How to Find Product-Market Fit with the PULL Framework: From Idea to First Deals

Snyder opens by saying that one of the main mistakes people make is thinking pull is a property of a company. In his view, pull is something a person has. He illustrates the point by saying a founder is not selling to SpaceX but to the head of legal at SpaceX, who has a project on her to-do list that she cannot do herself.

The distinction matters because it changes who a founder should target and what they should be listening for. Rather than asking whether a company has a need, the question is whether a particular individual is stuck on a priority that the product could unstick.

Snyder notes that when a prospect reacts with enthusiasm, saying something like "this is exactly what I need," this often does not lead to a purchase. For that situation he describes what he calls throwing the Hail Mary: asking whether there is anything on the person's to-do list that they would pay him to do right now.

The purpose is to see whether the prospect actually tries to buy. If they do not, Snyder says, the team has learned something and can keep iterating. If they do, the team may have something, and the next step is to check whether that demand repeats.

Later in the conversation the discussion turns to what comes after a customer shows real demand. Snyder says retention is very important and not obvious, and begins to outline his recommendations on that point.

Snyder walks through the example of Jump, a product for financial advisors. Advisors have to take notes from client meetings and put them into their CRM, and failing to record them properly could land them in jail, so they must do it, but they dislike the time it takes. Generic AI notetakers were on the list of options, but they did not integrate with the advisor CRM, and it was not clear they met SEC rules, so advisors could not use them. When Jump showed up, people reportedly pulled the product out of their hands. One user said it must be how Alexander Graham Bell felt when he invented the telephone.

Snyder then describes a contrast case. Insurance agents also meet clients often and have a similar project, but they are not required to put meeting notes in a CRM, so their existing option is simply not taking notes, which they already do. The Unavoidable element is missing, so the framework does not fill out and the force is not there. Snyder draws the lesson that all four elements are needed for real pull.

Snyder repeats that pull lives with a person even when the product is enterprise software. Lenny asks whether the point is that SpaceX does not need the product but a specific person there does, and Snyder agrees that this is a common mistake.

He gives an example of a frequent error: saying a company sells to mid-market private equity firms. Snyder argues that the seller should instead identify a partner or operating partner in that firm who is specifically trying to accomplish a particular task.

Snyder gives two reasons for approaching idea generation this way. First, the ideas that matter do not emerge from a whiteboard. He cites a friend's idea that was found by observing a problem, and says it is unclear that it would have surfaced in an interview either.

Second, he says building a company means serving a particular kind of person. The whole point is to help them do something, so a founder without firsthand experience will struggle to make the thousands of decisions required to build something that actually serves them.

Snyder then describes how the pull hypothesis is written. A founder states the kind of person they believe is trying to achieve something, what that person is trying to cut or accomplish, why the option of doing nothing is unacceptable, and why existing alternatives are not good enough. He gives a physical therapy example: a physical therapist trying to cut a specific cost, for which hiring a consultant was too expensive and too slow. The next step is to show the hypothesis to that person and ask whether it describes them and whether they would buy.

Lenny asks about the common advice to solve your own problem. Snyder says it works when a founder would have paid someone else to solve the problem. He cites a friend, Emily, who founded Clearly AI after working in security engineering at AWS. She had found a problem she would have paid to have solved, but no one was offering a solution, so for her the advice held.

Snyder says the approach falls flat when a founder has a problem they would not pay to have solved themselves but think the world should have fixed. He is honest about his own past: he has built things he thought the world needed, but if he looked himself in the mirror he would not have said yes, buy this right now.

Snyder also says he works mostly in B2B SaaS but has seen many consumer founders use the approach, and mentions a story about Spanx as an example of consumer success.

Snyder describes the core insight at this stage as coming from being hands-on with the people you will help. He refers to the idea of an earned secret. He says the ways to find one are either to have lived the problem inside a company, such as the AWS example, or to work side by side with the people dealing with it and look for the problems they have.

He adds that the method matters less than the source, and that the most likely approaches to work are getting your hands dirty and learning from reality rather than from theory.

Asked for a rough breakdown of where ideas came from across the roughly 50 companies he has worked with, Snyder says about 10% came from the founder's own experience and about 90% came from going out and getting firsthand experience. He says some of those companies sold the wrong product before finding the right one, but most did the work firsthand. He also says a retreat is mandatory as part of his advice.

Snyder says he tends to start on LinkedIn, because the kind of people founders want to talk to usually have active profiles. He suggests adding 100 to 150 people per week, about 20 a day, and messaging them once they accept the connection request.

His typical targets are that 30 to 40% of those contacted should accept the connection request, 30 to 50% of those should reply, and within the repliers a subset should say yes, with the majority saying something like "yeah, sure, why not." Lenny points out that if the acceptance rate is low, the profile itself may not look worth accepting, and Snyder agrees.

On the call itself, Snyder says the purpose is not to sell immediately but to see whether pull exists. If it does, the conversation turns into a sales call; if not, nothing the founder does will change that.

Snyder says the first sales call is almost always scheduled for 30 minutes. He has found that shorter calendar blocks are less likely to show up on the prospect's calendar and so are missed.

On demos, he is hesitant. If a company is very early, it probably does not have a demo to show. Later on, a very short demo might fit into the first call. He notes that this differs from enterprise sales, where the advice on waiting to demo applies to a different stage than the early hypothesis phase.

Asked how to tell a call went well, Snyder says the sign is that you surface pull, or that you learn why the prospect does not have pull and what would be relevant to them. Either way you come away with a filled-out PULL framework: either you understand their pull, or you understand why it is absent.

Snyder explains that the status quo is generally good enough for most buyers, who have many priorities. The bar to get someone to decide to try something new and buy it is high, and the framework is meant to get past that hurdle by finding people for whom the status quo is not okay.

He says the goal is to look for people who are already trying to change and do not need convincing. He is skeptical of the idea that a product has to be ten times better than the existing solution. The test is how stuck the buyer is with their current options. Using the Jump example, he says that if generic notetakers do not get the job done, Jump is infinitely better because it unblocks something.

Snyder also warns that a founder who asks whether a prospect's existing option is good enough and hears "yes, we can do it ourselves" should take that answer seriously, because it prevents pushing a product the market does not want for years.

Asked what happens once a prospect says they want to buy, Snyder says the next step depends on the size of the deal, but the founder should come back with a concept and a plan. He gives the example of a company selling into pharma. Its founder got a Hail Mary intro to a pharma executive, who said current market research was slow and bad. The founder had no slides, only the request.

The founder offered to come back in a couple of days with a rough mockup of what the solution might look like and to discuss whether it fit. The mockup was vibe-coded with no slides. After getting feedback, the founder asked whether anything else was needed to decide if the product was a fit for the organization. The executive asked about price, and Snyder says there is a talk track for pricing.

Lenny notes that starting with services has historically been considered a terrible idea, because founders fear they never automate and end up running a consulting business with poor margins. Snyder says he has seen the approach work, while acknowledging it has downsides and can get stuck as a service business.

In his example, a company may deliver a service-shaped offering first to validate that real pull exists and to figure out what to build, delivering a mix of manual and automated work. He says AI increasingly does more of this work, and that doing the work by hand teaches the team which problem opportunities are worth pursuing.

On early pricing, Snyder says he is mostly testing willingness to transact rather than willingness to pay in the first few customers. He gives the example of a company whose first deal was enterprise and whose founder wanted to maximize revenue ahead of a demo day, and says the approach is to name a number in a way that does not get the founder slapped.

Snyder recommends doing roughly five sales calls and then assessing whether pull exists. Usually there is no pull at first. If the calls show none, the Hail Mary question can surface new information worth following up on, such as asking whether there is anything the founders can do right now that is on the prospect's to-do list. The team can then run another set of five calls, or go back and gain firsthand knowledge and try again.

Lenny summarizes the process: form a pull hypothesis, use the template, do cold outbound with a message that gets attention, run about five calls, and if there is no pull, use Hail Mary questions or enter goblin mode to look for something. Then form another hypothesis, do five more calls, and repeat until there is pull, at which point there is a business.

Snyder describes goblin mode as an all-out, break-glass approach: going to extremes to find who will buy something right now, and then generating a hypothesis from what you learn.

Snyder says that if almost everyone is churning, it is unacceptable not to figure out why. He notes that for many horizontal products some user drop-off is common. He describes a startup that built a horizontal AI workflows product and found that one or two kinds of customers kept using it while everyone else churned.

He says that when 90% or more of customers churn, the team can often work out whether the problem is something it is doing or something about the customers, and that this reveals who actually has pull that can be sold, delivered and retained.

In that case, freight forwarders were using the product for one specific workflow, so the company pivoted to focus on them. It moved from a horizontal product with massive churn and low price to a premium product for freight forwarders.

Snyder describes the ideal early customer as someone who pulls the product out of the founder's hands and is weird not to buy and weird to churn. In his account, the adoption process is mostly the customer pulling rather than the founder pushing, and the founder's task is to find more customers who are essentially clones of that first perfect-fit buyer.

The founder's job is then to find more customers who are essentially clones of that first customer. Snyder also advises against proactively offering pilots, proofs of concept or design partner arrangements. If a customer says they need to pilot, the team can accept that then, but it should not offer one first.

Lenny raises the question of whether the approach applies to established companies building new products. Snyder describes advising founders who are launching second products by selling new features to existing customers before building them, and says a new-product leader told him the ideas that worked came from sales calls where customers asked for a related capability.

Snyder says the framework is hard to get right because finding an idea that actually works is difficult. He says it is meant to save founders the pain of building something that will not work and of spending a lot of their lives on it.

He compares it to a metal detector on a beach. It may not find metal where you are looking right now, but at least you are not given a spoon to dig through the sand. He adds that sometimes you may find something and then realize it is not worth anything.