Key takeaways
  • Product-market fit is when the market wants the product more than you can supply.
  • Measure it with the retention curve and the Sean Ellis 40% test.
  • There can be multiple PMFs, and beware the "almost" illusion.

Product-market fit is the most overused phrase in the language and the most under-understood concept behind it. Founders claim it after their first ten paying customers. Boards demand evidence of it before Series A. Product leaders write it on quarterly OKRs as if it were a feature to be delivered. None of that is what Marc Andreessen meant when he wrote the original essay in 2007. PMF, as he described it, is not a milestone. It's a moment, and one you mostly notice by what happens to your inbox.

Andreessen's line is worth quoting in full because it gets misquoted constantly. "You can always feel product-market fit when it's happening. The customers are buying the product just as fast as you can make it — or usage is growing just as fast as you can add more servers. Money from customers is piling up in your company checking account. You're hiring sales and customer support staff as fast as you can. Reporters are calling. You start getting awards. Harvard Business School wants to do a case study. Investment bankers are staking out your house." It's a physical sensation. Demand pulls. You stop selling and start triaging.

What it isn't.

It isn't strong early-customer interviews. It isn't a waiting list (you can manufacture one). It isn't the first ten paying customers, however excited they are. Those are early adopters, and early adopters will buy almost anything that gestures at their problem. It isn't a closed funding round, though one might follow it. And it isn't a state you achieve once and keep. PMF can be lost: when the market shifts, when a competitor compresses the category, when your product line drifts away from the segment that originally pulled. Slack had it. Then Microsoft Teams bundled the same functionality into the licence companies already paid for, and the pull weakened. PMF is conditional and it's segment-specific.

The trap is that an early-stage company always feels like it might have PMF. The team is shipping. Demos go well. A few customers love the product. The founder gives a confident answer when the investor asks. None of that is fit. It's traction, and traction without fit is the most dangerous thing in a startup's life because it justifies the spending that follows it.

The retention curve.

The most honest signal of PMF is the retention curve. Cohort your users by week of first activation. Plot the percentage still active each week after. Three shapes emerge.

Diagram · The retention curve as a PMF signal
The retention curve as the clearest signal of product-market fit A retention chart with weeks 0 to 12 on the horizontal axis and percentage of users still active on the vertical axis. Three cohort curves are plotted. The red-grey No PMF curve starts at 100 percent and falls steadily toward zero by week 12 — a leaky bucket with no floor. The amber-stone Partial PMF curve declines and flattens at roughly 15 percent — some users keep using it but most churn. The marigold Strong PMF curve declines and stabilises above 50 percent — a clear floor of users for whom the product is essential. A shaded zone marked PMF zone sits above the 40 percent line, the rough threshold Andreessen-era founders point at as the floor of a real retention curve. A second annotation references the Sean Ellis 40 percent test — when 40 percent or more of surveyed users would be very disappointed if they could no longer use the product, you have qualitative PMF to match the quantitative floor. 100% 80% 60% 40% 20% 0% W0 W2 W4 W6 W8 W10 W12 WEEKS SINCE ACTIVATION % STILL ACTIVE PMF zone — curve must flatten here No PMF · ~1% Partial · ~18% Strong PMF · ~52% Sean Ellis test · ≥40% "very disappointed" if the product disappeared = qualitative PMF.
PMF is the curve that flattens, not the curve that climbs. A retention curve still falling at week twelve is a leaky bucket, no matter how strong the top-of-funnel looks. The Sean Ellis survey gives you the qualitative version of the same signal.

The first shape is the killer. The curve falls steadily and never finds a floor. By week 12 essentially nobody is left. Most products at most stages look like this, and most of the time it's because the product doesn't yet solve a problem urgent enough to come back to. The team's instinct is to bolt on features. The actual fix is to re-frame the problem, which is much harder, and which is what early-stage PMF work really is.

The second shape is the one that fools teams. The curve declines, then flattens, but at 10 or 15 percent. There's a small group for whom the product is real. Founders interpret that as a beach-head segment to expand from. Sometimes they're right. More often they've confused fit-with-fifteen-percent with fit, and they scale the business as if the whole market behaves like those fifteen percent. It doesn't.

The third shape is what real PMF looks like. It declines and stabilises somewhere materially north of forty percent, and stays there cohort after cohort. That is the floor of a retention curve, and a floor means there's a population for whom the product has become indispensable. That population funds the company.

The Sean Ellis 40% test.

Sean Ellis, the early growth lead at Dropbox, Eventbrite, and LogMeIn, codified the qualitative version of the same signal as a single survey question. Ask your active users: how would you feel if you could no longer use this product? Options: very disappointed, somewhat disappointed, not disappointed. If 40 percent or more answer "very disappointed", you have PMF. Below 40 percent, you don't yet. The threshold is empirical. Ellis derived it from observing which companies he'd worked with that scaled successfully and which didn't.

The brilliance of the test isn't the magic number. It's that it forces you to ask users the right question. Most PMF research asks would-you-buy or do-you-like, both of which generate optimistic answers that don't predict behaviour. Asking what they'd lose forces a more honest accounting of the product's role in their life.

Rahul Vohra, founder of Superhuman, turned that one question into a measurement system: a continuous PMF survey running against every active cohort, segmented by use case, used to identify which segments scored above 40 percent and which didn't. They invested in the segments that did, and either fixed or de-prioritised the ones that didn't. The product team optimised the roadmap explicitly against the disappointed-users metric until it crossed the threshold company-wide. It's the most disciplined PMF process I've seen documented.

If you have to ask whether you have product-market fit, you don't have it.

Multiple PMFs.

A single product can have PMF with one segment and zero pull with another. That is the rule, not the exception. Salesforce has PMF with enterprise sales orgs and almost none with two-person startups. Notion has PMF with knowledge workers in tech and patchy fit with traditional SMBs. Each segment is its own PMF question, and each segment has its own retention curve. Aggregating across segments hides the answer.

The first product question a CPO should ask isn't do we have PMF. It's which segments do we have PMF with, and what's the relative size of each. The honest answer is usually one or two segments, plus several where the product half-fits. The strategic question is whether to deepen in the segments where you have fit, or expand into adjacent ones. And that's a different conversation from "do we have PMF, yes or no".

The "almost" illusion.

The most expensive mistake in product is scaling before PMF. Burn doubles, sales hires get made, marketing spend gets unlocked, the org structure assumes a market that pulls, and all of it is built on a curve that hasn't actually flattened. When the founder eventually accepts that the product doesn't yet fit, the company is now too big to do the rebuild that PMF actually requires. The team is now solving a smaller problem (sales productivity, conversion optimisation) than the real one (the product itself).

Brian Chesky describes Airbnb's pre-PMF period as flat-lining: a year of revenue that didn't grow because the product hadn't yet earned the pull. They didn't scale through it; they sat with it, talked to hosts, photographed listings by hand, and rebuilt the product until the curve turned. Marc Benioff at Salesforce describes the same period. The period before pull arrives is where you build the product the market actually wants. Most founders skip it because investor pressure makes it feel slow. The teams that come out the other side with PMF are the ones who treated that period as the work, not as a delay before the work.

How we use this at Product Pieces.

When a founder or CPO brings us a product and asks "do we have PMF", the answer almost never comes from one number. We look at three things together: the segmented retention curve (does it flatten, and where), the Sean Ellis disappointment score by segment (is there a population that would lose something real), and the source of new customers (is it referral, or is it spend). If all three line up, you have PMF in that segment. If one of them is missing (usually retention is fine but referral isn't, or referral is strong but retention erodes), what you have is the seed of PMF, and the work is to sharpen it before scaling.

The mistake we see most often isn't teams missing PMF. It's teams declaring it on too thin evidence and then hiring against the assumption. The PMF Set Piece is two weeks: segment the retention curve, run the Sean Ellis survey across active cohorts, source-attribute the last quarter's new customers, and give the founder a candid read on which segments are real and which are aspirational. The output isn't a verdict. It's a map of where the pull is — and where it isn't yet.

Next issue: N°17, Jobs to be Done. The switching question, and why people don't buy products.