Rob Snyder, a fellow at the Harvard Innovation Labs and the author of The Power of Pull, argues that customers almost never buy because somebody convinced them to, and that the difference shows up in a public company's income statement. Motley Fool analyst Rachel Warren takes him through his pull framework, the signals that separate structural demand from demand a sales and marketing budget is manufacturing, and which AI startups he thinks are actually working.
👤 Guest: Rob Snyder, fellow at the Harvard Innovation Labs
🎙️ Host: Rachel Warren, analyst at The Motley Fool
📰 Published: 30 August 2026
🔴 YouTube | 🟢 Spotify | 🟣 Apple Podcasts | 🔗 Show notes | ⏱️ 21 min | ✅ Time saved: 9 min
Key Takeaways
Revenue growth bought with sales and marketing is the growth that does not last
"revenue growth that is funded by aggressive sales and marketing, where it is a lot of push, is very hard to sustain over the long term"
The test is what the budget is for — converting people who already want the product, or persuading people they should want it
Net revenue retention is the one metric he trusts, and satisfaction scores are not
"those are stated preferences. Revealed preferences are in the post-sale usage."
Real demand looks like a customer buying despite the product, not because of it
A founder he knows reads the GDPR regulation aloud on sales calls and customers buy anyway
A compelling value proposition is not demand
"So the value proposition alone is kind of irrelevant." — what matters is what the buyer is already prioritizing
Products fail because nobody was prioritizing the problem, not because the idea was bad
"entrepreneurs will always go down this path of a product that makes total sense that nobody buys"
Companies take off before the product, the pricing or the sales process is any good
His own startup ran on a three-or-four-slide deck and a spreadsheet customers could not log into
Every software buyer now has do-it-yourself on their list of options
"can't I just do this myself with Claude or Codex?" — the startup's job is to find where that is not good enough, and will still not be good enough after the next model release
The AI products he likes are mundane, not visionary
A weekly pharma market-research report, an insurance quote comparison that took a person a week, a meeting note taker for financial advisors
He does not think anything in startups is durable right now
"The AI labs are coming for you. All the other startups are coming for you."
What he asks a founder claiming explosive organic growth is what happened after the sign-up
Viral traffic and an accelerator write-up convert, and then churn
The Model He Started With Failed Until a Restaurant Owner Called Him
Warren opened by putting the premise of the book to him: that the financial world has spent years treating consumer demand like a mathematical formula on a spreadsheet, and that his book takes the opposite view. Snyder said he started his first company believing exactly the spreadsheet version.
He thought demand worked the way the textbooks describe it — provide clear value and clear ROI, solve a problem, convince the customer, and they buy. It did not work. "we just got punched in the face for a couple of years when nobody would buy our product", he said.
The turn came from an unsolicited phone call from a restaurant owner: "Hey, I have no idea what you were trying to sell me, but here's where I need help." The caller named the thing he was focused on right then, said he would pay for help with it, and the company started to move
What he took from it: buyers do not behave the way he wanted them to, and they do not behave the way his economics textbooks said they would
"We went zero to four million in revenue in two years", he said, and he has since worked with a number of other startups on the same question — what is actually behind a purchase
His conclusion is that it is not that buyers pick whatever gives them the most ROI, the most value, or solves their problem
Push Is the Seller Convincing; Pull Is the Buyer Taking It Out of Your Hands
Asked to explain the distinction the book is built on, Snyder described the two states from the seller's side. Selling something he thought the market ought to want felt like force he was supplying himself — "it felt like I was pushing people to buy" — asking prospects on calls whether they had this problem, whether they wanted this value, and then doing all the following up. That is push.
When it worked it was almost none of that. People called him saying they had heard about the product and needed to buy it. "That is the buyer basically pulling the product out of my hands."
The difference carries past the sale. "After they bought, they would get set up. They would set themselves up. They would do all the work to implement the product" — where before he had to push customers to use the thing and beg them to use it
The framework itself: "buyers will pull a product out of your hands. If they have some sort of a project or a priority, they're trying to get done right now, but their existing options for getting that priority done are not good enough."
Absent those two conditions, a purchase is the odd event, not the expected one — the buyer would have to abandon whatever they were prioritizing, or decide their existing options are good enough and buy anyway
Demand Is Not a Line on a Chart, It Is One Person With a Priority
Warren said that investors evaluating public companies, and private ones too, tend to treat demand as something that moves predictably: "Demand is often treated like a line on a chart that moves predictably." She asked why that model breaks down at the moment of purchase.
Snyder's answer is that the line hides the unit. Every point on it is an individual person making a decision, and something in that person's life is causing it. He argues you can watch it happen — sit in on sales conversations and you can see whether the buyer is being convinced or persuaded, and if they are, the company is pushing.
Why that matters for a subscription business: "It's hard to convince someone to buy, then convince someone to use and repeat that forever and ever." Push is hard to scale and hard to retain against, especially in recurring revenue
What he says to look for instead: "What you're actually looking for is somebody who pulls the product into their lives and uses it as if they can't not use it, as if they're addicted."
A Compelling Value Proposition Is Not Demand
Asked why most new products fail even with a clear value proposition, Snyder said the business-school version — provide a compelling value proposition and people will buy — sounds right and is not. His counterexample is the sheer number of value propositions available to anyone at any moment that they are not acting on.
"There's 600 million SKUs on Amazon, right? All of them have some sort of a compelling value proposition. Just not to us right now."
"So the value proposition alone is kind of irrelevant." The question is what the person is prioritizing, and whether the value proposition is relevant to that
He says the failure pattern is visible at the Harvard Innovation Labs: brilliant products with interesting value propositions that customers look at and call exactly what the industry needs, and then do not buy, because they are not prioritizing anything related to it
"entrepreneurs will always go down this path of a product that makes total sense that nobody buys"
Companies Take Off Despite the Product, Not Because of the Pitch
Warren put the standard story to him — "I think the classic narrative we've heard is that a startup scales because of a genius founder's vision or an aggressive sales script." — and asked what he has actually seen at the point a company takes off.
Snyder said he used to believe the same thing, and that across companies he has watched go from zero to a million, zero to ten million and zero to fifty million, it almost never looks like that at the start. His own is the example. "We had a three or four slides sales deck." And: "The product for the first $100,000, $200,000 in revenue, was me in a spreadsheet that they couldn't log into."
What is present instead is a customer who has tried to do something and is blocked, buying despite the state of the product and despite a founder who is not good at selling
The example he leads with: a founder in Europe with the worst sales calls he has seen. "He basically reads the GDPR regulation on the sales call for no reason. And customers are desperate to buy regardless."
"You're looking for people who are trying to buy despite, not because of."
The fixing comes after, not before. Between a million and ten million in revenue you hire talented people to repair the things customers bought in spite of — the pricing, the sales process, and a product that actually does the thing
Not All Revenue Growth Is Created Equal
Warren asked how an investor would use the framework on a publicly traded company — to tell one that has genuinely cracked its market from one papering over weak demand with heavy sales and marketing spend. Snyder answered with a caveat first: "let's just put all the caveats out here of I am not an investor".
"What I would say is that not all revenue growth is created equal."
"revenue growth that is funded by aggressive sales and marketing, where it is a lot of push, is very hard to sustain over the long term" — and he says you can see it in a sales and marketing line that just goes up and up and up
The distinction he draws is not spend versus no spend. Companies with real demand often spend heavily too. "But I'd say like that sales and marketing spend is dedicated to converting people who already have demand rather than trying to convince people they should have demand."
Net Revenue Retention Is the Signal; a Satisfaction Score Is a Stated Preference
Asked for specific financial signals that separate a structural demand curve from a manufactured one, Snyder narrowed it to one number, and to B2B software specifically. He looks at customer retention, and "specifically net revenue retention is one of the big things that we look at".
Why that number and not another: it shows the company is not just convincing people to buy who will then churn, but converting people whose demand is large enough that they use more, pay more and get increasing value over time
What he weights less: net promoter score and customer satisfaction measures. "Those are fine, but those are stated preferences. Revealed preferences are in the post-sale usage."
Every Software Buyer Now Asks Whether They Could Just Do It in Claude or Codex
Warren noted that Snyder is co-founder of the AI cloud infrastructure company restack.com and evaluates early-stage B2B AI startups daily, and asked what he is seeing. He said the trend is AI, and that for any company that is not one of the AI labs, "the trend is how do we not get run over by one of the AI labs".
The question sitting in every software buyer's head, he said, is "can't I just do this myself with Claude or Codex?" In his own framework that is not a new problem so much as a new entry on the list of existing options — and it is often good enough.
The question that leaves for everyone else: "when is Claude Code, Codex, your AI products, when are those not good enough? And in what ways are they not good enough that the next model release will also not be good enough for."
Where he says the startups that work are concentrating: the point at which the customer has already tried it themselves. "they've gotten pretty far, but they haven't gotten all the way and they can't get the last mile to whatever it is they're trying to do"
The AI He Is Most Excited About Is Mundane
Asked where he gets most excited and where he thinks the use cases are overhyped, Snyder said his answers are smaller than people expect. "The things that I'm actually most excited about are way more kind of like simple and mundane than you might expect."
Pharmaceutical market research. A firm was paying a lot for a quarterly report on a niche of the industry that was decent but not quite what was needed. A startup focused only on AI for that space can produce it weekly or hourly instead. "It unblocks us to be able to get this kind of market research report that we in pharmaceuticals have needed."
Commercial insurance quotes. Assembling a comparison chart from the quotes that came back from insurers used to take a week. "now a very specialized AI that understands all the different kinds of quotes and insurance can do that for us in minutes instead of it taking a person weeks"
Jump, a company he says he hopes takes off: an AI meeting note taker built for financial advisors, working with their CRM and compliant with their regulatory requirements. Advisors had been typing notes into the CRM by hand for hours every day, and it removes a task they were already doing
He said the big AI labs are exciting too, but that his own interest is on the small-entrepreneur side — the little artifact or corner of a business where a startup takes off
What He Asks a Founder Claiming Explosive Organic Growth
Warren asked what questions he puts to a founder describing explosive organic growth, to work out whether there is a durable tailwind behind it. His answer starts at the far end of the funnel.
"I will focus on their metrics all throughout their funnel, with the biggest emphasis on what happens after these customers sign up or buy."
What he is screening out is hype. "It's not just we went viral on X or we were featured by Y Combinator" — a burst of inbound that converts and then does not last
Nothing in Startups Is Durable Right Now
On durability itself, Snyder was blunt that he does not think the category exists at the moment. Every startup in every category is in a race. The second Jump took off, he said, "There are a bunch of other AI meeting note takers for financial advisors."
There is no standstill in which a founder can call a business durable. "The AI labs are coming for you. All the other startups are coming for you."
Good metrics buy an obligation rather than a rest: once they look good, the job becomes staying ahead of everybody else
What that looks like in practice: "it's just staying close to customers, figuring out other ways we can unblock them and extending the product suite so that people want to stay with us longer"
Snyder's bottom line for an investor is that the shape of a growth number matters more than its size: growth that arrives because buyers were blocked and came looking shows up later as retention, and growth that arrives because a sales team convinced people shows up later as churn and a sales and marketing line that has to keep rising.
Products, Companies & Tools Mentioned
Claude, Claude Code and Codex (The do-it-yourself option now on every software buyer's list; Snyder says the question for any other startup is where these are not good enough and will still not be good enough after the next model release)
restack.com (The AI cloud infrastructure company Snyder co-founded, and the seat from which he evaluates early-stage B2B AI startups)
Jump (An AI meeting note taker for financial advisors that works with their CRM and meets their compliance requirements; his example of a mundane product taking off, and of how fast the copycats arrive)
Amazon (Cited only for the number of things on it that nobody is buying — his evidence that a compelling value proposition is not the same as demand)
Harvard Innovation Labs (Where Snyder is a fellow, and where he says he watches brilliant value propositions fail to find a buyer)
Y Combinator and X (Named as sources of the inbound surge he treats as hype rather than demand until the post-sign-up metrics say otherwise)
McKinsey (Part of Snyder's background as the host introduced him)
Books & Resources Mentioned
The Power of Pull – Rob Snyder (His new book, synthesizing years of building and analyzing startups into the pull framework)
Get the latest market chatter as it happens: