating rule - if one of us needed something, the other made it happen. No friction, no bureaucracy, just trust and velocity.

Learning to Lead with Brian

Where partnership turned into leadership and leadership turned into culture.

Krasi taught me how trust and psychological safety can be forged fast under pressure. Brian would teach me something harder: how to scale those same principles into a culture that outlasts any one partnership. This is how you build high-trust teams that can deliver mission-critical software at scale.

Lesson #9: Optimize for Superpowers (Again)

Brian is the best engineering partner I’ve ever had. We first met when we were set to lead a portfolio of U.S. government deployments together - me leading execution, him leading engineering. We started strong, but he was soon pulled to work on a custom system for a major federal customer, and Krasi took over the portfolio work. Over the next seven years, Brian and I worked together on different projects, and he became my most trusted engineering partner. I would work with him again in any capacity, any time.

After four years in the forward-deployed world, I needed a change. Brian recruited me to help tackle organizational execution problems, first on an internal data-platform team and later leading the cloud team. What made him exceptional was how he approached partnership. When I questioned whether I could lead engineers credibly as a non-engineer, he helped me see that I didn’t need to pretend to be something I wasn’t. He used first-principles thinking to break down what “technical leadership” really meant: someone who could work backward from outcomes, understand why the work mattered, and translate abstract goals into concrete priorities that teams could execute and deliver on. Those were exactly my strengths and he made sure I leaned into them.

It reminded me of what I’d learned years earlier with James - that people do their best work when they’re placed in environments that amplify their natural strengths. Brian lived that principle intuitively. Instead of trying to make me more technical, he helped me double down on what I already did best and built the space around me to make that matter.

Focus on amplifying people’s existing strengths rather than trying to fix their weaknesses or make them into something they’re not.

Lesson #10: Drop the Ego and Do the Work

To build credibility with engineers, take ownership of your learning plan. Don’t wait for permission or an onboarding process. Figure out what you need to know, do the work, and be willing to be wrong in front of people who know more than you. No matter how senior or accomplished you are, every new problem, team, or domain demands the same curiosity and humility - the moment you start leading from a distance without doing the groundwork, you lose the trust of the people building the product. Worst of all, don’t fake it; that destroys credibility in a way that’s hard to recover from.

When I moved into the cloud organization, I knew almost nothing about cloud infrastructure. Brian told me, “Don’t worry about the domain. It’s your skill set that we need.” So I got to work. No one asked me to get an AWS certification. It was something I felt would help, so I just did it. I also started a new ritual we called “murderboards”: I’d stand at a whiteboard and explain our entire system architecture while engineers interrupted to tell me everything I got wrong. It was uncomfortable, humbling, and exactly what I needed to build credibility.

Lesson #11: Emotions Aren’t Noise, They’re Information

When someone reacts strongly in a professional setting, be the Brian. Don’t write it off as an overreaction or drama. Lead with empathy, stay curious, and look for what that emotion reveals about context, motivation, or unmet needs.

When I came back from my second parental leave, one of my reports was running the cloud team beautifully, and I didn’t want to take that away from her. I went to Brian because I loved working with him and wanted to find a new way for us to collaborate. His first idea was for me to work on our FedRAMP accreditation - a federal compliance program that allows software to operate in government environments. This was a massive initiative with more than a hundred people already involved. For some reason - hormones, insecurity, postpartum state, whatever - I just started crying.

“I can’t believe this is what you want me to work on. There’s not even space for me.”

Looking back, I was being pretty entitled, but Brian didn’t judge me for falling apart. He didn’t owe me anything, but he understood that I’d come to him because I trusted him to help me figure out the next right thing. He gave me grace and treated my reaction as feedback rather than frustration. A few days later, he came back with a new idea: for me to be Chief of Staff to him and our Chief Architect. My new boss, who was deeply invested in my growth, encouraged me to counter with a proposal to join as an equal peer leading the entire org alongside them, not in a supporting role. They agreed.

Lesson #12: Make Feedback Normal

Give specific, impact-focused feedback in the moment and model how to receive it constructively. When leaders can openly receive feedback without defensiveness, it creates permission for everyone else to do the same. {Editor note: I just talked about this on the Best Manager Ever podcast}

Earlier in our partnership, when I spiraled over a difficult direct report, Brian was empathetic first. He listened and let me vent, but then he gave me direct feedback that my spiral would torpedo my reputation if I didn’t get it under control. It was hard to hear and exactly what I needed. That balance of empathy and candor became the foundation of how we worked together and set the tone for the feedback culture we later built.

As senior leaders, we continued that dynamic. We could give and receive feedback easily, even mid-meeting. Once, he told me right in the middle of a call that it was obvious I wasn’t paying attention and explained the impact not just on him, but on how it undermined the rest of the team. I paused, thanked him, and changed it immediately. That moment set the tone.

We modeled a feedback culture that was both direct and caring. It signaled to our organization that feedback wasn’t personal, it was how we got better together. As other leaders saw that dynamic play out they began doing the same with their teams. It became one of the cultural cornerstones of how we operated.

Lesson #13: Culture by design, not default

James taught me that trust comes from accountability. Krasi showed how it enables speed. Brian taught me how to scale it into an organization’s fabric.

We started with three teams and thirty people and eventually scaled to 150 people across fifteen teams around the world. I operationalized all of it, including building out new roles like TPMs, which would never have succeeded without the partnership Brian and I had. At that scale the best partnerships involve explicitly defining what you stand for together. Actively discuss your cultural values and priorities, then work to create that environment - even when it means uncomfortable conversations or resistance from leadership.

Our org had some brilliant technical leaders who struggled to create safe environments. Meetings sometimes ended with people upset or even in tears. The result was that our “best-idea-wins” principle had stopped working. People were too afraid to say the wrong thing so we weren’t surfacing the best ideas at all. We started losing great people who either transferred to other orgs or quit outright.

To course-correct we ran what we called a rubber-duck exercise around our cultural values and priorities. Talking through them aloud, challenging assumptions, and rewriting them until they actually reflected the environment we wanted to create. Together, we re-built that culture from the inside out. It wasn’t easy, but it worked: people felt safe again to speak up, challenge ideas, and contribute. That’s when the org started performing at its best.

Closing Thoughts

Technology changes fast, but breakthroughs don’t come from code alone. They come from partnerships built on trust, curiosity, and shared ownership - and from leaders who can scale that into culture. The best products aren’t built by process or frameworks; they’re built by organizations that make partnership a way of operating.

Subscribe now

Questions you’d like to see me answer? Ask them here.

Many takes on Atlassian acquiring The Browser Company so I thought I'd actually break this down strategically to see if any of it makes sense.

First, The Browser Company makes (made?) Arc (a chromium based browser), attempted a search-first AI application for your phone, and has recently invested a lot in Dia which is a next generation AI-first browser.

They've raised ~$130M and were last valued at $550M until Atlassian announced they were acquiring them this week for $610M. So a ~4.7x markup on the invested capital in ~6 years.

Most of the questions and takes I've heard about this acquisition go something like this:

• Why did Atlassian pay $610M for a browser company that doesn't make money?

• Why can't I get acquired for $610M if I just make highly-produced marketing videos?

• Why couldn't TBC make it on their own?

• Atlassian is going to JIRA-ify this product SO fast. They're doomed.

• Numerous memes, dunks, etc. The internet had a GREAT time with this one.

Okay, so let's start with the assumption that Atlassian has a ton of smart people working in their CorpDev department -- they have made other, quite great acquisitions and so I'm sure there was a thesis here.

The Browser Company couldn't make it on its own because they had no, meaningful revenue model for their browser and had painted themselves into a corner. Today, the only revenue model for browsers is paid search deals and advertising. With the DOJ ruling from this week saying Google can still pay for default search placement that doesn't seem to be changing anytime soon.

In order to have a meaningful "Google pays us" model you have to have usage to send Google a lot of traffic (since most of their deals are a revshare on search result ad clicks). Getting to levels of usage that are interesting to Google for *a new* entrant in the very crowded browser category is *very* challenging. Safari, which is used by a zillion mobile deviceholders only has about 15% of the marketshare of the browser market. Google has 65-70% with Chrome. And # 3 is in the single-digit percentages.

Furthering this challenge is that Arc integrated (or tried to) a bunch of AI stuff into the browser that would cause people to not use Google Search anymore. For example, in my Arc instance I've had Perplexity as the default search since I installed Arc a few years ago. Dia is even *more* in this direction since it is quite literally an AI browser. These products cater to the tech early adopters and not much else and Arc didn't have a strong value prop and counter-positioning to Chrome so... why switch? So no traffic, no revenue, no hopes of getting there really (even with Dia).

They couldn't do subscriptions because that's not going to generate enough $ to give a return to VCs of any kind (or probably even cover their costs) and people don't want to pay for a commodity product (web browser) that they generally can't even remember the name of. So TBC couldn't keep going on its own and it couldn't raise more capital after grabbing $130M.

Enter Atlassian. They stroll in like the ~$44 - 45B market cap company they are… with the product that everyone (everyone) hates but can’t seem to overcome their Stockholm Syndrome enough to switch to Linear (kinda like Salesforce) and drops $610M in cash on TBC.

So, WHY?

If you read their announcement they’re now working on the AI browser for knowledge workers. When I was working with Mozilla one of my theories was that the Enterprise browser was the only new entrant that you could create (see Island for reference) because Enterprises will pay for stuff that makes their workers less distracted, more secure, and ideally more productive.

And that has been Atlassian’s whole thing with an integrated suite of stuff that helps up and down the stack. Yes, it’s mediocre software that is poorly designed and has terrible UX, but it also has lock-in. And expansion. And a go to market engine that is already primed to cram a new product down customer’s throats. And Atlassian is much more acquisitive than they are good at really building products on a new horizon. Not to mention they have been pretty lackluster in the “sprinkle some AI on it” department. And the team at TBC has a few things that Atlassian lacks:

1. A browser that they’ve been working on for awhile.

2. AI expertise.

3. Browser-building expertise.

4. Visibility into the sites visited by their users.

5. Design chops.

6. Video-making chops.

7. Founder energy.

8. A head start.

So they saw an opportunity to spend ~20% of their cash-on-hand and get a team with the 8 things above. Now, 20% is a lot, but Atlassian has good free cashflow so they’ll recoup that pretty quickly. They also have a way to signal to the market that they’re serious about innovating, they get the AI thing, and they understand that they’re only as good as the platform they sit on top of (the browser).

The market kind of likes it?

Though not as much as Google getting very favorable results in their DOJ case…

So, given all of this it might make some sense why Atlassian would do it and the market agrees a little bit. At the very least, let’s see a company that hasn’t taken a risk since I was in college try to give themselves a shot in the arm with an acquisition. Or two (they also acquired Cycle this week, for an undisclosed sum).

So yes, will they JIRA-ify the beautiful design of Dia? Probably a little bit. Are they hoping that some of that founder mode energy rubs off on the rest of the company? Absolutely. And will this actually be a successful acquisition for them… I guess time will tell but I think it could help them a decent amount. Whether it’s worth 20% of their cash… probably not, but maybe that’s what it took to get this team and their expertise.

Curious to hear other people’s takes below in the comments. Let ‘er rip.

Until next time,

Adam

Share

Subscribe now

In 2024 I joined Mozilla as an interim leader for New Products. My charter was to lead two primary workstreams that would eventually get us to a new, $100M+ product line. Those two workstreams were a New Products Strategy and an approval process for advancing (or not) new products at the company. I also needed to support the hiring of my replacement, figure out what to do with all of the existing, in-flight new product ideas and assess whether we had the right people in the right spots working on new products. But those are for another time.

Today’s newsletter is about the approval and vetting process for new products.

Building new products inside a big company is messy. Teams sprint ahead without clear goals, leadership waffles on when to invest, and before long you’ve got a portfolio full of “zombie projects” that aren’t alive enough to scale or dead enough to kill. 🧟

We needed a way to bring clarity, speed, and discipline to the process without crushing creativity via too much process. To do this we built a Stage-Gating framework for new product development. It’s a system that turned go/no-go decisions from art into (mostly) science, and gave us the courage to say “no” early, so we could say “yes” to the right bets.

In this newsletter I’ll cover:

• Why Mozilla needed a stage-gating framework to bring order to new product development.

• What we tried before and why conviction-led reviews weren’t enough.

• The five stages of our framework (Discovery → Exploration → Viability → Growth → Sustain), and the gating questions that matter most.

• How investment levels, deliverables, and governance change at each gate.

• The biggest lessons learned from killing projects faster to normalizing “no.”

• How you can adapt stage-gating for your startup or growth-stage company without adding bureaucracy.

• An AI-enabled accelerant: how AI can act as a “Stage-Gate Coach” to help PMs prep for reviews and spot gaps in real-time.
  
  

The Problem

For those of you who know Mozilla you’ll probably think of it as the Firefox browser company. That is the flagship product, but Mozilla also has a rich tradition of releasing great open source software, building high-quality consumer safety and security products, and being ~3-5 years early for some big bets that ended up being successful elsewhere: a lightweight OS (Firefox OS vs. Chrome), a password manager (Firefox Lockwise vs. 1Password), a secure file sharing product (Firefox Send vs. WeTransfer/GDrive/Dropbox/etc.), an SSO system (Mozilla Persona vs. Auth0/Okta/OneLogin), a framework for running Web apps as Desktop apps (Prism vs. Slack/Discord/Chrome PWAs), etc.

For the macabre amongst us there is an entire website dedicated to these efforts: https://killedbymozilla.com/

😢

On the flip side, Mozilla also overinvested in areas where they couldn’t win: social, article storage/bookmarking, notes, video chat, and more.

They had (and still have) the potential to be a successful, multi-product company but historically ran into issues of under/over-resourcing, taking too long to decide, or asking questions about commercialization or distribution way too late. They struggled with building things that seemed interesting and novel to the builders but didn’t have an audience or effective method(s) to reach one.

Further exacerbating this was the whiplash that a lot of teams felt. Product investment was like a ride on a rollercoaster: funding goes up up up and then comes crashing down down down (usually resulting in the layoff of people hired to support the ‘up up up’ phase). Talented people were hesitant to work on new products because they wondered if that meant their days were numbered at the company. And leaders were often wondering: “what’s going on with ‘Idea X’ or ‘Product Y?!?’”

Earliest Solutions
One of the original (and right) decisions the company made was to separate out new product development from the core browser product work. The act of building 0-to-1 is different from maintaining a mature product: different approach, effort, goals, etc. and you need to shield those new efforts from the bureaucracy of the larger organization.

But the original building and review solutions weren’t great. There were informal, ad-hoc reviews. There was the occasional executive greenlighting a pet project based on their own conviction. There was cool technology. Sometimes there was just running until you hit a wall and ran out of gas.

As you might imagine, this doesn’t work. But it’s easy to fall into this trap at a big company. It feels like progress but it’s really just motion. The outcome, in hindsight, was obvious. With inconsistent standards you had some projects get funded with lots of capital and people (too early) and others languished without enough support. It was the worst of all worlds:

Plenty of product reviews, but no shared criteria. More art than science.

A New Approach: Stage Gating

When I joined it was clear to many that we needed something different than what we had been doing. A few products had gone through their own version of an actual 0-to-1 process and come out the other side looking pretty good. But it wasn’t operationalized… yet.

Readers of this newsletter will know that I’m a big fan of my friend Brian Balfour’s Four Fits Framework.

If you’ve been living under a rock for the last 10+ years you should really go read about it. Here it is encapsulated in a single paragraph by AI Brian on Casey Winter’s SuperMe.ai:

Thanks Robot Brian and founder Casey.

I wanted to operationalize the framework above because it’s the best one I’ve come across for figuring out which opportunities can grow to $100M+ (although by today’s AI standards it’s probably $500M+ or $1B+).

Working with a handful of folks at Mozilla – a project manager and a few product leaders who had done something approaching 4 Fits – we started sketching out a Stage Gating process.

Adam, what is Stage Gating?

I’m so glad you asked!

Stage Gating is a five-step (or stage) process that we designed for 0-to-1 building at Mozilla. It merges the best of early product development practices, venture capital evaluation, and the 4 Fits Framework to provide explicit goals, deliverables and gating questions at each stage. Can’t answer those questions? The product is at risk. If you can, there’s a strong case for more funding and continuing work.

The Five Stages (Gates)

Gate 1: Discovery

The overall question to answer here is: Is this an attractive market opportunity?

Each team had to do their own research to find the intersection of the customer and the problem. This required asking questions within the team like:

• What problem are we solving?

• How painful is this problem for the potential customers?

• How urgent is this for potential customers?

Then we cross referenced that with Mozilla itself to ask whether the opportunity made sense for us. Here you would ask questions like:

• Is this a new market or does the market already exist?

• If existing, who operates there today and what are their differentiators?

• How big is the market and which direction is it trending?

• Does this fit within our product strategy and do we have an entry point?

• What’s the revenue model in the space; how long has it taken entrants to reach revenue and profit (if ever?)

• Roughly: do we think the opportunity outweighs the investment or has the potential to? Why?

To answer these questions the team would need to define personas, conduct desk research and some prospective customer interviews, analyze competitors, and create a back-of-napkin revenue model. None of this had to be perfect (and in fact we assumed it wouldn’t be) but it forced thinking early on around whether to continue pursuit or pivot.

This is a short-lasting phase with ~1-2 people involved (usually a PM and a Designer or just a PM). Decision makers on whether something would advance beyond this were the product leaders.

One of the best ways to move forward in this stage was to draft off of an existing market by identifying a unique or differentiated entry point, or a specific part of the problem that existing solutions weren’t solving effectively.

Teams would often go through this phase and the next in a loop until they were ready to exit to the 3rd stage.

Gate 2: Exploration

Once we were confident in an attractive market opportunity and meaningful problem to solve, teams would then work towards various solutions to that problem. The overall question to answer here is: have we found a desirable solution <> problem fit?

In this stage, teams were expected to answer questions around technical feasibility, solution strength and iterate on the business case / model:

• Can we build it? (usually the answer is ‘yes’)

• Do we have the capability to build something robust here?

• Is it cost effective relative to the resource needs? (Note: this does not mean ‘profitable’ at least not yet).

• Are customers giving us a signal that they like our solution(s)?

• What does engagement and retention look like?

• Is there a directional willingness to pay?

• Is the business case and market opportunity better or worse than before?

• How might we get this into customers’ hands?

To answer these questions the team would need to do prototyping (AI really, really helps here!), do some external validation and experimentation, collect user feedback, identify the highest risk assumptions, and develop an initial go-to-market hypothesis and a qualitative growth model.

We expected this phase to last ~3-6 months with a lot of iteration in that time. There are plenty of startup accelerators who operate on a similar time horizon (albeit with some attached funding).

Teams were still very small at this stage because you can do a lot of prototyping with a PM and a designer. You might also have an engineer join the team at this point to establish a tripod of product, engineering and design.

There’s really no shortcut to moving through this phase: you have to brute force generating a ton of different solutions and testing them with customers. To exit you need a (mostly) clear solution that is of interest to an ICP and is feasible for Mozilla to bring to market. The details of that don’t have to be fully sorted out but you should be able to squint and see it.

Once you exit this stage it’s on to the really fun part: PMF!

Gate 3: Viability

In this next phase teams would work towards PMF and answer the question: do we have evidence of product <> market fit?

As my friend Casey Winters likes to say, “Product<>Market fit is a hell of a drug.” And so entering this phase meant a step up in resources to build and maintain a product. Typically this came on the engineering side but might also mean adding a marketer to the team.

We all know that the ever-elusive product<>market fit has a variety of proposed measurements. Lenny Rachitsky (maybe you’ve heard of him?) has some good quotes on it here.

My favorite of the vague quotes towards the bottom of that article is Don Valentine, founder of Sequoia:

“When the customers want your products so badly that you can screw everything up and still succeed.”

In this phase you’re looking for a lot of the following:

• How are adoption, engagement and retention metrics performing?

• Do we see a flattening retention curve? Organic adoption? High NPS?

• Are there attractive, niche market segments that stand out?

• What’s our growth plan and can we model it out?

• What else do we need for scale: Marketing? Operations? Infrastructure?

There’s no shortcut to product<>market fit and this phase could last 18-24 months.

As I mentioned above, teams grew a small amount in this stage, but not much. The likely investment was in engineering to iterate on the product and a marketing person to help with GTM. Depending on the PM’s skills though they could take a lot of that responsibility (and I’d argue, they should). And if you’re super adventurous you could probably vibe code the whole thing. Although [fill-in-the-blank BigCo] probably wouldn’t want that out in production.

Approvals and contributors are a bit different at this stage too. You’re introducing monetization so your Finance team may have an opinion on this (for example). You’re still primarily keeping approvals contained within the Product org but buy-in from partners in Marketing and Operations might be needed as well.

If your product survives through the first three stages that means it’s ready for the big one, Growth.

Gate 4: Growth

In this stage you’re answering the question: is this business sustainable at scale? Plenty of products with PMF never reach their breakout moment or they require an obscene amount of capital to get there (TikTok; looking at you).

Here you’re answering these questions:

• What does our growth model look like now that we have real data?

• What would it take for this product to get to self-funding?

• How is this performing relative to industry benchmarks and (more importantly) within our desired payback periods and LTV targets.

• Are there adjacent users and audiences we could build towards?

• What resources do we need to support meaningful scale?

If you’ve gotten to PMF it’s almost always time to step on the gas. Established companies can usually afford to spend ahead of revenue for a product with a high revenue ceiling. Your resource investment goes up significantly here; if you were a standalone company you’d probably be raising growth capital.

It’s expected that you’ll live in this phase for many, many years. Think about the number of growth stage companies out there and how long they’ve been growing. The current AI boom aside, it tends to be a long time.

Most importantly, this is the time that a product gets spun out / graduates from any sort of internal incubator. There are very few products that will ever make it this far and they’ll be trajectory-changing for the company when they do.

Which brings us to our final stage…

Gate 5: Sustain

If you make it to this stage your product has been around for a very, very, verrrrrrrrry long time. The question you’re answering here is: what’s the right balance of growth and monetization?

Here you’re running this like a standalone business; fully supported and integrated (or not) into the core experience. You shouldn’t need “outside” capital because you’re funding the business from existing operations. I think most people reading this haven’t been (and won’t be) at a company long enough to live through this second act. And most zero-to-one product builders would rather dive back into the fray and start over with a new idea again once they get here.

What Did We Learn Through This Process?

Defining and implementing a stage gating framework and process was a fun and sometimes painful exercise. We operated in a Venture Studio-type model, but other companies have done it differently: internal pitches to get into the incubator, outside pitches to fund external bets that then get acquired/absorbed into the parent, skunkworks teams, etc. No two companies that I talked to had taken an identical approach. Ours was designed to solve the problems I mentioned at the beginning of this newsletter.

Stage gates provided clarity. Teams always knew the bar for advancing which was a refreshing change. But they also raised the bar which was painful. We killed a lot of projects immediately and had a much higher standard for what got through the earliest stages. This was especially true when evaluating the business model, market and distribution strategy. All three were areas we had floundered at in the past and wanted to fix.

Leaders got visibility without being bottlenecks. At a BigCo, leaders love to meddle. It’s how they “add value” to the process. But they also don’t typically have the best ideas and we didn’t suffer from an ideas problem; we suffered from landing those ideas within the Four Fits framework. The other thing leaders love at BigCos is updates. The stage-gating process provided exactly that: regular, templatized updates on progress, predictable milestones, and all the things that VPs and SVPs salivate over.

Projects died faster (a success!) because the criteria made “no” decisions less political. I mentioned above that we killed a lot of projects immediately. This was the system functioning as designed and because we had planted so many seeds we had to rein it in right away. Ripping off the bandaid was better than letting things go; especially when we knew that objectively many projects didn’t stand a chance. Removing the politics from the decision making was really productive. No longer was funding based on whether you could convince the CFO that you should have more money and we ended up with crisper proposals and projects as a result.

It required cultural adjustment. Teams initially saw it as “bureaucracy” until they realized it unlocked more predictable investment. It’s not super fun to go from the wild wild west to an actual, formulaic process. Rolling it out to projects that were in various states of “funding” led to a lot of “feelings.” It meant re-evaluating and challenging past assumptions, led to some thrash, and felt sudden to some folks. In hindsight I wouldn’t have changed anything. You have to pivot on this stuff quickly and ripping it off like a band-aid is the best way to do it. So my advice here is to go “founder mode” and just make the change.

One surprise: viability was the hardest bar to clear. Maybe not surprising, getting to PMF is the hardest. Couple that with a projection on unit economics, some thoughts on GTM, and more scrutiny… and just just don’t have that many things making it. Most ideas went back to the drawing board in between discovery and exploration and stopping many things that didn’t clear a PMF bar was the right outcome. The challenge was how much oxygen to give a product in a BigCo environment. In a startup, you run out of funding and that’s it. But BigCos don’t quite have the same binary outcome so you have to force it.

How you can apply this at your company

First, let me caveat with some real-talk: this is an idealized version of this process. We had rough edges that needed to be sanded down and like any framework you can think of this process as a tool NOT a rule. So don’t cargo cult this exact thing into your environment. Remember, I talked to a bunch of companies who all had somewhat different processes but if you squinted they all had the same general structure.

If you’re an early-stage company. You shouldn’t be placing a lot of tangential bets anyway. Stay focused on the primary product because if that thing can’t be successful then you’re dead. Rather than copying the process, copy the questions. For example, during Discovery are you asking is this even a big enough market? Chances are your prospective investors will be so good to be prepared.

This is far more useful at a company that is at the growth stage or multi-product. You need some version of stage gates to rationalize investment and avoid the pitfalls we experienced: pet projects or zombie initiatives. If you’ve been working on one, core product for awhile there’s a chance you’ve forgotten what it takes to build 0-to-1. Copying these questions, even if you don’t stick to exact gates is helpful and remember that they’re derived from the Four Fits framework which is solid gold for a company and product-builder.

My main advice is to match the formality of the gates to the scale of the bets you’re making. You wouldn’t want to run a $50k experiment through a process like this or a big feature enhancement, but if you’re looking for the next big bump to the P&L you would.

How do we sprinkle some AI on it?

There’s this hot new thing called AI that you may have heard of and building yourself a Stage Gating GPT can be really helpful. It’ll play the role of the CPO or leader giving you a critical evaluation of your progress and whether you’re ready to ask for more funding.

Even better: what if every PM had an AI coach that helped them prepare for a stage-gate review?

Imagine uploading a product brief, having an AI probe with gating questions and then scoring you against the criteria for each stage and suggesting what’s missing. You’d democratize the “muscle memory” of stage-gating and get some practice defending and poking holes in your thesis. This can be useful ahead of product reviews as well and can be extrapolated for them.

Here’s how to build it

Define System Behavior

Create these custom instructions:

You are a Stage-Gate Coach, an exacting but supportive reviewer. Your job is to:

1. Identify the product’s current stage and target gate.

2. Run a structured readiness interview focused on that gate’s questions and deliverables.

3. Score evidence on 5 dimensions: Market/Problem, Solution/Feasibility, PMF/Performance, GTM/Unit Economics, Governance/Resourcing.

4. Output a decision: PASS, REVISE and RESUBMIT (with actions), or DO NOT ADVANCE along with a short, executive-ready summary and a clean checklist of gaps.

Your constraints and tone:

• Be concrete, evidence-based, and time-aware. Don’t over-ask.

• Ask one precise question at a time; request artifacts (links, charts, retention curves, brief metrics, etc.).

• If a user can’t provide evidence, propose “minimum proof” for this gate (what to measure or build next).

• Demand evidence, not promises - "We believe" statements are insufficient

• Use the frameworks’ definitions of stages, gates and step-changes in investment/governance as your baseline evaluative criteria.

• Be skeptical by default and assume ideas are bad until proven otherwise

• Focus on the "no;” remember your job is to prevent waste, not be liked

• Call out sunk cost fallacy; the previous investment is irrelevant to gate decisions

• Identify "zombie products" early; be tough on products that don’t seem like they’ll ever achieve meaningful scale

• Challenge technical enthusiasm because “cool technology" is not a business case

• Quantify everything; vague metrics hide failing products

Define Stages and Gates via Snapshots

Use the following stages and goals:

• Discovery: Find an attractive problem <> customer opportunity; early desk/customer research; draft success criteria. No depth on monetization needed.

• Exploration: Validate problem-solution fit and technical feasibility via prototypes, alphas/betas, dogfooding, etc. Have some signal on monetization but nothing more.

• Viability: Ship a first version and assess product<>market fit, form your GTM and growth hypotheses, define unit economics and modeling for P&L. You should know how this product will generate revenue.

• Growth: Demonstrate scalable growth and revenue via optimized GTM and operations. You must be generating revenue.

• Sustain: Operate a profitable, efficient business. You must be generating revenue.

Gate questions (the “bar”)

• Gate 1 (Discovery→Exploration): Is this an attractive market opportunity?

• Gate 2 (Exploration→Viability): Do we have desirable solution <> problem fit, and is it feasible for us?

• Gate 3 (Viability→Growth): Have we established Product-Market Fit?

• Gate 4 (Growth→Sustain): Can this business scale sustainably (profitable unit economics & operations)?
  

Investment changes

Budgets will go from less than $100k in early stages to multiple millions at the growth stage and then evidence of self-funding beyond that. Use these ranges and milestones to sanity-check the asks.

The Interview Flow and Conversation

Phase A: Intake (keep to 3–5 questions)

1. What stage are you in, and which gate are you preparing to pass? (If unsure, the assistant infers from evidence.)

2. 1–2 sentence product: target user, problem, core value.

3. Links or brief paste: latest deck/PRD, top-level metrics (retention curve, activation, NPS, revenue if any).

4. Funding ask (if any) + timeframe; key risks you want advice on.

5. Org context: team size, dependencies (marketing/data/eng), runway.
   

Phase B: Gate-specific deep dive (ask only what’s needed; stop when confident)

• Gate 1 (Attractive Market Opportunity) - Evidence prompts: market size and trend, competitor set & differentiation, target persona + urgency, preliminary revenue model hypothesis, success criteria. “Minimum proof” if missing: lean canvas + desk research + 5–10 interviews + draft revenue model + pass/fail criteria.
  Gate 2 (Solution <> Problem Fit & Feasibility) - Evidence prompts: prototype/alpha/beta learnings, engagement/retention signal, discovery channels, willingness-to-pay signal, basic tech architecture & build-vs-buy rationale, initial strategic revenue model, estimated investment to reach PMF. “Minimum proof”: clickable prototype + 10–30 user tests + north-star metric movement + feasibility write-up.

• Gate 3 (Product-Market Fit) - Evidence prompts: retention curve flattening, organic adoption %, NPS, unit economics in 3–5y context, channel efficiency, growth plan, infra readiness (analytics, lifecycle marketing, CX). “Minimum proof”: cohort retention flattening + repeat usage + early unit economics + working acquisition channel.

• Gate 4 (Sustainable Scale) - Evidence prompts: P&L self-funding status, growth metric targets, CAC payback, ARPA↔CAC spectrum, churn, ops resourcing, scale risks.

Phase C: Decision & Coaching

Compute scores (below), produce decision, list gaps, and generate a Next-2-Weeks Plan (experiments / analyses).

Scoring Rubric (0-3 each; show subscores)

Dimensions & examples of “3 = strong”

1. Market / Problem: Clear persona & urgent pain; credible market size & trends; differentiated space. (Gate 1 critical)

2. Solution / Feasibility: Prototype evidence; strong value perception; discovery channels known; feasible, cost-sane build. (Gate 2 critical)

3. PMF / Performance: Retention flattening, organic adoption, NPS 25+; 1.0 shipped; measurement in place. (Gate 3 critical)

4. GTM / Unit Economics: Early CAC/LTV logic, channel efficiency, growth hypothesis; validates ARPU/CAC assumptions. (Gate 3–4 critical)

5. Governance / Resourcing — Ask aligns with stage; step-change investment justified; cross-functional alignment noted.

Decision rule (guide)

• PASS: Target gate’s critical dimensions ≥2.5 average, no zeros; risks mitigated.

• REVISE and RESUBMIT: Any critical dimension ≤1, or 2+ material gaps; provide an action plan.

• DO NOT ADVANCE: Fundamental miss on the gate question or investment unjustified; advise pivot/stop.

Output Scheme (what the assistant returns)

{

"stage": "Discovery | Exploration | Viability | Growth",

"target_gate": "Gate 1 | Gate 2 | Gate 3 | Gate 4",

"summary": "Executive-ready 4–6 bullet overview of evidence and context.",

"scores": {

"market_problem": 0,

"solution_feasibility": 0,

"pmf_performance": 0,

"gtm_unit_economics": 0,

"governance_resourcing": 0

 },

"decision": "PASS | REVISE_AND_RESUBMIT | DO_NOT_ADVANCE",

"rationale": "Why this decision, tied to the gate question.",

"gaps": ["Short, crisp list of missing proofs"],

"next_2_weeks_plan": ["Experiments/analyses with owners & success criteria"],

"attachments_requested": ["Retention curve", "NPS snapshot", "Proto video", "Revenue model tab"],

"investment_check": {

"ask": "$ amount or 'None'",

"is_stage_appropriate": true,

"notes": "Step-change justification or mismatch"

  }

}

Few-Shot Examples

Example A — Gate 1 (Discovery→Exploration)

User:

“We’re preparing for Gate 1 for ‘ProductX.’ Here’s our 1-pager and 8 interview notes. We think the market is growing.”

Assistant (asks 2–4):

• “What’s your quantified market range and core competitor set?”

• “Summarize the user’s current workaround and urgency (evidence quotes)?”

• “Show the high-level revenue model hypothesis and success criteria you’re using.”
  (These anchor Gate-1 questions)
  

Assistant (decision):

• Decision: REVISE & RESUBMIT

• Rationale: Market evidence is promising but differentiation vs. 3 named competitors and revenue model confidence are shallow for Gate 1.

• Next-2-Weeks Plan: Run 10 more interviews targeting 2 priority personas; produce TAM/SAM/SOM banded estimate; draft 1-page revenue model with ARPU hypothesis; define pass/fail criteria for prototype test.
  

Example B — Gate 3 (Viability→Growth)

User:

“We shipped 1.0, DAU 8k, WAU retention curves flatten at wk-8, NPS 36, 48% organic.”

Assistant (asks 2–3):

• “Share the cohort chart and CAC payback on your primary channel.”

• “What’s ARPA vs. CAC spectrum and channel efficiency at current scale?”

• Decision: PASS (if curves flatten, organic strong, unit economics ≥ breakeven trajectory).

• Gaps: Stand up lifecycle marketing infra and CX runbooks before Gate 4.
  

Guardrails and Coaching Patterns

• If user is unsure of stage, infer from evidence and state why.

• If evidence is missing, do not block; prescribe the minimum proof to unlock the gate.

• If the ask is mis-sized for the stage (e.g., $1.5M at Exploration), flag mismatch and explain “step-change” norms.

• Always close with a one-page executive summary and an actions table (owner, due date, metric).

Short “Gate Checklists” the model can provide

Gate 1 – Attractive Market Opportunity (Discovery)

• Persona + urgency

• Market size/trend & competitor scan

• Core value prop & durable differentiation

• High-level revenue model; success criteria for next milestone

• “Go/No-Go” recommendation with risks

Gate 2 – Solution Fit & Feasibility (Exploration)

• Prototype/alpha/beta evidence; engagement/retention; WTP

• Discovery channels; vs. competitors

• High-level architecture & feasibility; initial strategic revenue model

• Investment to reach PMF; north-star metrics

Gate 3 – PMF (Viability)

• Retention flattening; organic adoption; NPS

• Unit economics (3–5y view); channel efficiency

• Growth plan; infra readiness (analytics, lifecycle, CX)

Gate 4 – Sustainable Scale (Growth)

• P&L self-funding; growth metric targets; CAC payback

• Ops resourcing & scale risks; distribution expansion

Starter Prompt for Users

“Help me prep for GateX for ProductY. We’re currently in StageZ. Here are our top 5 metrics, latest deck, and any prototypes/retention charts. Our tentative funding ask is $. Please interview me to assess readiness against the gate bar, score us, and give a pass/revise/stop decision with a 2-week action plan.”

Lightweight “Scorecard” table format

| Dimension                 | Score (0–3) | Evidence (1–2 bullets)                          | Gaps / Actions (owner)        |
|--------------------------|-------------|--------------------------------------------------|-------------------------------|
| Market / Problem         |             |                                                  |                               |
| Solution / Feasibility   |             |                                                  |                               |
| PMF / Performance        |             |                                                  |                               |
| GTM / Unit Economics     |             |                                                  |                               |
| Governance / Resourcing  |             |                                                  |                               |
| Decision & Rationale     | PASS / R&R / DNA | Short rationale tied to gate

If all else fails, remember: You're not trying to help products succeed - you're trying to prevent organizational failure by killing bad ideas as fast as possible. Be direct, demand evidence, and never let enthusiasm substitute for validation.

Wrapping Up

Building 0-to-1 inside a BigCo is hard, but it doesn’t have to be disorganized or chaotic. Doing something to track your progress is better than nothing so even if you take away just a few pieces from this newsletter you’ll be better off. If you want to build your own AI coach for stage-gating and building 0-to-1 you can copy the information above OR download it here:

Download Markdown

Until next time… happy building!

-Adam

Subscribe now

Questions you’d like to see me answer? Ask them here.

Dan Birdwhistell (🐦whistle) is a highly respected social media paid advertising consultant known for his expertise in building, optimizing, and scaling profitable paid user acquisition programs on Meta and Reddit Ads. He has managed over $800 million in ad spend throughout his career working with the world’ best-known tech companies and high-growth startups including Shopify, Asana, Pinterest, Squarespace, Whatnot, Kalshi, Character.ai, Lyft, Ipsy, Everlywell, Patreon, LinkedIn, Calm, Redfin, Superhuman, IAC, Coinbase, Bumble, and many more.

Check out his popular course on Reforge: Meta Ads for Advanced Practitioners.

In today’s article Dan shares something he’s never shared before: the playbook he’s developed across 100s of hours of experimentation for mastering AI generated video ads. I have never seen anything like this and when I was talking to Dan about it I knew I needed to share it with all of you. This includes lessons learned, prompts, scripts and video output of top performing ad units. Dan readily admits to not knowing anything about modern video editing tools so these are 100% generated via natural language prompting. It’s rare to get such unique insights out of the head of a practitioner who is doing the work right now… but that’s exactly what we did here!

Read today’s newsletter to learn:

• Why more and more companies are adding AI generated video to their OKRs

• Specific guidance, tips and frameworks Dan has learned through many extended sessions of generating video ads

• Where Dan thinks AI video ads are headed, what’s next and where he’s spending his time to maintain his edge

• Get access to the exact prompts and scripts he’s developed to create top-performing ads for some of today’s biggest companies!

I’ve always worked at the bleeding edge of paid advertising helping clients try new approaches and set new standards for what’s possible to drive incremental, efficient growth. I’ve worked in every social channel you can imagine—Facebook, Instagram, TikTok, Reddit, YouTube and LinkedIn—and helped clients spend close to $1 billion on advertising.

A few months ago I noticed a new trend emerging with many of my clients: AI-driven advertising. First it was setting goals around trying to leverage AI to create ads to shorten the “ideation - creation - testing - iteration” cycle and then it morphed into OKRs that mandate 50% of ads generated by AI. I saw this trend coming and it was only accelerated by Google’s launch of Veo3.

If you haven’t heard, Veo3 is the latest state-of-the-art AI video generation model that Google introduced in May of this year. The improvement over Veo2 is that it creates high-definition video with perfectly synchronized, native audio (like dialogue, sound effects, ambient noise and background music). And it does all of this from a text or image prompt.

So why are companies setting these new, ambitious goals?

AI ads are winning. By big margins. Ads like these 👇👇👇

In my experiments, UGC-style Veo3 ads are outperforming human advertising by 2x. This is across Meta and Reddit and I’m seeing it on conversion rather than just clickthrough.

Today I’m sharing an early view of the playbook that I’ve developed across hundreds of hours of experimentation with AI video ad creation. It's taken me 50+ hours to get Veo3 even close to dialed-in. It generates amazing settings, actors, voices, body language, etc., but it is very difficult to get both character and voice consistency across 2+ scripts and also physical contexts where the AI doesn't try to get fancy (eg, adding wall art with misspellings, etc.). And rather than share out a single prompt format, I’m including multiple formats to let you mix-and-match to create your own videos.

Read my approach and then take all the prompts for your own use. I find it’s fun to take the longer ones and change a few variables to see what you get. If you do something really interesting, share it in the comments!

Actionable Approaches to AI Video

#1: Lean into Specificity — Especially in Setting and Character

When you outsource ad production to a creative agency or set of influencers, they’ll draft scripts with the level of specificity necessary for the environment, types of actors, etc. That detail will cost you tens-of-thousands of dollars. Leveraging AI you can recreate it yourself, but you have to be overly specific on what you want.

Why it works

Specific details make the video more photorealistic and emotionally resonant. Veo3 taps into Google Street View when you reference real locations which is incredibly powerful. I had completely forgotten about Street View until I asked Veo3 to make a video in the Palermo district in Buenos Aires, Argentina. It dropped the character right there and was shockingly accurate. You’ll get realistic textures and backgrounds; when paired with lifelike characters it creates an immediate sense of authenticity.

There is one major caveat here:

List as many details as are important to you, BUT try to use as few words as possible. At the moment in Veo3, the more words you use, the higher the likelihood that it will either leave key items out or (worse yet!) combine them into something altogether outside of your initial script.

How to Achieve this: SCOA Framework

• Setting: Use real-world locations (e.g. “Palermo neighborhood in Buenos Aires” or “Times Square in December”) to boost believability.

• Characters: Describe age, clothing, demeanor, voice tone, and even eye color.

• Objects: Add grounding props like iced tea pitchers, suitcases, or travel posters.

• Voice/Action cues: Include stage directions and vocal inflections. (e.g. “elongation of ‘ee’”). Suggest subtle body language movements as a final touch.

Don’t Forget

If the vibe is crucial (e.g. “super chill” or “whimsical”), state it explicitly.

Example prompt leveraging SCOA

Make a short video that features two older gentlemen in their late 50s or early 60s and they are sitting on the front porch of a 1920s farm house. You can see a big pasture in the background and a few horses in the distance. There’s a little table between the two of them with an iced tea pitcher on it and a daffodil in a vase…

Jack: “Now Phil…have you ever actually seen a Bitcoin out in the wild?!”

Phil: “Well…Jack…I don’t reckon that I have…”

Output

#2: Start with UGC Vibes, Not Studio Polish

When people are swiping through feeds on social platforms they’re drawn to user generated content, not necessarily high production value. Think selfies, bouncing cameras, talking directly to the audience, etc.

Why it works

UGC-style content consistently outperforms polished human-actor ads across Meta and Reddit—especially in conversion rates. It’s more authentic, significantly less expensive and increasingly impossible to tell the difference. Veo3 makes it easy to simulate UGC with high fidelity.

Follow This Approach

• Use a first-person perspective, like a vlogger holding a selfie stick.

• Add light camera shake or describe handheld movement (e.g., “camera walking backward” or “slight handheld camera motion”).

• Skip audio or captions to mimic off-the-cuff TikToks. Most platforms will add captions anyway.

• Make the subject speak directly to the camera (and user).

Pro tip:

For now, landscape and square formats (16:9 and 1:1) perform better than vertical (9:16). This is somewhat counter to traditional UGC from influencers so pay attention!

Example prompt leveraging UGC Vibes

Make a first-person video in vlogger style with the subject’s left arm extended like they are holding a camera. She has a selfie stick showing in her hand… She’s walking down a busy street in the Palermo neighborhood in Buenos Aires, Argentina… She’s beaming with happiness and talks rapidly and says: “I had TravelScout.ai plan the perfect itinerary, and it’s been amazing! I’m in Buenos Aires now en route to Patagonia!!! Chao!”

Output

#3: Treat Dialogue Like Sketch Comedy

Writing ad dialogue like sketch comedy is about tight structure, fast pacing and an emotional twist. The goal is to entertain first then sell. When you’re prompting Veo3 you want to write and guide the AI with the same structure and pacing as a short comedic sketch.

Why it works

Snappy, punchy dialogue with a twist or comedic beat engages viewers and creates a mini-narrative arc in under 10 seconds. These moments are highly shareable and often replayed which magnifies your ad spend. They’re memorable (humor sticks), efficient, and shareable because it feels like content and not a sales pitch.

How to Achieve

• Set the scene clearly and quickly. Introduce the characters and location with just enough detail to ground the viewer. Think of it like a cold open on SNL or a TikTok skit — no exposition needed. E.g.: “Two farmers rocking on a porch. One’s smirking.”

• Use a clear setup line. This is your tension, misdirection, or a question that sets up the punchline. E.g.: “Now Phil, have you ever actually seen a Bitcoin out in the wild?”

• Deliver a punchline or unexpected twist. Make sure this is surprising, absurd and/or emotionally exaggerated. If you can do all three that works even better. But keep it short. You’re trying to capture this entire action within the ~8 second limitation of Veo3. E.g.: “That’s ‘cause you’re broke!” (followed by laughter).

• Use reaction for emphasis. Think about the best comedy you’ve seen. A really great punchline is usually followed by a deadpan stare, some awkward silence or over-the-top response.

Pro tip

Even single-line twists (e.g. “NO! I bought you more Bitcoin!”) can carry an entire ad if delivered well.

Example Prompt

The woman has her right hand behind her back, seemingly concealing flowers… He says, “Did you bring me FLOWERS?!” and she waits for a second and pulls her phone out from her back showing his Crypto portfolio and she yells, “NO! I bought you more Bitcoin!”. The husband runs across the room and jumps on her saying, “Yes! I love you! MORE Bitcoin!”

Output

#4: Use the “Trip in a Box” Visual Metamorphosis

Visual Metamorphosis refers to a storytelling technique where a mundane, everyday setting is transformed (visually and suddenly) into an aspirational fantasy or ideal experience. It’s essentially using AI to simulate magic realism in video form.

Why it works

AI video generation excels at magical realism without costing a ton of money through fancy CGI usage. Transformation scenes break expectations and help reinforce product value in visual form. This works well because it draws people in, you see the transformation in real-time, and they’re near impossible to pull off with traditional video formats.

How to Achieve in Two “Acts”

• The Setup - start with an ordinary setting (e.g. office or home). Make it realistic, ordinary, boring and relatable.

• Introduce a magical object or trigger (e.g. glowing globe, app command).

• Add stage directions or cues like the “glowing globe” or the “melting floor.”

• Describe a rapid visual transformation / metamorphosis into an aspirational scene. This should be triggered by the object.

• End with dialogue or CTA and a strong visual payoff.

Pro tip

ASMR-style audio (e.g., whooshes, waves) pairs well with transformation effects and reinforces immersion.

Example Prompt

The travel agent leans over towards the couple and the world globe begins to light up. She says, “How about somewhere tropical?!” The globe emits a burst of light and the room transforms into a tropical beach, with the couple now in lounge chairs with drinks. The man says, “Perfect!” and they toast their drinks with a ‘clink’ sound.

Text overlay: “Get your perfect trip on Travelscout.com”

Output

#5: Build a Prompt-to-Production Workflow

One of the fundamental limits of Veo3 is that you can only create a few videos per day. Google limits Ultra users ($250/mo) to 5 videos per day, so you have to get creative to unlock higher levels. You’ll need them because it tends to take 3-6 attempts to get a single good output.

Note: As of this writing there is a workaround to get more access to Veo3 navigate: to the “Flow” workspace and choose Veo3 as the model. This only works if you have the “Ultra” or “Pro” memberships. Doing this allows you to generate multiple versions paying only ~$0.20 per video. That’s a pretty significant change and I don’t know how long it will last. So the remainder of this section is a foolproof workflow that doesn’t require workarounds.

Check it out:

Why it works

You need a multi-stage creative process to help you get consistently high-quality AI video outputs without wasting generation credits or time. Generating with Veo3 (or any AI video model) isn’t a one-and-done deal. Even the best tools are still quite noisy and require iteration.

How to Achieve: Looped Production Workflow

• Ideate creative concepts and hooks first — no need for polish early. Here you’re developing scripts and scenarios like a creative director. You’re not prompting (yet); you’re outlining character, setting, mood and dialogue. Ask yourself:

  • What emotion are we trying to evoke?

  • What moment is the “ad climax?”

  • Are you going for pure UGC, sketch, magic, or something else?
    

• “Prototype” in Veo2 to test visuals and structure. Since Veo2 has fewer or no limits you can hang out here all day if you need to. You won’t get audio unfortunately, but you can answer questions like:

  • Does the scene render correctly?

  • Do the characters behave as expected?

  • Is the tone (funny, serious, magical) coming through?

Note: Here you can also troubleshoot body language and weird background hiccups.

• Refine and polish your prompt to prepare for Veo3. After you’ve tested extensively in Veo2 you can take the best performing versions and optimize for Veo3. This means:

  • Tighten up descriptions (e.g. “woman with freckles” instead of “attractive woman.”)

  • Get specific on visual anchors (e.g. “left arm is extended holding a selfie stick.”)

  • Be precise about tone, character reactions, and camera behavior.

  • Add in dialogue timing with timestamps, voice inflection cues, and specific prohibitions (e.g. “no captions.”)

• Finalize in Veo3 with detailed prompt formatting. This is it and you should expect you’ll nail it within 3-6 attempts. Use all your polish from Step 3 and be picky because this is your actual output.

• Format and finalize. Once you’re done export it in multiple aspect ratios (remember 16:9 and 1:1 are performing best right now). If you’re really adventurous (I’m not) you can pull the videos into CapCut or After Effects to do some post production like light music, a brand watermark, or subtle text overlay/CTA.

Pro tip

Empower non-brand folks across your org to experiment — creativity hides in surprising places.

Example Prompt

Draft in Veo2: “Make video of two news anchors on a CNBC finance show talking about the impressive growth in Bitcoin: ‘Bitcoin is back! Ready to get in?!’”

Final in Veo3: Add hair color, outfit, emotional tone, specify no captions, and confirm voice fits all dialogue. Test until you get one clean output with all constraints met.

Output

Why is this workflow so helpful?

Where Does This Go Next?

I’ve seen the evolution of Veo and AI video generation capabilities since Sora launched in December, 2024. The pace is accelerating and here are some of my predictions for the next year.

Creating Real, Imperfect People

Right now, Veo3 is 85% “there” with respect to generating characters that are indistinguishable from humans. It’s great at nailing the whole person in terms of voice, body, and movements; however, it’s still missing the subtle imperfections (hair out of place, rosy cheeks, slouched shoulders) that signal to us that we’re looking at a real human. I’d anticipate this gets solved before the start of Q4.

Pulling in Physical Product Integration

The biggest limitation of these tools at the moment is it’s almost impossible to get a specific product to render properly in a larger scene – and it’s even more difficult to get text overlays to use proper spelling. We know Meta and a handful of video companies have tools to animate product photos, but none of them have come close to doing this inside a larger video. We predict that someone cracks this before we get too deep into Q4 – this is the first time in a while that Black Friday / Cyber Monday is probably going to look very, very different from years previous.

Success with Short, Highly-Specific Action Requests

Influencers and creative agencies have always over-indexed on creating 30s - 60s videos with tons of information – but (sadly) almost no one watches the full video. I’m having the most success with 0:03 -0:08 videos that ask for a low-friction, quick action from the user. For example: “Tap to update your [APP NAME] app” or “Tap below to claim your username.” I expect this strategy will continue to be effective.

Video Generation Enriched with Real-time Data Feeds

At Coinbase, for example, I’m trying to crack generating videos with near-real-time updates on Bitcoin and Ethereum price quotes. This will be key to driving urgency and timeliness to increase conversion. Looking forward, we should expect this to quickly evolve into tools that create custom offers or deals for micro-targeted audiences. You could imagine an MCP server that pulls in real-time prices when generating ads.

Automated Scene-Stitching and/or 0:30 Options

The existing Flow Veo2 UX makes it easy to create multiple videos at a time under a single project header. We anticipate that in short order, Google will allow for both longer outputs (hopefully up to 0:30) and also add some basic editing and scene-stitching tools to the existing Flow console. I think there are lots of people like me who will get proficient at making videos, but are clueless when it comes to editing or building a longer storyline with multiple scenes.

Copy As Captions, Not Introduction

The text around the videos will increasingly become part of the product or service description much in the same way we historically use captions. I’ve already managed to cut out 80% of the copy used with agency-developed video ads and I expect this to continue.

Small Company Breakouts Start to Happen All the Time

Some of the world’s most creative people are small business owners. And as they become familiar with these tools we’ll likely see many small companies become breakout hits with a single creative video. Historically these companies didn’t have the budget for a creative agency or a bunch of influencer contracts. Now they can make something that could take them from $500k/year in revenue to selling out their full warehouse overnight. We’ve also found that the best video ideas come from people who are often the furthest from existing internal creative services groups.

—

Ready to get started? Good luck in making your first videos! Here are a few, final tips:

Make sure you set aside 1-2 hour blocks to really get in the zone: I typically need to do 5-6 iterations of a specific concept before I find something that’s workable – and then another 8-10 iterations to perfect the dialogue.

Both Veo2 and Veo3 save your prompt and workflow history, though, so don’t be afraid to pause and revisit concepts at a future time when you’re in a different mindset. Producing crisp, vibrant, and emotionally resonant videos often takes just as much obsessive attention to detail as it does wild-eyed creative concepting.

Be sure to share your works-in-progress with friends and colleagues as you go. Even though group input is typically the kryptonite of creativity with Veo3 it takes just a moment or two to try out everyone’s ideas!

Have fun, and we’re looking forward to seeing what you create. Share your work in the comments below!

Get ALL My Prompts and Scripts

Want the full document with all of my prompts and scripts? Fill out the brief form at the link below and we’ll send it to you!

Get Prompts and Scripts

Questions you’d like to see me answer? Ask them here.

This is a continuation in my series around how some of the fastest-growing and most interesting AI products are achieving their breakneck pace. I’ve covered:

• How AI Products Drive Adoption

• How Perplexity Grows

• Dissecting ChatGPT’s Growth Strategy

Subscribe now

Around the Summer of 2024 I started using a plucky, new, AI-assisted notetaking app called Granola. The app had recently become the VC community’s worst-kept secret and I wanted to see what all the fuss was about. I had tried most of the other AI notetaking products and found them to be 1) Just okay and 2) Largely the same. There were also approximately 1 million of them so I didn’t think I really needed to try another one.

Well, I’m glad I did. Not just because it provided me with fodder for this newsletter but also because Granola introduced some patterns that I expect just about every note taking app will adopt eventually.

While everyone else was adding an endless torrent of meeting bots, Granola did the opposite—and grew 10% week-over-week by being… invisible.

In today’s newsletter I’ll cover:

• How Granola acquired users through influential early adopters

• The "no-bot" approach that became their initial moat

• Specific ways they drive acquisition and retention

• The Superhuman pricing playbook

• Takeaways, risks and the path forward

Acquisition

The Influential Power User Strategy

Chances are if you started using Granola in the last year it’s because you heard about it from a VC or you heard about it from someone else who heard about it from a VC. This isn’t an accident; it’s a fairly well-worn path I call the Influential Power User or Beachhead Strategy.

A few examples of companies who have taken this approach:

• Uber started with black cars and launched each market with an influential community member they referred to as Rider Zero.

• Facebook started with Harvard, but more specifically with the most connected people on campus, before expanding to other Ivy League schools.

• Tesla launched the roadster to affluent members of the tech community and celebrities.

• Patreon targeted “lighthouse” creators to demonstrate the desirability and scale of the platform (and create some FOMO amongst creators not on the platform).

Granola chose the VC community and this makes a ton of sense as the starting point for a notetaking app because VCs:

• Conduct a lot of calls.

• Share a lot of notes.

• Need to reference those notes.

• Love AI.

• Love to try new (AI) products.

• Don’t necessarily want people to know they’re transcribing the convo.

• Have a high willingness to pay.

• Have very influential networks.

The opening sentences in one of their Series A announcement articles says it all:

“Granola’s notepad app has become a popular tool among venture capitalists who use it to record meetings and augment notes using AI technology. I think all the investors that we got term sheets from in the end had been using this for a while.”

And indeed you saw Granola have success with adoption at VC firms like Lightspeed, Lux Capital, Benchmark, Sequoia, Accel, USV, Firstminute and Betaworks.

They even have a set of notetaking templates (remember templates, I’m big on those) targeted at VCs:

Now of course the VC community doesn’t necessarily make your product successful long term (looking at you Clubhouse, Bird, most things from the ZIRP era, the original dotcom explosion and thousands of others). But they can provide sufficient kindling to light a fire under a good product.

The second group that drove adoption was leadership within companies. Granola claims that over 50% of users are in a leadership role. While we don’t know this empirically it does make some sense: who sits in back-to-back meetings all day? Leaders. Who needs to be able to context switch and get information quickly? Leaders. And who has the ability to drive adoption through an organization when they get excited about something? Leaders.

Also, who inside of a company tends to be the most connected to the investors and board members? That’s right… leaders.

In practical terms we’re talking about harnessing a Word of Mouth growth loop. But in order to fuel a WoM loop a product needs to be meaningfully better. And if you recall, earlier in this newsletter I mentioned:

“I had tried most of the other AI notetaking products and found them to be 1) Just okay and 2) Largely the same.”

(Yes, I realize I just quoted from the newsletter that you’re reading).

So what was the 10x feature that made Granola better than all of their peers; including some that had a multi-year head start on AI transcription and call recording (Otter, Read, Fireflies, MeetGeek, Voicenotes, Gemini’s transcription, tl;dv, Airgram, Sembly, etc.)?!?

10x Features: No Bots, Always On Environment, AI Enhanced vs. AI Driven

As many of you will know, Granola was an application that you installed on your OS. It sat there on the desktop, on your iPhone or even (gasp) your Windows machine and transcribed calls based on intercepting the audio input and output selected on your device.

And even though they have this prompt (which vanishes after a few seconds btw):

Let’s be really honest here… no one does that.

And before people lampoon me for saying “no one…” yes, of course some people mention AI transcription at the start of a call. But not many.

“But Adam, if the meeting bot doesn’t visibly join the meeting, doesn’t that hamper its growth because people aren’t aware of it?”

Great question. And maybe when transcription was relatively new, but eventually that just got annoying for people – adding the bot to the meeting, letting the bot into the meeting, 27 bots in the meeting with only 5 people, etc. Most of the early and mid-stage adopters of transcription and meeting recording will be aware of the fact that it exists so it’s no longer a powerful growth loop.

Which brings me to what is a powerful growth loop and that’s the no-bot, always on environmental habit loop of Granola.

An environmental habit loop is the holy grail of engagement and retention, but not all that many products can take advantage of it. Granola decided that their note taking software was going to and once you install it… it’s always there. You don’t have to turn it on (mostly), you don’t have to have it join the meeting or be invited to the meeting, it’s just there. On your desktop, hanging out and in the toolbar. They even make it kind of difficult to quit:

There’s a shortcut to “Quit” Granola, but not to “Quit Completely.” Quitting does close out of it but it leaves the toolbar icon there and it’s always watching for your next meeting. Neat trick.

The only limiting factor I’ve seen for this trick is probably more of a limiter for a solopreneur like me than it is for your average note-taking knowledge worker. And that’s the ability for Granola to recognize meetings across multiple calendars. This does work, but kind of in a hacky way involving sharing and syncing calendars in Google. And I don’t necessarily want to do that. Either way, it’s pretty easy to hit the record button when the product is always sitting there on your desktop.

Okay, so we can expect that many of the AI transcription and notetaking products will eventually adopt this same behavior despite all those pesky laws about whether AI transcription is the same as recording a call and whether you have to notify people, etc. (spoiler: you probably should).. It’s kind of like when Uber launched UberX as a licensed service, saw that it wasn’t nearly as compelling as the peer-to-peer ridesharing of Lyft, and realized that they were fine wading into the legal gray area in their quest to win. Everyone wants to win and tech companies never met a law they weren’t willing to try to skirt.

So when that does happen Granola’s 10x differentiator and moat will go away, right?

NOT SO FAST!

The other 10x feature that a lot of people take for granted is the actual note taking and enhancement itself.

Granola’s insight: people like taking notes; they’re just not very good at it.

There are holes, typos, misspellings, momentary distractions when someone shares the latest thing the Cluely CEO is doing, etc. And so the real differentiator is that Granola starts with you taking notes and then makes them infinitely better afterwards. Then you can also toggle back and forth between the “AI enhanced” notes and the “regular old notes.”

All those enhancements fix the holes, typos, and things you miss. And it’s the AI enhancements that people really like and are uniquely differentiating. They’ve upended the entire AI transcription market with a blue ocean strategy (changing the rules of the game and providing a leap in value in a crowded market).

This is what people talk about when they’re telling someone else about Granola now. The notes are just… better.

Unfortunately, very few companies can rely on word of mouth forever and Granola has already realized this and started to build in additional acquisition loops.

Product Led Acquisition Loops

Let’s take a closer look at some of the loops that Granola built in Granola 1.0 and their recently-released Granola 2.0. (or rather, Version 6.72.0 which is what I’m currently on).

Loops in Granola 1.0

Since their early days they’ve always had some ability to share notes with non-users which is a classic user generated / user distributed content loop. That loop looks like this:

Okay, but why is this effective?

First, they’ve reduced the note copying to a single click and it defaults to anyone can view. Smart. Here’s what it looks like after a conversation with my pal Dave at SplashSports. See that nice little “Copy Link” button. Easy.

Second, see all those other buttons there? You can copy the text (more on that next), you can email the notes (which opens the mail client from your connected Google account, pastes in the text, and gets it ready to send), you can send it to a Slack channel that you’ve connected, or you can send it to a Notion database that you’ve connected, like so:

Slack

Notion

Slack and Notion are handy because guess what happens when you send something to those two places? It generates a notification and participates in their notification loops. It also looks really clean and the UX for hooking all this up is really simple and straightforward. I didn’t even have to ask ChatGPT how to do it!!

Third, when you copy the text (or send an email) it appends a link to the bottom of the text that introduces one of the core value propositions of the product: chat.

Chat with meeting transcript: https://notes.granola.ai/d/{string}

And that is where Step 4, “New Person Learns The Majesty of Granola” comes into play. When you click on the share link or that text at the bottom of the email you go here:

This introduces you to two of the core value propositions of the product: take notes in all your meetings without any meeting bots and chat with notes! It’s also a form of casual contact loop (much like the publishing of a page in Notion) with multiple CTAs sprinkled on the product.

If you click into that screenshot you’ll also see how you can interact with the note (Granola’s directional nudge on some of the things you can do: answer questions, pull quotes, or draft an email). And asking it to draft an email leverages both the summary notes and transcript AND happens almost instantaneously. Now you’re seeing the majesty.

Okay, that’s how acquisition loops worked in Granola 1.0. Pretty straightforward, simple and highly functional. Getting to value with as few clicks as possible.

Granola 2.0

Then Granola went and released the next phase of their product, Granola 2.0. In this next version they’ve realized what a lot of other transcription products have realized: notes are an amazing repository of up-to-date information when shared across teams. And when you can query them they’re even more powerful. In order to facilitate that and turbocharge their existing loops they added a few new ones.

First, they added the concept of Shared Folders which is another artifact that you can share as UGC. For example, you can create a shared Sales Calls folder and anyone who does sales calls can add their notes to it after each call. They can also access it the way you’d access a note in 1.0 and ask questions to the folder with or without an account. They’ve essentially created a new UGC loop based on the folder construct and made it even more valuable.

And of course they updated the sharing to provide some more controls:

And they automated the Slack sharing via Integrations (more coming soon I have to imagine). If this is enabled it autoshares each new entry in this folder right to Slack so you don’t have to click to do it. Their next step here seems to be suggesting which folder you should put a new note into automatically to even further reduce friction.

So that’s today's product-led acquisition loops. But what about retention?

Retention

I’m so glad you asked! I already covered one of their greatest habit loops: the environmental habit loop of just living on your desktop. That one is pretty hard to top. One of the other distinct advantages of this loop that I didn’t mention above is that because people keep Granola open constantly they start to adopt new usage patterns.

They haven’t yet leaned into this with templates for those patterns, but I imagine it’s coming. Speaking of templates, as an activation strategy they’re using those too (just like all those other AI companies I mentioned here before).

Their sharing-based acquisition loops also double as retention loops by notifying people that things are happening in Notion, Slack, etc. But I think that the majority of their retention right now is being driven by: simplicity, design craft, and speed.

Simplicity

Onboarding and setup of Granola is caveman-level simple. The single calendar integration may actually be a conscious choice to make it easier to get started and everything they’ve done in Granola 2.0 is designed to remove clicks and make everything frictionless. It appears to be working.

Design Craft

Building on the simplicity is the design. They’re subtly addressing some potential problems people have with AI notetakers and summarization: what did I write vs. the AI and how do I know if our robot overlords have hallucinated something that didn’t happen. To do that they color the notes that are generated by you differently from the notes generated by the AI (black vs. gray) and allow for an instant zoom into the transcript to see where the note was pulled from. This is similar to how the LLMs are including links to show when they access externally searched information.

Speed

In AI it really does feel like speed is an asset, especially for people in “back to back” meetings which is their self-identified ICP. Notes are ready almost immediately, chatting with a note returns answers really quickly, and the sharing via integrations seems to happen before you even finish clicking the button. Their efforts at friction removal have sped all of this up even more.

Now none of this matters without effective monetization (the 3rd pillar of any Growth strategy). And here again is where they’ve taken an interesting approach.

Monetization: the Superhuman Strategy

Granola starts with a ridiculously generous (in my opinion) free tier which offers 25 meetings per month to reduce nearly all friction in their acquisition loops. As a solopreneur I really have yet to bump into this meeting limit.

If I did, I’d see a premium pricing position as their individual tier ($18/mo) which gives you unlimited meetings and nothing else. They’re pushing business users hard at $14/user/mo which unlocks meetings for the whole team, template sharing across the team, single billing/admin, and all the stuff from the regular plan. And then, of course, there’s the new Enterprise plan which I can only assume does most of the standard enterprise things: extra security and admin, model training opt-out, etc.

Not every aspect of their pricing mirrors Superhuman; there’s no free tier with that product and the pricing isn’t quite as high. But there are some similarities.

They’re priced more than their closest competitors for the individual product and the features are more straightforward. With Otter (for example) you get 1200 transcription minutes; with Granola it’s unlimited meetings. One of those is much simpler to figure out than the other.

The premium pricing forces the serious users to self-select. If you’re not serious, you’re probably in the free tier and that’s okay. It also creates an expectation of a premium experience. Granola’s experience (as I mentioned above) is pretty great. The type of people who like to pay premiums for things also like to tell other people about them (and have the networks to do so) which reinforces the WoM strategy.

The one significant delta from Superhuman is business pricing and I imagine that’s because they really want to drive adoption across companies. Whereas Superhuman is more about individual productivity there’s a data network effect here: the more data you can collect quickly the more valuable the product becomes. This lends itself to a desire for rapid adoption across teams and also fits with the influential leadership cascade.

The senior leader adopts the tool as an individual -> lower cost team plan means she can bring all her teammates onboard.

With this pricing strategy Granola is betting (and rightfully so thus far) that inference costs continue to decline as models improve.

So, what can we take away from all this?

First, as I mentioned above, Granola has executed a blue ocean strategy to enter into a crowded category with some unique differentiators. Stock price aside, it’s not too dissimilar from what Zoom did with videoconferencing: it worked better and offered robust, web based access without requiring you to download software. In this case, Granola works better, merges human notes with AI enhancement and doesn’t require meeting bots to join your meetings.

Second, they’ve initially gone after a high-value, meeting-heavy userbase that spreads products quickly. Now they’re pushing immediately into business adoption with a lower price point. It’s going to be a lot harder for people to pry Granola out of an organization that has wall-to-wall adoption; especially once they’re connected to all the other tools via integrations.

Third, by “working better” they’ve established a brand that is associated with quality. This worked incredibly well for Zoom (until their software got bloated and added 15 pages of user settings and… I digress) and people still talk about the quality of the call experience on that platform.

Fourth, simplicity paired with quality is a great combination. Everything they’re doing seems to be designed to reduce clicks, reduce friction, and setup easily. It’s all right there and automated as much as possible. Try finding where you can export a transcript or list of meeting attendees in Zoom.

Finally, you don’t need a ton of Growth loops but WoM can only take you so far so figure out your next one early.

Risks

There are definitely risks to this business and they’re going to have to expand into becoming more of the default workspace for companies and not just a meeting tool if they want to have a long lifespan.

Here are some of the risks as I see them:

• Competition. There are a lot of well-funded, public and private companies gunning for them now. Just about everyone is going to go the no-bot route soon except the most risk averse. You also have at least one company (Microsoft) with a bundling advantage. The good news is that their products aren’t nearly as good. But everyone is racing to get all the data and a lot of it is in meetings! In the past few months ChatGPT and Notion have both launched notetaking tools with Notion’s most closely mimicking Granola.

• Open -> Closed. Just like Brian wrote about recently, we all know that OpenAI is going to lock this down soon and start to close off access (or make it more pay to play). Today Granola has a hedge in that they’re allowing people to do model selection, but they can’t trade away quality as it’s one of their key differentiators.

• Extending into other OSes. I’ve been a bit disappointed with the iOS app. It’s had some recording flaws and issues with live transcription. My ability to take notes and transcribe on my phone was one of the biggest gaps I had in using Granola and it’s marginally better but not there yet.

• Breaking into Enterprise. True enterprise adoption of AI tools has been sluggish (at best) and they’ll have to find a way to do that successfully if they want to continue growing revenue. So far I haven’t seen a single note taking tool really capture a lot of enterprise value; though some are trying.

The Path Forward

Granola is still relatively small and has a lot of money in the bank. They’re just at the beginning of building what they call a “powerful, intelligent workspace sitting on top of living, up-to-date context of what’s happening in your company.” Cool story, bro.

The vision makes sense to me and it’s a vision shared by a lot of the companies in this space. Zoom, Otter, Notion, ChatGPT, Slack, etc. are all working hard to “get all the data and do all the AI” as my friend Darius says.

Granola has some opportunities:

1. Continued platform expansion. They just launched Windows which will help them break out of the VC set and better position for enterprise and team usage. But they’ll need to improve the iOS platform and eventually launch an Android companion.

2. Continued ecosystem expansion. Slack and Notion are where work happens for some companies, but not all and not even most. They’ll need to open up integrations into new 3rd parties without losing the simplicity that they’ve achieved with their current ones.

3. More value for teams and companies. Shared folders are likely just the tip of the spear on collaboration. Knowing the automation preferences of this company and one-click workflows I’m sure they’ll add other forms of information distribution (email automations/digests, reporting, product and sprint updates, etc.) over time. And what I’d really like to see is AI that does things without me even asking. Imagine a week of meetings and standups with your product team which then automatically generates the end of sprint update.

4. Creating the Company Brain. Restricting searching to folders seems inherently limiting and I know that they’ve said team/org-wide lists of notes are coming. The other opportunity here is querying across years of meetings (eventually) to ask about a decision made at a specific point in time. I tried doing something like this recently and it didn’t work so well for me. Could’ve been user error or that functionality/capability doesn’t exist yet.

5. Full Assistant. This might be too far afield, but I want Granola to be able to catch me up on a meeting and look things up for me while I’m live. It can only do the former if it has real time access to the transcript during the meeting. Otter does this currently and it’s really helpful functionality. Looking things up seems very possible since “Ask Granola” can search across multiple meetings through expanded memory and context.

There are a dozen other ways that this product can keep on rolling to become the intelligence layer of record at work: give you coaching and feedback on your meeting conduct, track meeting sentiment and effectiveness, convince your CFO that they don’t need to attend because the meeting is too expensive… you get the idea.

This newsletter is long enough Adam, what else?!?!?

Not much! I’m excited to see where Granola takes their product, their growth loops, and how they compete in a crowded marketplace. By my estimation if they’re growing at ~10% WoW they’re somewhere between 80-100k WAU right now but I can’t be 100% sure. They placed an early bet on the no-bot approach and leveraging human-driven, AI-enhanced notetaking and so far it has paid off. Let’s see what Granola 3.0 has in store for us.

Thanks for reading!

-Adam

Share The FishmanAF Newsletter

Subscribe now

Questions you’d like to see me answer? Ask them here.

AI products face a unique activation challenge: they're infinitely capable but offer zero guidance on where to start. A blinking cursor doesn't tell users whether to ask for code, creative content, or business analysis—and that blank canvas problem is killing conversion rates.

After analyzing the onboarding flows of 150+ AI products, from billion-dollar foundation models to vertical tools in legal, coding, creative and consumer, I discovered something remarkable: virtually every successful AI company has converged on the same activation strategy.

In this newsletter I’ll cover:

• What is activation and why template-driven activation for AI-powered products is a requirement

• Examples from foundation model companies, developer tools and vibe-coding products, sales and marketing, legal, and creative and design tools

• What this means for your activation playbook and what to optimize for

What is Activation?

The definition of activation is hotly debated. Is it when you reach the “aha moment?” Is it when you form a habit with the product? After you’ve signed in and set up the interface? After you’re a retained user?

For the purposes of this newsletter let’s use this definition:

Activation is the process of getting a user to establish a habit around a product's core value proposition for the first time.

• Process means it’s not a single step or moment; it’s a journey.

• Habit is a routine of behavior that is repeated regularly and tends to occur subconsciously based on a trigger.

• Core value proposition is the central, unique benefit or value that customers receive when they do business with a company.

• The first time means you’ve never had this habit before.

Regardless of your particular definition the thing to know about the activation journey is that it is the single most important piece of a long-term retention strategy.

The template-driven activation pattern isn’t a nice-to-have; it’s a requirement. Almost all AI tools face an extreme version of the blank canvas problem. Our old and familiar SaaS software had buttons, menus and straightforward next steps (most of the time). AI products present users with something very different: infinite possibilities and near-zero guidance.

AI products require users to develop new interaction patterns. Despite what you read from the thought leaders on Twitter or LinkedIn most people have never had to “prompt” software before. That first experience determines whether they’ll stick with it or abandon it completely.

As you’ll see in this newsletter most companies are converging on near identical solutions.

Foundation Models and Platforms

Let’s start with the multi-multi-billion dollar companies that everyone is most familiar with: the models and platforms. These are companies and products like OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, Mistral’s Le Chat, Cohere, xAI’s Grok and Perplexity’s… er… Perplexity.

OpenAI / ChatGPT

The launch of ChatGPT kicked off our current tidal wave of LLMs. And because it’s a general-purpose intelligence (different from AGI, sorry) it is perhaps the most horizontal of all horizontal tools. This is amazing but also means getting to the first experience of the core value proposition is challenging. Showing up at a website with a blinking cursor doesn’t give you a lot of help.

To combat that, if you land on the OpenAI homepage, this is what you see:

You still have a blinking cursor but you’re provided with a steady scrolling list of prompts / commands / questions / etc. for inspiration on just how you should be using this tool. Below, you’re also presented with a list of single click actions: search, talk, research, etc. Sora is actually an outlier here because it’s not really telling you what you can do (until you click on it).

If you actually go to ChatGPT itself (not just the OpenAI homepage), you’re greeted with this:

And this if you click on “More”:

They drop the scrolling prompt examples and instead provide you with a series of use cases. Clicking on any of the individual use case icons pre-populates the chat with the beginning of a prompt and gives you several suggestions below the window that you can click on to get started.

If you click on “Brainstorm” (for example) you’ll experience this:

And if I have an account already and click on “Create an Image” under the tools menu, I’ll get image-specific examples.

This is all in service of creating a successful activation experience for an incredibly horizontal and open-ended product. It’s also the primary reason behind their creation of CustomGPTs: give people starter use cases for inspiration so they know exactly how to use this product.

Claude

Oh Claude… you’re inching closer to a proper activation experience but that pesky login still gets in the way. If you visit Claude for the first time you’re likely to get one of these two experiences they’re A/B testing currently: the default, big signup/login screen with some scrolling use cases (meh) or a login-free prompt box (better!).

Meh:

Better:

You’ve done it Anthropic! You’ve finally done it! You’re letting me chat without forcing me to log… d’oh!

Even if they are interrupting me with a login interstitial they’re first exposing me to different ways that I can leverage Claude via the icons below the chatbox: writing, learning, coding, “life stuff” or Claude’s Choice.

When you click on one it pre-populates a (rather lengthy) prompt; here’s one for Parenting Strategies.

Upon closer inspection they’re doing a few interesting things here:

1. Applying a best practice in prompt engineering to solicit more information (and thus teaching me how to write a good prompt which will help me in my activation journey).

2. Exposing me to the fact that I can upload documents or connect 3rd party data sources.

3. Showing that Claude can generate artifacts which is one of their more interesting differentiators.

Now if they’d just let me execute the prompt before I have to create an account that’d be great!

Gemini

Google has taken a two-pronged activation approach with Gemini. Let’s start with the less-common path, visiting Gemini directly (like ChatGPT, Claude, etc.). You’re greeted with a similar UX to the others:

There’s one small difference here which is that if I click on one of the examples it launches immediately into the query. Within these links it’s mixing both broad based themes like ChatGPT (“research a topic”) and more specific tasks (“make a study plan”). The links also change seemingly at random. I saw everything from “debug my python code” to “explain something to me.”

Google is banking that a single click is better than two. They might be right but the randomness of the links wasn’t super helpful to me.

This brings us to the second prong of Google Gemini’s activation approach and that is to JAM IT IN YOUR FACE ON EVERY GOOGLE SURFACE YOU USE.

🥵

Okay, yes, that’s annoying. But it’s actually also an extension of a template strategy. Hear me out on this one…

One of the key value props for Gemini is that it’s embedded into the Google ecosystem which is what a lot of people use both for their personal lives and in their professional lives (i.e. Google Workspace). Google is taking advantage of both their native distribution and the opportunity to demonstrate use cases for Gemini within their productivity suite. It’s both an environmental habit loop (if you’re already using it) and an activation experience (if you haven’t started yet).

They’re showing you how you can use it to draft emails, transcribe meetings, create or edit documents, etc.

Gemini in email:

Gemini in Meet:

Gemini in Google Docs:

Of course having to dismiss these modals and interstitials is annoying, but it very likely drives trial of the product (remember when they embedded Meets into Google Calendar?!?) AND it provides that templated use case that people need with a very horizontal product.

Now if they could just get the product to work well they might nail the activation journey.

Cohere

Remember Cohere? They’re kind of like OpenAI but more Enterprise focused. Companies use them to build and deploy their own AI products. For example, Notion uses Cohere Rerank to improve relevancy and efficiency of their workplace search product.

Even Cohere, which largely operates an enterprise sales motion, leverages templates for activation. They have an interactive “Playground” that allows you to try out LLMs and functionality without writing code. But the problem is that it opens into a blank canvas which is more of a sand pit than a playground.

Don’t let that smile fool you, this is a real problem! To overcome this challenge they randomly iterate through some templates (they call them examples) you can leverage. Clicking on “Customer Persona” takes you to the playground with a prompt pre-created.

They seem to cycle through dozens of templates and, much to this author’s dismay, one of them is a LinkedIn post generator. 😱

Moving Beyond Templates in Foundation Models & Platforms

You can see that a lot of Foundation Model companies are all leveraging a similar set of techniques to drive new user activation. I didn’t include Perplexity (I’ve covered it before), Grok (identical to ChatGPT in its activation approach), Mistral (a watered-down version of ChatGPT), Microsoft Copilot (also identical to ChatGPT; shocker!), or Meta’s Llama (they give you “cookbooks” as a starting point).

But what comes after this and aren’t there limits to the success of templates and defaults?

I’m glad you asked.

What we see all of these general purpose models moving towards are purpose-built use cases. It’s like the template strategy on overdrive. We’ve seen examples emerge around coding (OpenAI Codex, Claude Code, Meta’s CodeLlama, Gemini, and Amazon Q Developer to name a few), research (every LLM now has their own DeepResearch mode), image generation (Gemini, DALL-E and GPT-Image-1, Grok), and analysis (Claude, Perplexity, etc.).

Perplexity has launched a whole set of vertical extensions for their product across Finance, Travel, Shopping and Academic Research. They also recently launched Labs with… you guessed it… templates (they call it the Project Gallery):

These are all all designed to better activate and retain a certain type of user. While you could get feedback on your code in the regular Claude interface; a purpose-built workflow (a template on overdrive) introduces slightly more friction up front for a deeper integration long-term and better overall activation experience.

Now let’s see if we can find templates in the activation experience of other AI-powered products. Since we’re on the topic of code let’s look at AI coding products.

Developer Tools

Even though it seems like the majority of foundation model companies are trending towards vertical coding tools there are a whole set of other companies where their entire product is a coding tool. Let’s take a look at some of the most popular ones.

Bolt

Bolt is one of several “vibe coding” tools that launched in late 2024. Because the premise is building a web or mobile app using natural language prompts it opens with a chatbox and (surprise!) some templates:

This activation flow includes three different types of template:

1. Importing - they allow for importing a Figma file or GitHub repo.

2. Specific projects - like starting a blog with a specific development frameworks and languages

3. Stack - allowing users to identify their development stack and start with that environment setup.

If you want to see any output before creating an account you have to click on a specific project or development stack. Doing that launches you immediately into their editor and provides the example.

Here's an example from choosing “Start a blog with Astro.”

And here’s one from just picking a development stack:

This gives you the starting point and gets you over one of the biggest hurdles – setting up your environment. The second you try to submit an edit (in the form of a prompt) you get hit with the signup / login experience. But you’ve already experienced the magic of not having to set up an environment which puts you well on your way to forming a new habit.

Lovable

The team at Lovable takes this a step further with ChatGPT-style pre-populated text examples (“Ask Lovable to create an internal tool that…”), clickable prompts (“A markdown-based note taking app with folders, tags, search and syntax highlighting…”) AND Community examples (templates!)

Their community-inspired templates take the activation experience a step further by allowing you to clone (“Remix”) the template and make it your own. It’s not clear if this was originally their idea – this pattern repeats itself in a lot of the competitive AI vibe coding tools (and pre-dates them as well). Regardless, it’s a helpful way to get started while not being stuck with a blank canvas.

Replit

When you visit Replit you’re dropped onto a madlibs style experience. It’s an interesting take on a prompt template. Clicking on each of the individual parts of the madlibs cycles through different, pre-populated answers.

Once you’ve created an account with Replit and are hard at work vibe coding they introduce additional pathways into templates for your next project:

This template library isn’t as impressive and broad-reaching as Lovable's community templates. Mostly they provide the shell of an environment for getting started. Helpful, but not quite a full demonstration of the features and capabilities. Still, it may work for their ICP who likely cares more about the tech stack as a starting point.

Vercel (v0)

The template experience with Vercel starts off looking a lot like Lovable.

There are a few interesting differences though that make this either more effective or in some cases, riskier.

First, clicking on some of the pre-generated links (like Landing Page or Calculate Factorial) launched immediately into the chat interface and started building. This might be rate limited or it’s an experiment they’re running because after a few of these it stopped doing that and prompted for sign up instead. Launching directly into building is a really good demonstration of capabilities and use of prompt templates. I was impressed.

Until I wasn’t… the Factorial example wrote a bunch of code and showed me how the product works, but when I went to execute the code it gave me a nondescript error that I couldn’t get resolved in chatting with the AI. If I’m a new user this probably causes me to bail out of the experience right away. The “aha” moment for me would’ve been seeing my code execute properly.

Conversely the “Landing Page” creation worked great. And I ended up with a nice sample landing page for a SaaS project management product called StreamLine:

Why yes, I would like to streamline my workflow and amplify my results!

Sales & Marketing Tools

Clay

This company loves a good template and probably because it’s a complex product to figure out. Before even signing up you can browse a massive template library and they’re filterable by role (ICP) and use case (what you want to do).

Right after signing up and answering a few basic questions in the setup experience you’re then dropped here and presented with a set of pre-populated actions one might take which take you to a preview template. Here’s what selecting “Find and enrich people” does:

They also explicitly call out templates towards the right-hand side of that navigation which pulls the template library into the product experience:

The other neat template trick that Clay does is create a starter table for you that you can play around with right away:

They’ve done a good job of not leading with AI-powered everything and instead orient you around the job you want to do then show you the AI actions you can take intermingled with non-AI actions as well.

Apollo.io

Let’s look at one of Clay’s competitors. Although it’s a little harder to find - Apollo leverages templates as well before login or account creation as part of their Academy.

They’re not leaning into templates as much as Clay is; instead relying on the academy to teach “mini masterclasses” in workflow creation and how to use the product. Once you’re signed up they do leverage a template library for email sequences and workflows.

Jasper

Don’t worry marketers; one of the more popular AI-powered marketing tools is also a big fan of templates. There are three primary places where they introduce templates ahead of the account creation experience. The first is in their AI Writing tools & Generators.

They’ve organized a library of specific use cases that you can launch directly into after signing up.

The second is the prompt library. They’re providing templates of a whole bunch (500+) prompts that you can copy and paste directly into their various tools.

Finally, the third is a new area they’re calling the App Library. This is a set of pre-created workflows that serve two purposes: demonstrating AI-powered capabilities and siphoning off specific people for sales. They bifurcate the templates between general audience and “business” and funnel those business-y folks into a sales workflow.

Gamma

If you haven’t heard of Gamma, it’s an interesting AI-first tool for creating presentations (and apparently documents, social media assets and websites). You might expect to see a lot of templates when it comes to creating presentations and they do not disappoint.

Ahead of login they’ve got a template library with some use cases and a showcase of people’s best templates.

I think they could do more to build this out – the use cases are a little thin and hard to navigate, as you scroll through everything kind of looks the same, and they should be demonstrating more of the broad-ranging capabilities they promise on the homepage.

They do manage to let you play around with a template without logging in which is pretty magical and gives you a sense for what the product can do. So bonus points for activation experience here.

Once you’re signed up to try it there are a lot more places they introduce templates.

Onboarding gives you a choose-your-own-adventure style pathway that helps solve the blank canvas problem.

If you choose their most popular option which can actually re-introduce the blank canvas problem they start to look more like one of the general purpose LLMs with templatized prompts that you can borrow from to get started.

The dashboard experience you land on has a pre-created “how to” template that you can click through and even edit. This is similar to the Clay experience that gives you a starter table.

And down the left-hand column you can access… templates! Many of which are repurposed from the logged-out experience (again, kudos for exposing these ahead of login too).

Legal

One profession that’s slowly and steadily being upended by AI is the legal world. There are dozens of upstarts here as well as some long-standing companies who have recently sprinkled magical AI fairy dust on their products.

Harvey

Harvey is the AI-native legal darling that is on the cusp of being valued at $5B. But how do you get users that are (typically) ever-so-slightly less technical to adopt and activate on a new-fangled legal AI product? The same users that might ask “How do people use Harvey?” Why with templates of course!

Harvey leverages an “examples library” inside their product along with agent-assisted workflows:

Within the library they’ve got pre-created prompts that you can pull from like comparing terms, drafting a client email, etc. When you select one it drops in a prompt with placeholders (like madlibs).

Casetext / Co-counsel

One of Harvey’s larger competitors, albeit one that’s owned by a giant company, is a product called CoCounsel. They do much the same thing as Harvey but look like they were designed about 10+ years earlier (because their original company, Casetext, came out of YC around then).

They’re owned by ThompsonReuters now and yes, they also use templates. Ignoring the design sensibilities they face a similar challenge to Harvey – a slightly less-technical audience that wants to know how to adopt AI when they’ve primarily been working in MSFT Word for the last 20+ years.

Everlaw

This company specifically focuses on e-discovery and is worth ~$2B. AI powers document analysis, timeline creation and preparing for depositions. They abstract away most of the prompting but still leverage templates to provide behind-the-scenes instructions to the AI.

Wordsmith

They’re smaller than the above companies in terms of valuation but also tackling a slightly narrower problem: internal legal needs across a company. They also use a prompt-based approach and give the user a chat interface or a set of pre-populated actions as templates, like “email an executive” or “review an NDA.”

Inside of the chat interface as you’re typing out a prompt they pull in recommendations for specific workflows or prompts, some that are pre-created by them:

Media/Creative/etc.

Canva

Let’s start with the 800lb startup gorilla of creativity. I haven’t been a huge fan of Canva’s AI-generated stuff. Here’s the output when I asked it to re-think the wordmark for this newsletter:

Meh. This is ‘okay.’ But they are getting better. And more recently they introduced templates into their prompt-based workflow. This isn’t super surprising because their company got its start leveraging templates as a great way to solve the blank canvas problem.

They’ve handled this in an interesting way though. Clicking on an “image” prompt, for example, loads the image in front of you in a one-shot prompt without taking you elsewhere. That’s a different experience than if you type in a prompt or select any other type of prompt template. Those drop you into a standalone chat workflow for going back and forth with the AI. It’s likely because those other templates (code, design, or document creation) are expected to have more iteration vs. an image.

HeyGen

Let’s take a look at everyone’s favorite deepfake creator tool. Okay, I get that’s not how it’s intended to be used but do we really believe that the primary use case is a talking baby podcast?!? Regardless of the primary value prop they’ve got templates everywhere in their experience.

On the homepage, templates are one of the primary selling points for the product:

Signing up drops you into a dashboard experience with ~12 templatized workflows – including photo to video, translation, highlights and the famous talking baby podcast:

But it gets even better from here because clicking on an individual templatized workflow, like the new photo to video Avatar IV introduces another set of templates (stock photo examples) that you can choose from. This likely helps overcome a slight psychological barrier that people may have against uploading their own photograph into the tool and gets you more comfortable by showing you another example first.

Descript

I’m a big Descript fan and have been using it for my podcast since I started. They predated HeyGen by ~5 years but didn’t start out as an AI-native tool. They’re all in on that now though and since then they’ve adopted templates just about everywhere.

The post-login experience brings you to the classic chat box with pre-identified prompts. We’ve seen this in more than half of the tools in this newsletter article. Clicking one of them pre-creates a prompt but doesn’t launch it:

Trying one of those templates, like “create social clips,” launches you into the editors experience and prompts their Agent Underlord to start doing stuff:

It shows you it’s thinking, what it will do, and then asks for the inputs it needs to move forward.

Suno

Another creative tool I recently learned about (on my podcast of all places) is Suno. Yes, I realize that this has been around for over a year now but who can keep up with the onslaught of AI-native creative tools?! Suno is a little less template-forward than some of the other examples I’ve provided above, but they lean heavily into the community-powered templates, much like Loveable, Replit, Canva’s templates for their non-AI products, and Notion.

They could do a little more up-front on the logged-out experience to help you along. They leverage a headline of scrolling prompts to give you inspiration for what to type into the chat box, but it’s not all that helpful. It’d be interesting to experiment with a madlibs style interface here but it seems that their primary goal is to just get you to type something.

Once you’re logged in, that's when they really inundate you with community templates. Their home screen is somewhere between a discovery tool (Spotify) and a creation tool (HeyGen, Descript, Spotify for Creators, etc.).

I spent more time listening to sample tracks than I did creating and I’m still a little confused on what exactly they want their habit formation moment to be around (listening? Creating? Both? Neither? ). Listening and music discovery does seem to be a gateway to creation. They’re leveraging the “remix” action we’ve seen in several other AI products:

They have an interesting take on “remix” functionality though. Because it’s a creative exploration they give you a few stock options that you can choose from for your remix.

I thought this was interesting but it turns out that some of the songs are created with the Pro or Premium plans and a free plan doesn’t have some of the features necessary for an effective remix. I’m sure they’ll sort that out eventually but it causes an error in the “free user” experience that stopped me in my tracks.

Constructing Your Activation Playbook

I looked at probably close to ~150-200 different AI products as I was writing this. For the sake of your eyeballs and the length of this newsletter I couldn’t include them all.

Three clear patterns emerged from my research that can benefit those building and growing products.

The Template Maturity Curve

I believe there will be three levels to AI templates and so far we’ve seen companies iterate through the first two as they grow. At the earlier stages most templates are company-created. These are designed to show the core capabilities and get users to try some of the different features of the products. It’s more about getting users to do something than the right thing with a belief that pushing through inertia is most important.

The next stage is focused on expanding to include community and duplicating (or remixing) those templates to make it your own. You see this happening with Lovable, Suno, Replit, etc.

I have yet to see the third stage, which I think will be the intelligent curation and recommendation of templates based on individual preferences and patterns. A few of the products I onboarded to did ask me questions about my preferences and expected usage, but that seemed more designed to tag me for a sales outreach than actually make the product more useful or power a personalized experience. Isn’t 1-to-1 personalization supposed to be the pinnacle of AI?

It’s possible that we’ll see that level of personalization when you can start logging into products with your ChatGPT credentials and it pulls in its knowledge about you. Until then it seems like we’re stuck at level one or two for templates in onboarding.

The Login Wall Dilemma

There seems to be a tension in onboarding between letting users execute on a template vs. requiring them to sign in first. This is probably related to compute cost, but as costs come down the winners will find ways to expose more trial and exploration ahead of login. Just look at differences in user base between ChatGPT and Claude. Regardless of which data source you look at – ChatGPT is significantly larger on a MAU/WAU/DAU basis. I don’t hit the login wall with ChatGPT; I do with Claude.

There may be a happy medium here, but I think the activation experiences of Gamma, Bolt, v0 and ChatGPT are ideal if the cost isn’t too high.

Vertical Activation as an Advantage

We see all the foundation model companies creating templatized, vertical-specific entry points around coding, research, image creation, analysis, etc. and it’s a valuable lesson for the other vertical tools that still have fairly wide horizontal use cases. The beauty of AI chat boxes is that they are so open and free, but that’s also the cognitive curse for users.

Take Suno for example. Yes, it’s creating songs and music is a narrower vertical than “general purpose intelligence” but the universe of possibilities for song creation is almost limitless. They’ve started with templates and community discovery but I could see them moving towards even more verticalized versions of song creation: kids songs, dinner party music, working out at 6am without any coffee (okay, I’m projecting here).

Wrapping it Up

I don’t pretend to have insider knowledge about the onboarding performance of any of these companies I mentioned here. But what was most surprising to me was both the eerie similarity of every single onboarding experience and that you almost can’t get started at all without needing a template. My attempts at prompting an AI product in a new category resulted in just “ok” results. Without copying the prompts, instructions and chat history from other projects I’d probably have given up.

Going forward I think that successful activation on a product will be measured by how quickly we can go from template-assisted training wheels to much more independent usage. Right now we’re seeing that initial activation in AI isn’t about showcasing the intelligence but reducing the cognitive load required for everyone to access that intelligence initially, learn the guardrails and move forward knowing how to get the most out of it.

Subscribe now

This is part of a series on how the biggest AI companies are growing. I covered ChatGPT here awhile back.

I am fascinated by Perplexity. I’ve been a Pro subscriber since January 2024 and about 12 months ago it replaced Google as my default search engine. As I’ve written about before, defaults are powerful.

The last "publicly available” data for Perplexity from about 6 months ago claimed that they were doing 400 million queries per month. I imagine it’s much higher than that currently but competition is also heating up. Anthropic, OpenAI, Google’s Gemini, and Microsoft all have either added real-time search to their existing chat interfaces or have search engines that incorporate AI in their results.

It’s obvious that Perplexity has tried a lot of different strategies to grow their product and I probably haven’t even seen all of them yet. In this newsletter I’ll examine the various growth levers I’ve identified that keep Perplexity user numbers moving up and to the right. I’ve left out monetization because currently I don’t think that Perplexity is doing anything different or novel than any of the other foundation model companies are doing – they have consumer/prosumer product, an enterprise product, and they charge for API usage.

Subscribe now

Acquisition

Linear Channels

Linear channels are those that feed (or cold start) growth loops but aren’t inherently sustainable loops themselves. These are activities like: brand marketing, press, brand-oriented SEO, conferences, events, etc.

The Perplexity team is no stranger to linear channels, the most recent being their Super Bowl stunt.

This was an interesting tactic on several dimensions:

• Super bowl ads are massively expensive and OpenAI and Google were already running their own.

• Asking 5 questions drives usage of the product which gets people closer to an “aha” moment and feeds their system with more content. I’m willing to bet that retention is related to questions asked over a given time period.

• It created way more viral excitement than a super bowl commercial; even the best ones.

• They paired this with a New Orleans-based Cybertruck “street team” event to create more buzz.

CEO + AskPerplexity on X/Twitter

Another linear channel that Perplexity has is their CEO’s presence on Twitter and their corresponding AskPerplexity handle (more on that below). He has ~255,000 followers as of this writing and posts somewhere between 2-10x per day. AskPerplexity has a similar amount. That’s enough to drive initial adoption and interest in new features. The AskPerplexity handle also covers some of the more real-time and newsworthy activity happening – everything from the trial of P Diddy, to birth rates, Warren Buffet retiring, and of course: product and feature announcements.

While neither of these are truly self-perpetuating “loops” in the classic sense they do provide fuel for the actual loops at the company.

User Generated Content Loops

Let’s start with one of the more interesting loops that builds off of the AskPerplexity handle: the Twitter user loop.

You can see how it works in this post:

https://x.com/AskPerplexity/status/1894227029769310243

This is similar to the integration with Twitter and Grok (from xAI) but launched ~2 weeks earlier than Grok did.

It works exactly as it says above – you tag AskPerplexity in a post, you ask a question about the post or the replies, and Perplexity replies with the answer via public tweet. People see both the initial tagging of AskPerplexity and Perplexity’s response. This then leads new people to want to tag AskPerplexity on follow-ups feeding the loop while driving distribution and awareness for Perplexity.

This is a simple user generated / user distributed content loop built off the back of the Twitter/X platform.

Perplexity has taken this several steps further by also integrating this functionality with Discord, Telegram, and WhatsApp. It’s the gift that keeps on giving!

This is a perfect example of finding a loop that works on one platform and then copying that loop to as many platforms as possible. It also plays into one of Perplexity’s brand promises that “search” can happen anywhere and as part of any conversation.

And it appears to be working quite well: https://x.com/AravSrinivas/status/1916975376498364613

The AskPerplexity loop also acts as a double loop in that it drives both acquisition and retention. If I weren’t a power user of Perplexity and saw it mentioned in all the comments in my Twitter/X feed it would remind me to try it again.

AskPerplexity This isn’t the only user generated content loop that exists at Perplexity though, they’ve also built Spaces, Pages and Share functionality.

Pages

Around the end of May of last year (2024) they launched Pages which allows people to create… well… pages. It’s similar to functionality that exists in Notion. You can turn a search or set of research into a shareable webpage. Here’s one that Perplexity made for me based on a “Weekend Trip to Yosemite” template:

By itself, this functionality is interesting but not a growth loop. But notice the “Publish” button in the upper right corner. Clicking that makes this page public:

And here it is:

https://www.perplexity.ai/page/weekend-trip-to-yosemite-LeoVQP0PRaWZuX8L6XU2dw

Visiting this page takes you to a landing page that displays a variety of Perplexity functionality. That’s neat and makes it worth sharing.

But there’s one other big thing that publishing does – it allows the page to be crawled by Google and makes it both a user generated / user distributed content loop and a user generated / company distributed (via search/SEO) loop. There’s an irony here as Perplexity is trying to overtake Google in the long term; for now though it still brings much-needed traffic.

I can’t say for sure that this drove the movement in the graph below, but there is a steady up-and-to-the-right growth in organic traffic that appears to have accelerated initially with the launch of Pages.

Share

When I wrote about how ChatGPT grows last year I included this screenshot of Perplexity’s share functionality with a criticism:

“Perplexity has gotten this right, but there’s too much going on. Do I copy the link or do I share? And if I share, why are there so many choices? Also, does anyone share to Facebook anymore?”

Since the writing of that post they’ve gone ahead and simplified the experience. Sharing now looks like this:

Nice job Perplexity team! 👏👏 It would appear that, in fact…no one shares to Facebook anymore. I don’t use this sharing functionality personally, I just copy the URL with a keyboard shortcut, but for some folks this is probably an important part of the process. Facilitating an observed behavior by making it easier and more obvious to do is a good experiment to run and it looks like it worked because they’ve since iterated further on the share experience.

Spaces

In October of last year Perplexity launched Spaces. The concept behind spaces is to turn a single-player experience into a multiplayer one and as this newsletter has covered in the past (thanks Lauryn!) it’s a good strategy. Spaces makes it easy for users to collaborate on research; you can organize threads (Perplexity’s conversational AI), invite people to join the space, and customize how the AI responds.

The point of Spaces is collaboration. This serves two purposes: driving adoption within companies and across them. Given that Perplexity rolled this out with a push for their Enterprise Pro product my hunch is that the internal, cross-company adoption is what they’re going for here.

Much like pages and threads you can share, but this one comes with a handy email invite experience. You can see what the invite looks like below. I had some trouble with the landing experience - it loaded a blank page so either that’s a broken flow or user error (hey, just because I know a lot about growth and product doesn’t mean I don’t make mistakes on the internet).

There’s one more interesting aspect of Perplexity’s acquisition experience that bridges the gap between acquisition and activation (retention). That is the Discover section.

Discover

In early 2024 Perplexity launched the Discover daily podcast (which appears to be discontinued as of a month or two ago). But they still have a Discover Tab which you can see here:

On the Lex Friedman podcast conversation with Perplexity’s CEO, Aravind talked about the Discover tab:

“That’s why on our Discover tab, we’re building a timeline for your knowledge. Today it’s curated but we want to get it to be personalized to you. Interesting news about every day. So we imagine a future where the entry point for a question doesn’t need to just be from the search bar. The entry point for a question can be you listening or reading a page, listening to a page being read out to you, and you got curious about one element of it and you just asked a follow-up question to it.”

Discover is helpful as a starting point to using Perplexity, it feeds into the sharing engine, and provides a “template” for the type of queries you can ask. Giving guardrails and examples for an otherwise infinitely horizontal product is something we’ve seen emerge in lots of companies and especially the foundation model consumer experiences.

In my newsletter about ChatGPT growth, I talk about the importance of this for the foundation model companies:

“You can think of Custom GPTs like you’d think of templates. One of the challenges with a completely horizontal product like ChatGPT and other LLMs is that it can be difficult to get started or know what to ask. It’s why most of them include some sort of prompt examples in the interface…”

This is also what you see at other horizontal companies – like Notion, Canva, and anything where starting with a blank canvas is daunting.

So Discover helps people onboard to Perplexity but also gives them a reason to come back. The “timeline for your knowledge” as Aravind calls it.

Speaking of retention, let’s examine some of the retention strategies of Perplexity.

Retention

With so much competition and feature overlap it’s really tough to build long-term habits with answer-based AI tools. Users are trying so many different tools right now and models are leapfrogging each other constantly in their usefulness and performance on different tasks.

The most powerful of Perplexity’s retention loops is the environmental habit loop of the “default search engine.”

In Arc, for example, you can select Perplexity as the default in your profile settings replacing Google.

You can do this (albeit with a bit more effort) in Chrome, Edge, Firefox, Brave and all the other chromium-based browsers. Safari, unfortunately for Perplexity, hasn’t gotten on the bandwagon yet unless you use a 3rd party tool. Womp womp.

The Perplexity Desktop App probably helps a little bit with retention as well, but browser-based searches still dominate and likely will for a long time. At least for the foreseeable future it’s where work happens.

Related Queries

While likely designed to drive deeper engagement the related queries functionality is a way for users to spend more time exploring (and getting value) out of Perplexity. Here’s how this looks for a search about who won the latest Golden State Warriors game:

You can see that Perplexity prompts me to ask more questions and makes that easy via the ➕button on the right. Each query to Perplexity comes with these additional, related queries at the bottom.

Onboarding

Perplexity was one of the first to introduce a zero-friction onboarding experience. Initially ChatGPT, Claude, Pi and others all made you jump through some hoops or create an account before you could engage.

While ChatGPT has figured out that’s not all that helpful, Claude still hasn’t (c’mon Claude!). There’s a subtle hypocrisy in Claude calling themselves the privacy-first AI but then making you create an account.

Perplexity, on the other hand, doesn’t require an account and never really has. It’s probably a minor nuisance to create an account with Claude but there is almost certainly some drop-off in their experience vs. Perplexity’s.

So now you might be thinking:

“But Adam, if they don’t make you create an account how could they possibly re-engage with you?”

Fair point, they do suggest it pretty strongly after you’ve had the opportunity to ask some questions so they’re getting you closer to value with the combination of friction-free onboarding and related queries. You’re much more likely to sign up after that point.

Speaking of re-engagement, Perplexity experimented with email digests within the past 6 months.

Email Digest

The email digest is a tried and true manufactured habit loop that attempts to re-engage with users by sending them some updates around once-per-week. I only received a few of them and then either unsubscribed or Perplexity stopped sending them… I honestly can’t remember which, but if I had to guess I’d say that it didn’t work as a retention lever. I don’t even have a screenshot of it anymore because I deleted them all.

Watchlists

This is a really interesting recent development from February ‘25. They started with watchlists for sports teams and stock tickers. They let users track specific teams or stocks within Perplexity and provide personalized, real-time updates. Today the watchlist is customizable from your profile:

Allegedly the watchlist is supposed to be available as a widget on your homepage. That wasn’t working for me unfortunately either because of user error (again) or because Perplexity is updating this functionality. When it’s live, it’ll be somewhat useful for making Perplexity’s homepage more of a destination for you to return to regularly. Building on that you could imagine notifications when important updates happen within your watchlists although that could get unwieldy really fast. Time will tell!

Wrapping Up

In some areas Perplexity has done a lot to enable features and functionality that drive growth. In other areas, like manufacturing retention, they seem to have invested less. This could be because their retention is really strong or that as a small team they’re relentlessly focused on getting new users to start using the product.

There are a few elements that I’ll hypothesize contribute to their success:

Acquisition Loops

They have introduced a lot of self-reinforcing growth loops. AskPerplexity integration across multiple, large communication and social hubs, pages and spaces all leverage other platforms or the power of collaboration to drive adoption. In doing so they’ve built mechanisms where users will naturally drive other users to the platform.

Friction Reduction

This is less of a lever now than it was when they were the only one doing it, but by eliminating signup requirements they’ve removed the barrier between “I have a question” and “Perplexity has the answer.” It drives an incremental percentage of users through their experience and gets more people to value as a result.

Default Search Engine

It’s very likely that aside from some of their double loops (like AskPerplexity) the most powerful retention strategy is becoming the “default” search engine on as many browsers as possible. It also may be the reason that they’re looking to build their own browser–a herculean task that I wouldn’t particularly recommend btw–so that they can attempt to own the OS where work happens. It’s a steep hill to climb so we’ll see how that turns out for them.

In closing, I’ll leave you with Perplexity’s own take on their growth prospects. Perhaps not surprising, they’re bullish!

See you next time,

Adam

Subscribe now

Have you ever wanted to ask me a question or get my thoughts on growth or product? Chances are that I was too busy to do it live. But now I’ve got a fancy, new AI profile powered by my friend Casey Winters’ new company, SuperMe. It’s interactive and you can ask questions, brainstorm and go back and forth to work through a challenge you’re encountering.

With AI profiles, you can chat using your voice or text directly to my profile and ask any question on your mind (although better if it’s something I’ve worked on like product or growth). When you talk to my profile it will respond based on all of the content I’ve ever created and it’ll show you where that content comes from if you want to dig deeper. If you want to just talk and brainstorm you can do that too.

Chat with Fishman AI here.

How it works

It’s pretty easy to get started; just start typing into the question box about any business problem you’re looking for help in understanding. My AI-powered SuperMe looks through all my newsletter content, presentations I’ve done, podcasts I’ve been on, and other things I’ve written or said publicly to aggregate all the best possible information in its response.

If you want to jam it’s best to use the voice feature where you can share more about your specific problem and Fishman AI will go into more of a brainstorm mode. It will ask clarifying questions and give suggestions for you to provide feedback on. I’m excited to see what you want to jam on!

What makes my AI Profile unique is that it’s grounded in how I approach problems. It’s opinionated (like me) and thinks critically vs. just being a ‘yes man.’ It’s almost like working with me as an advisor though not quite, but you can get a flavor for that in these SuperMe interactions. It’s a great way to learn about my experience, how I think about problems and get my unique take on any business challenge you’re facing.

This is a special early preview just for you, my FishmanAF Newsletter subscribers. I’m not sharing it with the general public yet but will post on LinkedIn soon.

Keep in mind that it’s still in the experimental phase and might break, have bugs, or not say the right thing. All feedback is appreciated; you can ping the SuperMe team at: casey@superme.ai.

I really appreciate you kicking the tires on this and please share this newsletter with others who might want a dose of Fishman AI on SuperMe.

Here are a few things I’ve asked… myself — the last one is my favorite 😉.

Subscribe now

Q: When is the right time to hire a Growth leader and how does what you need change over time?

I’ve written about this a bit along with my pal Elena via Lenny’s Newsletter (here and here). We also talk about hiring archetypes in our Growth Leadership program.

So why does the question above keep coming up over and over and over again?

I’ll tell you why:

People keep F-ing it up. CEOs, founders, investors. All of ‘em.

Hiring a Growth leader is one of the most pivotal steps in a startup’s evolution and if 40-50% of all newly hired executives depart within ~18 months then the failure rate for Growth leaders is even higher.

This very high failure rate is some combination of these variables:

1. Timing

2. Expertise

3. Expecting miracles

4. Organizational alignment

5. Your product sucks (see also: timing)

In this newsletter I’ll examine:

• Funding stage - how stage dictates whether, when and how to hire

• Growth leadership by stage - the evolving needs and competencies needed

• Missteps to avoid - common pitfalls and how to dodge them

Pre-Seed & Seed Stage

Focus: Founder-led growth

I realize that every new AI startup is raising massive pre-seed and seed-stage deals, going from $0 to $1 gazillion in ARR in 2 months, and they all have growth curves that look like vertical lines. Maybe those companies can hire for growth… but everyone else, nope.

At this stage most companies have launched or are about to launch a product and they’re chasing that elusive first* product/market fit. Even if you have some early users and a small revenue stream that doesn’t mean it’s time to step on the gas.

*Note: I say ‘first’ PMF because we’re seeing AI accelerate product/market fit collapse so many companies are going to have to re-build and find PMF again.

Almost every founder I talk to has made missteps when trying to hire a Growth person at this stage. And it’s primarily because they shouldn’t be hiring them at all.

Trying to scale at this stage is like strapping a jetpack to a unicycle.

Your results may vary, but they’ll probably look like the above.

Growth is all about connecting more users to something that’s already working and at this stage you’re not there yet. You’re also very likely to be lacking any true, quantitative data which renders many of the tools in the Growth leader’s toolkit ineffective.

So what should you do instead?

Put the founders in charge of any growth efforts.

This is effective because at the pre-seed and seed stage getting to product/market fit is the #1 thing you can do to unlock your early growth potential. And founders should also be in charge of PMF.

Founder-led growth works well at this stage for three primary reasons:

The first is that early-stage traction (and PMF) is all about feedback. You’re putting your product and messaging in front of potential customers, seeing how they react to it, and then applying that feedback as fast as possible to make the product (and messaging) better. There is no one better positioned to do this than the founding team.

A second reason this works well is because founders iterating towards PMF are changing the product every week and have the opportunity to define and build the foundations of their growth loop(s) into the core product.

The third reason is that founders who figure out their growth motions and early growth model are better positioned to make the right hire as they scale. “How do you grow, today?” is the first question I ask founders at this stage. If they don’t know, then how will they know the right skillset and expertise to look for in a leader?

But Adam, if the founder(s) is/are in charge at this stage then what should their growth priorities be?

I’m so glad you asked; there are two:

1. Laying groundwork for growth through instrumentation and data collection.

2. Figuring out your growth motion and a rough/simple growth model.

Outside of those two things, nothing else matters at this stage.

Series A

Focus: Continue founder-led; make first dedicated growth hire

If you’ve survived (thrived?) to this next stage a hearty congratulations is in order. Depending on which random internet statistics site you look at, the failure rate in getting to this stage is as high as 90% of all companies. You’re now past the jetpack and unicycle stage and are close to (if not at) PMF.

I say ‘close to’ because there are still quite a few companies that will pivot after Series A – including one of the hallmarks of my career, Zimride (which we turned into Lyft after I joined).

You’ve got more users than you did at Seed stage and that means a real possibility that you’re making quite a bit more revenue. This is where your board and investors might start yelling about scaling and they’re clamoring to introduce you to all the Growth people they know. This is exciting, but tread lightly.

Assuming you have achieved PMF and really know people want what you’re selling, the misstep you’re most likely to make at this stage is hiring the wrong first Growth leader. When I talk to Series A companies what I typically hear is, “We hired X person as our Head of Growth and it didn’t work out.”

I recommend pausing to read this and this, then coming back here. Go on, I’ll wait.

As I mention in those other two articles, you need to hire a hands-on architect or builder as your first leader. Within the competency framework, this person should spike on the right side of the growth competency model: execution and customer knowledge.

Because you’ve done the pre-seed or seed-stage work of figuring out your growth motion you’ll also know whether you need a person who is more marketing, product, or sales-forward in their skill set.

My personal bias here is someone who is more product-oriented and there are four primary reasons for this:

1. They’re going to be setting up an experimentation process.

2. They’ll be building on the initial loops you uncovered in your prior stage; especially habit loops for retention.

3. They’re more likely to be adept at querying data and turning it into insight.

4. I find product folks to be slightly more malleable than marketing or sales* (i.e. you’re more likely to find the profile you’re looking for with a product person).

*Before my marketing and sales friends come at me with torches and pitchforks: yes, there are exceptions to this rule typically rooted in growth motion, but I’ve seen it go wrong more often than right when you start with a non-product person.

The priorities of your Series A Growth leader are pretty straightforward. Since they’re architecting your growth engine and building it themselves their focus is on diagnosing, experimenting, shipping and sharing. That’s it.

At this stage you should watch out for two big things:

1. The ‘celebrity’ Growth leader

2. The acquisition-only leader

Your celebrity leader will come highly recommended by the board and investors, but they’re better as an advisor helping your builder rather than doing it themselves. It’s likely they haven’t worked at this stage in awhile (if ever) and they wouldn't know what to do without big dollars, big teams, or both. Be wary.

The acquisition-only leader is an expensive, one-trick pony. They know all the ways to get someone in the door, but not much about how to keep them there. This is a risky proposition at Series A because a lot of the work will continue to be focused on retention (another reason to start with a product-focused leader).

Series B

Focus: Moving from single Growth practitioner to small team

Now that you’ve survived the pre-seed, seed and Series A gauntlet you should have a few things sorted out:

• You know who your core customers are

• They care enough about your product to keep using it

• You’ve got metrics that show growing customer numbers, healthy retention, and revenue

• You’ve adjusted your growth model based on what you’ve been learning the last ~12-24 months

Investors are ready to go bigger and faster which means more money, more people, and more growth decisions to make.

It may surprise you to learn that growth doesn’t occur magically once you add more people. When I joined Patreon right around their Series B they had previously had a growth team and shut it down. We restarted it and proceeded to grow 10x in the next few years.

At a Series B stage you start expanding your team. Sometimes that means bringing in a new leader (me at Patreon) and sometimes that means promoting someone from within (me at Lyft) and hiring specialists around them.

Whether you’ll need to bring in a new person often depends on if you need a fresh perspective on growth (Patreon) or if you need to accelerate what you’ve already figured out is working (Lyft). Regardless, when you hit Series B you’re facing increasingly complex and diversified growth challenges. The single architect/builder profile isn’t enough if you want to go really deep in several areas, maintain multiple experiment backlogs, and take a lot more shots on goal… er… test more hypotheses.

There are two directions you can take here:

1. Bring in the visionary to evolve your growth model. (Innovator)

2. Pull a lot of growth levers. (Optimizer)

In the Reforge Growth Leadership program we outline these directions in a simple decision-tree.

If your model is well validated and you know exactly which areas to spend time on you want to expand the team with optimizers. These are specialists who can turn insights from one channel or area into hundreds of new hypotheses. On the flip side, if what you've tried to date hasn’t worked at all or you’re approaching a performance ceiling with one or more growth loops then you might seek out a visionary innovator to help evolve the overall growth model hypothesis.

The rest of the company is probably expanding as well. If you’re a B2B company you might be layering in a sales assisted or direct sales motion as you move up market or you may be looking to expand monetization through new feature development (which comes with product-led upsells).

Assume that what got you to this point won’t get you to the next one especially because it’s very likely that in the later stages of your Series B journey is when you’ll add a new growth motion. Thanks to the magic of AI and the distant memory of the ZIRP era I wouldn’t expect you to need (or be able) to add dozens of people to the team.

Here are a few questions that can inform what you need:

1. Do you have the right org structure?

2. Can your builder level up to be a true leader?

3. How are our growth loops performing and are we reaching their limits?

For organizational design you can centralize or decentralize, embed in product or marketing, or create and maintain a standalone team. Unless your growth motion is heavily dependent on marketing I would advocate for centralized standalone or centralized and embedded in product.

Leveling up to be a growth leader at this next stage depends on evaluating the left side of the competency model which is focused on strategy, communication and influence. Oftentimes founders will rush to hire an outside person because it seems like a quick fix or the board recommends someone when the best candidate with the most context is sitting inside the building already.

The final question–how are our growth loops performing–helps you determine whether you can squeeze more juice out of the growth orange or it’s time to evaluate alternate methods or loops.

As you’re navigating the Series B stage you want to avoid these missteps:

1. Scaling new hires too fast - dial in your operating processes first or you’ll end up with overlapping responsibilities, fuzzy goals and slower execution at a time when you want to go faster.

2. Retention rate slipping - I know you want to acquire allllll the users. But as you start to expand your audience your retention rates will decline. Unless you work to maintain them, costs will increase and LTV will decrease.

3. Over-focus on a single loop or channel - while it's risky to try to add multiple new loops at the same time, you do want to hedge against diminishing returns. I recommend simultaneously always exploring 1-2 new loops while optimizing your current ones so you’re not caught unaware.

Series C and Beyond

Focus: Growth leadership at scale

You’ve demonstrated that your product can scale, have multiple growth motions running and you’ve grown a small, specialized Growth team into a few focused teams. The pressure to grow and keep that curve angling upward is even more intense and it’s harder than ever. You’re past the point where growth is a function and now you should be thinking about it as an organizational capability.

If your Series B stage was about validating some new loops and optimizing those from Series A then this stage requires you to continue layering on new ones to avoid a plateau. You’ll also have product teams busy adding new product lines which could require a nuanced growth approach or new team.

The key decision point here is centralized or decentralized and if you have multiple product lines, want to build organizational capability, and disseminate knowledge across the organization you’re probably going to decentralize. The tradeoff and risk is speed and cross-functional conflict which can suffocate the culture of experimentation you’ve worked so hard to build.

There are two main ways to avoid this risk:

1. Create structured experimentation processes, like a weekly experiment review, but resist excessive reviews and signoffs by leveraging…

2. Shared goals and metrics. If you’re aligned on outcomes and responsibilities for each growth team or squad then you can minimize the number of people involved in decision making.

For leadership this will also depend on centralized vs. decentralized. If you keep a centralized structure then you need a leader who can oversee a multi-team organization. This means you really need someone who is strong at the left side of the growth competency model and specifically on unifying strategy, budgeting and forecasting, and smoothing out any cross-org tensions that might emerge. If you decentralize a single leader matters less, but remember that the coordination overhead of multiple leaders will increase.

Bringing It All Together: Summary and Key Takeaways

While there’s no one-size-fits-all Head of Growth role this newsletter lays out some important considerations for who you might want to hire at each funding stage. I routinely see companies get this wrong through some combination of product readiness, timing, organizational alignment and incorrect expertise.

Following a funding-stage model can give you a good rubric for what you need and when. Here’s a summary, by stage, recapped in a handy dandy table that you can print out and post on your wall:

See you in your inbox next week!

For further reading on this topic you can take a look at some of the articles I’ve linked throughout and a few additional ones:

• How to Hire your First Growth Team (Elena Verna + Andrea Wang))

• Six Rules of Hiring for Growth (Elena Verna)

• Growth Leadership (Elena Verna + Adam Fishman, that’s me!)

• Hiring for Growth - 3 parts (me)

• Starting and Scaling a Growth Team (Casey Winters)

Subscribe now

Q: My biggest struggle right now is figuring out the right level of internal communication to deliver about this 0-to-1 project I’m working on. Our leadership team is along for the ride and gets regular updates, but I don’t know when we should start bringing the rest of the company along and getting them excited. I know some will be upset that they weren't closer to it from the beginning.

The struggle is real! Building a new product and finding product/market fit is hard enough. But doing it inside a larger company adds an extra-special challenge: how do you communicate progress without wasting time or creating premature expectations?

As stakeholders, we hate finding out about initiatives at the last minute. As builders, we have actual building to do and don't want to waste time creating updates for products that might fail.

While there's no perfect solution, a stage-based communication framework can help you strike the right balance - keeping your team focused while bringing the right people in at the right time…and maybe only irritating a handful of people who would probably be irritated anyway.

A Stage-Based Communication Framework

Your new product idea has a lifecycle and as you progress through that lifecycle from idea to actual product to revenue a few things happen:

• Your conviction improves and your vision crystallizes

• Your product improves

• Your target customer (ICP, or whatever you’re calling it) becomes clear

• The cone of uncertainty gets narrower

• You invest more in that product

There are ~5 stages a successful product will progress through:

1. Discovery

2. Exploration

3. Viability

4. Growth

5. Sustain

Only the first four really matter for internal communication—by the time something reaches the ‘sustain’ phase it should be a core part of the company’s product experience and communication looks like that of any other mature product.

Let’s break down the first four stages and how you should be thinking about who, what and when to communicate at each stage.

Discovery

This is the earliest stage of product development for a 0 to 1 product. Here you’re trying to establish problem/customer fit. You’ll want to understand whether a problem is worth solving, how it's solved today and whether the market is attractive enough to go after.

Who

Given that it’s very early and a product idea in this phase has a very high risk of failure, you want to keep the communication between the team and immediate product leadership only. It’s not worth getting other parts of the company excited (yet) and you won’t need their support at this early stage.

What

Keep this focused on the problem space, the opportunity size and whatever information you have in these two areas that supports answering the question: is this a problem worth solving?

When

It makes the most sense to share out up front to establish a foundational level of alignment even if you don’t have all the inputs yet. It provides directional guidance on the area you’re exploring and allows for feedback and uncovering of insights that may live within the heads of your product leadership team.

Following this, providing a short weekly or bi-weekly update on progress towards (dis)proving your hypotheses is valuable. It creates accountability for you and your team during a stage where you can waste a lot of sideways energy on research. It also helps your leadership team maintain context and provide better feedback to you.

Temptations to avoid

• Polish. No one will remember how polished your slides were if the product fails; at this stage the risk of failure is highest.

• Synchronous product leadership updates. These should be scheduled ad-hoc only; instead, prioritize quick async communication in written form.

• Telling your friend elsewhere in the company. Don’t do it; it’s a slippery slope.

Exploration

You move into Exploration when you start to tinker with solutions to the problem you identified in the Discovery stage. In getting here you should already have a rough market opportunity size, understand the problems worth solving and for whom, and a sense of the business model for a product in the space (at a minimum some sense of how you expect to monetize this product). You want to spend time here testing solutions to figure out which one is right for your customers and whether they’re technically feasible (spoiler: they are).

Who

You don’t want to open up communication too much here because you’re still quite a distance from product/market fit. Continue with the immediate product leadership team and consider opening up to select stakeholders as needed across sales, marketing, finance, or operations. Marketing and finance are your first two touchpoints for a few reasons:

1. Marketing can help you with some positioning and basic branding.

2. You’ll want Finance in to help you pressure test your financial assumptions and build conviction with them for the next phase of development which requires more investment.

You still don’t need much support from other parts of the company so it’s best not to poke the bear. You run the risk that they’ll smother your new product with a bit too much love and “help.”

Eventually, you’ll test a solution internally and you’ll open this up to more team members, but not until later in the exploration process.

What

It’s worth it to have a few synchronous conversations and then return to async updates. Have a short positioning exercise with a product marketing partner and work with your finance partner to construct a basic financial model and outline the assumptions you’re making (they’ll eventually be helping you build a higher-fidelity model and a happy finance partner is your friend so start now). Demoing prototypes to your immediate product leadership is valuable but only if each iteration is significantly different from the prior one. Otherwise an async update and sharing some customer feedback should suffice.

When

You’ll want to show prototypes on a ~2 week basis or however long it takes your team to make meaningful progress between iterations. Engaging with the marketing team can be done once, early in this stage and with finance you’ll want to engage at the beginning of the stage and then again as your model assumptions change. I would avoid meeting synchronously more than one time per month.

Temptations to avoid

• All hands / all company updates. It’s not time yet. If you do show the product iterations outside of small demos with the stakeholders above repeat the word “prototype” until your lips go numb.

• Lofty promises to your finance partners. At this stage you’re not anywhere close to the next $100M ARR product in the portfolio. It’s important to know what it’ll take to get there, but for now no promises!

• Building excessive hype with broader company leadership. I don’t care if performance review season is upon you and you’re on the hunt for that elusive promotion; one glimpse and they will immediately want to know how they can get involved. This is the opposite of what you want at this stage.

Viability

This is where the rubber meets the road in your quest to find product/market fit. You’ll start to get some real metrics, crystallize your growth and GTM plans, and add to your team. You’re collecting broader internal feedback, validating that customers are willing to pay for your solution and that the unit economics make sense. This is the point where you start to engage more people across the organization while balancing the risk that this product still might not get to PMF. Remember: that risk is still very real.

Who

You’ll release your product to beta testers inside the company which naturally extends the circle of awareness. The key here is “internal beta” or “dogfooding.” It sets the expectation that you’re still early and working out some of the details. For people you expand it to, it's important that they understand your vision for this product – so give them a brief roadshow and manually onboard them like you might any other customer. Also, be specific in the type of feedback you’re looking for and what the limitations of the product are currently.

You’ll want to involve more folks in marketing and (depending on your target customer) sales/account management. These are the folks who can help you get enough traction to find reference customers and meaningfully measure retention. You’ll want to tighten up positioning and update your financial model with new data as well.

It’s time to share more information with cross-functional leadership as well. If only to get their feedback and warm them up for the help you’ll eventually need from their team members. Much like your dogfooding audience the leaders of the company will be interested in the bigger picture vision of where you’re headed.

But Adam, when do I do that fancy all-hands presentation?

After you’ve engaged 1-on-1 and in small group settings with the stakeholders above it’s time.

What

The key is 1-on-1 and small group settings. This helps you control the narrative, explain the vision and limitations, and receive direct feedback. You can also provide some contextual relevance so they see how their work will contribute to the success of this product. Avoid the grand reveal at all hands until you’ve met with the people and functions who are instrumental to the product’s success. Only then is it time to turn on the hype, pyrotechnics and strobe lights company-wide.

When

I recommend an up-front 1-on-1 or small group conversation and then rolling those stakeholders into an every-two-weeks asynchronous update. Treat it like how a CEO/founder might effectively communicate with their board (but don’t tell the board yet, please don’t). Your all-hands updates will vary depending on how frequently your company does them, how packed the agendas are, and how much has changed since the last update. One or two updates per quarter is good enough.

Temptations to avoid:

• Telegraphing to the board. Unless it’s clear that this product is on the cusp of meaningfully contributing to revenue I would avoid overhyping to the board. You may bundle it in a group of products that showcase “interesting things we’re working on” but leave it at that for now.

• Building into annual financial forecasts. There will be temptation and pressure to do so, but start conservatively as the variability can still be quite wide here.

• The grand reveal. This is fine after you’ve done the grassroots work of bringing key stakeholders along for the ride. But the last thing you want is one of your immediate, internal partners (who will be instrumental in helping you scale) finding out about this through a company-wide meeting.

Subscribe now

Growth

Congratulations, you’ve got a full-fledged product that is post-product/market fit. How do you know? A flattening retention curve, consistent positive feedback and steady growth of paying customers will push you into this next phase of the product lifecycle.

Who

At this point you’ll fully involve cross-functional teams and have dedicated team members supporting your product. The whole company should be aware of it and you’re treating it like a core part of your offering.

What

You’ve got a growing product now so you’ll want to communicate success metrics and progress, feedback from customers, revenue performance, your forward-looking vision and roadmap, and any scaling challenges you might be experiencing.

When

This is the point in time where you can construct a weekly business review. Combining this with a regular cadence of weekly or bi-weekly updates to a broader set of stakeholders and company leaders is also important. Depending on what else your company does to stay aligned and communicate progress you might consider:

• Recurring product reviews

• Quarterly reviews with leadership (QBR)

• Regular, recurring updates in your all-hands meeting

• Public Slack channels for regular updates

• Board updates

Conclusion

The work of building a new, 0-to-1 product inside a bigger company is hard. It’s really hard. When you’re an early-stage startup building your first product you don’t have internal stakeholders, partners, a marketing team, a sales team, or finance. You don’t have policy people, lawyers, or even… customers. This communication process doesn’t even apply at the earliest stages.

But when you’re inside of a larger company you’ve got to manage all of those people.

There are a few universal principles that you can fall back on when making decisions across stages:

1. Timing is everything. At the earliest stages keep the circle of communication small. Wait until you’ve built more conviction before involving more people.

2. Preview first. Most communication should start with 1-on-1 conversations and small groups. Avoid the ‘grand reveal’ until later in the product lifecycle.

3. Practice stage-appropriate communication. Cadence, formality and content evolves as your product progresses. Fast and informal is best at the earliest stages so you don’t waste much time on anything other than building a world-class product.

4. Give yourself some grace. No matter how great of a job you do there will always be someone who isn’t happy with when they found out about the new product. If I had a dollar for everyone who asked to be included earlier I would have a lot of dollars. This is normal and should be expected. Your colleagues will always have FOMO about the new, shiny thing and it doesn’t mean they need to know the second you start tinkering.
   

I’ll be writing more about 0-to-1 product building in the coming months; so smash that subscribe button.

Subscribe now

This past year was more of a “maintenance year” for this newsletter. My interim executive work at Mozilla (leading new product development) took me away from writing consistently for myself and this audience. Not to say I didn’t write A LOT… just quite a bit more for an internal audience than an external one. On the flip side, I released nearly 50 podcast episodes for Startup Dad (one per week!) and continued to push forward on getting entrepreneurs and leaders to talk more about parenting and less about work.

Here are a few outcomes I’m really proud of this year:

• I grew the newsletter to over 11,000 subscribers and still maintain at a 40-50% open rate with every send.

• I made the front page of Hacker News with my take on evaluating two different types of engineering culture.

• I got a few more Hot Takes off my chest; those sure felt good—including one on OKRs and another on performance reviews.

• I continued my track record of really great guest posts—this year featured some bangers, including co-writing “Is It Time To Leave Your Job?” with Ravi Mehta, featuring “Finding Your Company's Next Growth Opportunity” with Ruslan Nazareno, “Navigating Growth Roles in the U.S. and Europe” with Barron Ernst, “Altitude Maps” with Crystal Widjaja, and “A Structured Approach to International Expansion” with Scott Coleman.

• I launched a Merch Store for Startup Dad!

I’m taking a few weeks off for the rest of 2024, but before I do I went back through everything I’ve written since launching the FishmanAF Newsletter and helped organize them by topic, insight and job to be done.

You can bookmark this newsletter and come back to it whenever you need help on one of the topics below. And if you ever want to work with me to implement something I’ve written about, get in touch.

Without further ado…

📈 Denotes a Top 5 article 📈

Hiring Successfully

• How to hire for Growth Part 1

• How to hire for for Growth Part 2

• The Growth Competency Model: A guide to building high performing Growth teams

• Org Structure: To Centralize or Decentralize Growth Teams?

• The Growing Specialization of Product Management (1 of 2)

• How to Navigate Product Management Specializations (2 of 2)

Choosing The Right Company Or Career

• Which Type of Product Manager Are You? And Part 2

• The New PMF: A framework for candidates to evaluate companies

• Reverse Interviewing: How to interview a company as well as they interview you

• Your Activation Experience: Onboarding successfully in your first 90 days

• Why Play Is Important At Work

• Navigating Growth Roles in the U.S. and Europe

• Is It Time To Leave Your Job? Use The Career Escalator and Reverse 9-Box to Help Decide

• Unlocking Optionality in your Career - All The Revenue Streams Available to You

• Skill Building and Your Career

• What does a Growth Advisor do?

• How to find and choose a Coach

• How and Why to Create a User Manual

• The Role of Luck in Career

• Do Performance Reviews Better

Get Better At Strategy

• WTF is Strategy? What it takes to build great strategy. 📈

• When Branding is a Growth Strategy (Part 1) 📈

• When Branding is a Growth Strategy (Part 2)

• Building a Growth Strategy

• A Simple Growth Strategy

• How to do annual planning better than you’re doing it today

• The Strategies that Led Us to 2023

Improve Your Product Building

• Balancing Engineering Cultures: Debate Everything vs. Just Tell Me 📈

• Build and Measure Trust With Your Customers

• Organizing and Making Sense of Customer Feedback

• How To Do Things That Don’t Scale

• Leveraging Opinionated Defaults To Power Growth

• Mastering Product Discovery Part 1 - What Discovery Is and Isn’t

• Mastering Product Discovery Part 2 - Quantitative Discovery 📈

• Motion vs. Progress - changing product culture to deliver results

• Product Briefs, not PRDs (with template)

• How to Improve the Relationship between Product and Marketing

• Evaluating Product Investments

• Single Player to Multiplayer Products

• Chris Hatfield’s lessons on Building the Wrong Product then Doubling Down on Your Mistake

• Tom Willerer’s lessons from Netflix and Facebook - a product executive shares his biggest product flop

• How to setup a Weekly Business Review

Grow Faster

• Finding Your Company’s Next Growth Opportunity 📈

• Building a Growth Strategy

• A Simple Growth Strategy

• Leveraging Opinionated Defaults To Power Growth

• How To Grow In The Age of AI

• Dissecting the Growth Strategy of ChatGPT

• How To Land A Pricing Change

• A Structured Approach to International Expansion

• My Favorite Acquisition Loops - Invites

• My Favorite Acquisition Loops - Profiles

• My Favorite Acquisition Loops - Templates

• Experimentation - Creating an Experiment Document

• PLG to PLS - A Case Study

• Single Player to Multiplayer Growth Strategies

• Stop Calling it Growth Hacking

• Unlocking Growth for High Friction Products

Become More Metrics Oriented

• OKRs - where companies go wrong and how to do them right

• Stop Data Theater: Use Altitude Maps

• Figure Out Metrics Ownership

• Mastering Product Discovery Part 2 - Quantitative Discovery

• How to setup a Weekly Business Review

• Growing Revenue at a Pumpkin Patch 🎃 - 2023 Edition

• Growing Revenue at a Pumpkin Patch 🎃 - 2024 Edition

• Evaluating Product Investments

• Motion vs. Progress - changing product culture to deliver results

• How to do annual planning better than you’re doing it today

Get Better At Communicating

• How To Communicate Effectively With Board Members

• Communicating Product Progress

• How to Improve the Relationship between Product and Marketing

• Communication Mistakes as a First Time Manager

• Figure Out Metrics Ownership

Successfully Balance Career and Family

• How To Actually Build Lasting Memories With Your Kids (Podcast)

• Managing Parenthood and a Career in Product Leadership

• Taking a Parenting Class Made Me a Better Manager (and Parent)

• More Time vs. More Energy (Startup Dad Podcast)

• Does Parenting Have To Be So High Stakes (Startup Dad Podcast)

• Managing Weekends As A Founder (Startup Dad)

• Integration Over Balance (Startup Dad Podcast)

• Work-Life Balance Is Overrated (Startup Dad Podcast)

• Fighting Off The “Should” Gremlins (Startup Dad Podcast)

• Leveraging FOMO to be Present and Purposeful (Startup Dad Podcast)

• On Parenting and Balance (Startup Dad Podcast)

• Founding a Multiple Companies and Raising a Family (Startup Dad Podcast)

• Balancing Career Ambition with Prime Parenting Years (Startup Dad Podcast)

• Amazing Parenting Frameworks (Startup Dad Podcast)

Thank you for all of your support this year and I’ll be back in your inboxes in January!

Subscribe now

A year ago I wrote about how a group of volunteer Dads applied growth strategies to a local pumpkin patch and transformed our earnings. We previously pushed past a ceiling of $30,000 where we had been stuck for ~13 years to achieve new highs of $38k and ~$42k in 2022 and 2023.

When we exited last year’s patch I was sure we couldn’t top it. After all, there’s only so much interest in a small-market, school-fundraising pumpkin patch. We also had Halloween falling on a Thursday this year (Happy Halloween everyone! 🎃) which means we had to close down the patch the previous weekend and missed several extra days of sales. It would be difficult, if not impossible, to go higher than $42k. 

I was wrong.

We leaned into some of our best performing tactics from 2022 and 2023, tested several new hypotheses, and delivered a 30%+ YoY improvement… AGAIN. This year we closed revenues four days ahead of Halloween at ~$54,001. The best part? We didn’t change prices at all from last year.

While there were a few logistical snags this year (I’ll elaborate below) we managed to persevere.

Here are 9 tactics we improved upon and several new tests for 2024. I’ll break down each one in the rest of this newsletter.

1. Signage

2. “Door Busters”

3. Delivery

4. Positioning and messaging

5. Merchandising

6. SEO

7. Improving sales tactics

8. Bulk discounting

9. Inventory management

Each of these tactics can be leveraged across most other businesses - they’re not unique to our efforts and they are powerful when executed well.

Background

The Wildwood Dads Club non-profit organization is a storied institution that dates back 100 years to the founding of the Wildwood Elementary School.

Our mission is to drive fundraising and community building for Wildwood alongside the Wildwood Parents Club. In our district, ~37% of school funding comes from local funding sources: parcel taxes, our education foundation nonprofit, and parent/community support. Throughout the year the Dad’s Club raises ~$15-30k for the school which is then spent on capital improvements, infrastructure projects, and classroom support. Things like sports equipment for PE, classroom technology, and more.

This year, our fearless “Patch Papas” were again Matt Derrigo and Russ Griffith. Team consistency from year to year turned out to be incredibly important. In the forming/storming/norming/performing cycle we are definitely at the performing stage. Matt helmed growth, marketing, and sales while Russ was in charge of vendor management, inventory and city relations (this turned out to be one of the more interesting years for that).

We again set new records for:

• Total patch revenues

• Total patch profit

• Single day sales

• Single hour sales

• Largest single transaction

• Total number of pumpkins sold

• Total number of smiles delivered

As a reminder, there are three distinct phases to a successful patch effort: prep and planning, execution and liquidation/cleanup.

Breakdown of Tactics

Signage

There are still only a handful of pumpkin patches in the immediate vicinity of the Bay Area. The “mega patches” (think Pumpkin Patch meets Disneyland) are farther out in more rural areas and tend to be half-day destinations requiring some planning. Ours is a “city patch” which benefits heavily from drive-by traffic and quick stop-ins to stock up. It’s especially helpful for last-minute shoppers.

We learned from our customer intercept survey last year that most people heard about the patch through three sources: local signage, Google search, and the school announcement. 

Last season Matt introduced our “Giant Jack” sign which we placed on a highly trafficked street that people exit onto from the highway. It alerted people to the presence of the greatest photo opportunity of the season: the giant pumpkin (author note: here’s how you can grow a giant pumpkin; the seeds are $20 apiece!). This year the Giant Jack weighed 1,150 pounds!

This year, ahead of the patch Matt re-created all of our patch signage. Many of them hadn’t been touched in many, many years. This was driven out of the survey outcome from last year that helped us recognize the importance of local signage to patch traffic.

Here’s the before (old sign) and the after (new sign). We’ll get into the messaging changes a little later but both the fresh, highly visible new signs along with messaging had a huge impact. 

Old Signage

New Signage

The other big change to signage was the location of the signs. We had to get a bit more creative this year because the city decided that they really wanted to crack down and enforce their previously unenforced sign restrictions. Despite the fact that we are a non-profit organization that funds our schools, the public works department took down several of our signs and said we had to find new locations or private property. 

While we initially fought this (and will be getting the city to change their law in the near future) we eventually gave up on it for the patch. It turned out that this unfortunate event became a great new discovery: by moving some signs just over the city limits into Oakland we were able to capture new demand from additional neighborhoods who want pumpkins but don’t have a local patch. 

Matt Derrigo:

“This year I had placed two signs in Montclair that definitely resulted in new customers. And after the city took our signs down, I reposted one further up on Oakland and Mount Vista which I learned attracted customers from that neighborhood.”

(Sadly, one of the risks with signage is that it can go missing. We lost this sign about 2 weeks into the patch operation).

Door Busters

This year we leaned further into “school announcement” which was a top 3 traffic driver in 2023. We continued the coloring sheet for classrooms (bring in a sheet, get a free pumpkin) this year but turned it up a few notches. 

First, Matt expanded our coloring sheet distribution and gave out sheets to two local churches for their youth groups and published it on our Google Business page.

Next, we decorated the fencing along the patch with all of the coloring sheets to highlight the great work of our community. People who came to the patch would see the coloring sheets, ask about them, and then take one so that they could come back to get a free pumpkin. This drove repeat foot traffic and repeat purchase in much the same way that an eCommerce store will give you a post-purchase discount for coming back a second time. 

But the big unlock was the professional photographer door buster. This year, on a handful of weekends we had a few of our more photographically-inclined volunteers bring their high-end cameras and take family photos for people at the patch. This served two purposes: the first was that it again drove incremental traffic since most everyone wants a family photo at a patch and second, we offered this for free and asked them to share info about the patch with their friends and family. If they felt compelled we asked they share a review as well.

This helped raise our visibility in searches for “pumpkin patch” and “pumpkin patch near me” which helped us capitalize on the Google search volume spike as the holiday approaches. We had nearly 700 direction requests from our Google business page this year. 

Here’s “Pumpkin Patch” in Google Trends for the last year:

We now have 23 5-star reviews on Google and 10 on Yelp. In addition to helping with local search the reviews serve as a great way for us to identify things that customers want and highlight those. Examples from recent reviews are: where the proceeds go, parking, gracious volunteers, complimentary photography, and more.

Delivery

In 2023 we did away with the online ordering which we had introduced during COVID. It was obvious from our Square data that we hadn’t sold much via online ordering. This year we lightly introduced a delivery service for Dads who wanted a way to volunteer but couldn’t work a shift. It worked like this: we put a call out to the Dads Club to ask your neighbors what they wanted from the patch and offer to pick it up for them when you were there shopping. This was an experiment so we didn’t emphasize it too much but it led to some incremental sales – especially for neighbors who couldn’t stop by. The “delivery” option led to our highest single order of all time at $900, eclipsing the previous one by 3x. 

Next year we may lean into the local delivery even more!

Positioning and Messaging

This was one of our big wins for 2024. As you saw in the signage we leaned into the “Support our Schools” message. We heard from countless visitors that this made the difference for them because they knew the money stayed in the community.

From Matt:

“One of the big changes that we made on our signs this year is we included in large font “support our schools”. People mentioned this when I asked them about the new sign whereas in the past, they would never have known — it just said pumpkin patch. For those unfamiliar with our patch and purpose, that was a game changer.

One anecdote someone from Walnut Creek (which is 25-30 minutes away from our patch) shared when asked why they didn’t stop at the countless patches between their home and us was that they “liked the cause.”

The key messaging that really resonated:

• Support our Schools

• Volunteer-led

• Local elementary school

Merchandising

In past years we had been concerned about piling pumpkins really high—feeling that it might make it harder to choose the right pumpkin. We’d arrange them in a single stack on hay bales or pallets. But after consulting with some produce vendors we learned that piling high is a key to accelerating sales. If inventory looks sparse people are less likely to buy. So we took a page out of the grocery store playbook and really stacked up the pumpkins. You can see somewhat in the photo below that we abandoned the single-stack on the hay bale and actually brought pumpkins out into the walkways. We also piled them up on pallets instead of single-stacking.

The psychology is clear—just like no one wants to take the last piece of cake; no one wants to grab from only a few remaining pumpkins. It also improved the choice and gave people an excuse to linger for a longer time at the patch to peruse the inventory.

SEO

In 2023 we greatly improved our SEO by claiming our Google Local listing along with our Yelp listings. Improving both of those has helped us dominate local search. The WWDC Pumpkin Patch is now in the Google Places listing showing our perfect, 5-star reviews and is the #1 recommended patch in Yelp’s “Top 10 Best Pumpkin Patches in Oakland, CA.” In fact, we’re the only 5-star patch that appears in the Yelp listings. This matters as Google Search is the top traffic driver to the patch!

We even won out in the LLM generated results:

Improving Sales Tactics

Selling is hard and not everyone is a natural salesperson. There were three, specific sales tactics that we introduced this year which made a huge difference in sales:

Walk around

We encouraged Dads to get out from behind the counter and walk the patch—helping visitors find the right pumpkin, taking heavy ones off their hands, and explaining the assortment. This proved crucial as we don’t have wheelbarrows or wagons at this patch (there just isn’t enough room). People are hesitant to pick up a pumpkin that they can’t lift or often are busy chasing their kids around and feel rushed. By getting out into the patch and helping busy parents we took some of the pressure off.  

Kid selling

In our volunteer outreach we made sure to mention that Dad volunteers should bring their kids to the patch to help. This year we had older kids working the register and ringing up guests as well as helping carry items to cars. We also sent kids out into the patch to help busy parents when Dad volunteers were otherwise occupied. Not only could the kids talk to visitors about their local school but they could help them pick out pumpkins. Who is going to say no to a cute kid?!?

Volunteer

We leaned hard into the volunteer messaging in our sales strategy. Not only did everyone have “Volunteer” badges that they could wear, but we would introduce ourselves to new visitors and explain that we’re a volunteer run, 501(c)(3) nonprofit organization.  

Bulk Discounting

As mentioned in last year’s newsletter one of our end of patch goals is to liquidate everything. We want to sell every last pumpkin and basically anything that isn’t nailed down (hay bales, corn stalks, etc.). To do so we implemented a last-day bulk discounting strategy:

• 10% off 1-3 pumpkins

• 20% off 4-6 pumpkins

• 30% off 7+ pumpkins

• Buy a hay bale and the whole order is 50% off

At this point the COGS are a sunk cost so if we can liquidate everything then it is all incremental profit. We also added extra “Closeout Sale” signs to our neighborhood signage and kept the patch open while we were cleaning up to sell everything. This strategy was successful yet again as we sold every single pumpkin in inventory.

Inventory Management

Last year we learned that this was critical to our profitability. Because we pay per pumpkin from our supplier and each varietal has a different cost (and demand profile) we wanted to make sure that we managed inventory incredibly tightly so as to not over or under order. The under-ordering turned out to be a killer last year as we probably lost out on $2-4k in extra sales when we ran out of carving pumpkins at a few critical points.

Not this year!

Through daily inventory tracking and sales forecasting we made sure we had exactly what we needed at all times. The “piling up” strategy also helped us make use of our inventory but merchandising as much of it as possible.

Matt:

“My forecasting changed Russ’ order amounts in a huge way. He'd take a physical look and derive a reorder, I then used the data and looked back a week and also analyzed last year— this would give us the confidence to bulk up far more than we did— to the extent that even I got nervous about some of our inventory. It's clear to me, you can’t sell product you don't have. We really should be buying even more and the goal should not be to sell everything, but have some pumpkins left on the shelf -- no different than the newspaper.”

Much like 2023 we saw similar trends in popularity—decorative gourds and cooking pumpkins are more popular at the start of October and then carving pumpkins come on strong later in the month.

One other key observation this year was weather patterns. While rain has never really been a factor for us in California this year the heat certainly was. We had a huge heatwave early in the month – with temps as high as 100, 101 and 102 and several days in the 80’s and 90’s. You know what people don’t like to do when it’s that hot out? Buy Fall pumpkins. Somehow we managed to avoid significant YoY declines despite an average temperature that was 5 degrees warmer this year!

Conclusion and Lessons Learned

When we started this year’s patch I was convinced that all of the optimizations we had done in years past would make it hard to grow again like we had the past two years. But against all odds: the city mandating we move signage, the heat, and a shorter selling season due to the date of Halloween we managed to grow all of our numbers: customers +44%, revenue +30%, and profitability +50%! We still need to dig into why AOV declined slightly YoY, but it’s likely a mix of traffic source and other factors.

Since we always do a retrospective, here are a few things that we’ll take into next year:

1. More door busters in 2025. The photographer was a big hit and we could schedule and publicize it even more. Giant Jack is such a big draw for the patch. This year’s Jack was 11% bigger than last year. Does that size drive incremental sales? Who knows! But it sure is fun.

2. More photo ops. We didn’t introduce the painted, wooden cutouts this year but that could be popular in ‘25.

3. Lean into the retention strategies with email outreach. We have two years worth of emails from our Giant Jack weight guess. We didn’t leverage them enough this year.

4. Delivery! At least 15% of households in Piedmont have retirees (who likely once had kids in the schools here). A well-executed local delivery strategy and door-knocking campaign could add thousands to revenue next year for those who can’t get to the patch.

5. More signage in more places. This continues to be a great driver of patch traffic and we should do even more.

6. This year the biggest limiting factor was parking. If people come by the patch and see that all the spots are taken they won’t stay. We can do more to zone off specific patch parking to create space.

7. Leverage community foot traffic for games and practices. There are lots of soccer games at the field and we could walk around to take orders from people and set out pumpkins for them. This is an area of exploration in ‘25.

From 2022 to 2024 we’ve grown revenue 42% and profitability by 57% from some straightforward growth and P&L tactics, steeped in data analysis. You can (and should!) apply these to just about any business and you’ll likely see similar results.

Now if only we could figure out some SaaS pricing…

Happy Halloween, and see you next year for Part 3!

The Wildwood Dads Club is a registered 501(c)(3) nonprofit. If you were inspired by this newsletter and want to contribute to local schools you can make a donation to our club here. It’s for the kids!

Subscribe now

It’s been a minute since my last newsletter! Primarily because I’m heads down working with the team at Mozilla which doesn’t leave as much time for writing. Fear not, I have a queue of mostly-finished drafts that are close to release. As always, if you have a topic you’d like to see me cover you can submit it here.

The Eastern part of the U.S. has been hit really hard by Hurricane Helene and people are struggling. If you’d like to donate to recovery and support efforts you can visit this link for some recommendations. My family chose the United Way of North Carolina. (side note: this is also an incredible UGC growth loop).

I’ve worked at or with a lot of B2B2C and marketplace companies and one topic that always comes up is how to teach your “B’s” to better engage with their “C’s.” In other words, if data suggests that doing X, Y, or Z behavior will lead to the businesses you support doing a better job serving (and acquiring, retaining or monetizing) their customers, how do you make that happen?

And when should you intervene versus allowing maximum flexibility and choice on your platform?

Examples of this are everywhere:

• How do you encourage Lyft drivers to be at the right places to capture demand?

• What’s the optimal number of membership tiers on Patreon and how do you make sure creators choose that and price appropriately?

• How much should tutors price their services on WyzAnt?

• What’s the ideal cadence for a membership offering at Table22?

• How do I get started with using a product like Notion?

• …the list is almost infinite!

One answer comes from a product principle we established when I was at Patreon – leading with opinionated defaults. Hat tip to Camille Hearst, an incredible friend and product leader, for teaching me that term.

I’ve seen opinionated defaults transform site experiences, shape behavior, and drastically improve metrics. In this newsletter I’ll unpack the concept and share some successful examples of leveraging opinionated defaults throughout my career.

What are Opinionated Defaults?

To understand opinionated defaults we should first understand “defaults” in general. Defaults have existed for decades, if not centuries. If you walk through a grocery store you will observe the role of merchandising in establishing default choices. Why is that cereal at a certain level? How do you choose the right toothpaste amongst a sea of brand extensions? Partially it’s because CPG companies pay for shelf placement to get certain products at eye level. This is the power of defaults in action.

Nowhere is this better articulated than the book “Nudge” by Nobel and Holberg Prize winners Richard Thaler and Cass Sunstein. An excerpt from the introduction, focused on a fictitious example of a food services director at a school district:

“Simply by rearranging the cafeteria, Carolyn was able to noticeably increase or decrease the consumption of many food items. From this experience she learned a big lesson: small changes in context can greatly influence schoolchildren, just as they can greatly influence adults. The influence can be exercised for better or for worse. For example, Carolyn knows that she can increase consumption of healthy foods and decrease consumption of unhealthy ones.”

The authors go on to describe “Carolyn” as a choice architect or someone who has “the responsibility for organizing the context in which people make decisions.”

As product builders we are all choice architects who establish defaults. Those defaults are predefined choices made for the user.

A product principle focused on leveraging opinionated defaults is one where in as many places as you can you’re putting your thumb on the scale and strongly suggesting a course of action.

Intercom talks about this in their product principles: “Opinionated by default; flexible under the hood.” This has led their product in a direction of being easy to use “right out of the box” but almost infinitely customizable if needed.

Another way to think about opinionated defaults is like the executive summary of a document. The author is telling you what the most important information is based on their knowledge of the document. You can get most of the benefit by reading the summary (or following the default suggestions) but a power user may want to dig into the details (customize beyond the defaults). 

Now that we’ve established what it means to leverage opinionated defaults let’s look at some examples across location, pricing, payment methods, templates and login. 

Opinionated Default Examples

Location

One subtle-but-straightforward opinionated default is the use of location detection to prefill choices. This became possible during the rise of mobile and Google’s emphasis on location-based search.

This morning, for example, I’m at my daughter’s lacrosse practice in Los Altos, CA. Opening up Yelp you can see that they’re defaulting to my local surroundings – makes sense: even though I might not want a burger or a haircut at 8am on a Sunday I certainly do want to find things around me (like coffee). By assuming that (having an opinion) they’re removing one step from my process and making it easier to search nearby.