Early November, an online booking service for beauty salons in Almaty, charging 12,000 ₸ a month per salon after a two-week free trial. In October the marketer doubled the Instagram budget, and 320 new salons signed up instead of the usual 200. An investor writes to the founder: "The market wants you, this is product-market fit, double it again." The founder asks the analyst for a single number that says whether there is PMF. The service's numbers are made up.
There is no single number. But there is an answer, and sign-ups are not part of it.
What product-market fit is
The term took hold with Marc Andreessen's 2007 essay "The only thing that matters": product-market fit means being in a good market with a product that can satisfy that market. Without it, Andreessen writes, customers are not getting value out of the product, word of mouth is not spreading and deals drag on. With it, customers buy the product just as fast as you can make it.
As a working definition that is too broad, so it gets translated into behaviour: people in a clear segment get value again and again, and pay for it. The words that matter are "again" and "segment". One-off interest can be bought with ads; repeated value cannot.
The picture shows this on a different work task, with invoices and a report. For the booking service, the next cycle is the next month of subscription.
Sign-ups are not PMF
The 320 October sign-ups show that the ads kept working on twice the budget. That is interest in an ad. The breakdown by channel showed that the share of salons turning on online booking did not fall in any channel, but even that says nothing about PMF: turning on booking is a first step, and repeated value is still a long way off.
The worst case looks like this: twice as many sign-ups, and a month later almost nobody is taking online bookings. Doubling the budget again buys even more sign-ups and even more salons that leave, and the ads only speed up the leak. While people are not coming back, you first find out whether they use the product, whether they stay and why they leave, and you leave the budget alone.
The value event and the cycle
The value for a salon is clients booking themselves, without the receptionist. That happens every day, but the salon pays once a month. So the service measures retention monthly: a salon is retained in a month if it paid for that month. That is also how the service defines an active salon in its investor report.
Payment alone is not enough: a salon may simply have forgotten to cancel. So next to it you check whether the retained salon took online bookings that month. If it pays but takes no bookings, that is not retention but a delayed exit.
The cycle comes from the job. A service people use once a quarter will look dead on weekly retention, and a booking service measured on yearly retention will spot problems far too late.
There is also an earlier step. A first online booking during the trial splits new salons into those who will pay and the rest: in September, 46 of the 60 salons that got one paid, against 4 of the other 140. PMF is the next question: do the salons that paid stay?
The retention curve has to level off
Retention is built by first-payment cohort: salons that first paid in March are kept apart from April's. If the product is needed, the curve falls at first, as the salons it did not suit leave, and then levels off: the ones who are left stay. Amplitude's article on retention puts it plainly: if the curve does not stabilise but trends towards the x-axis, the product has a product-market fit problem.
Here is the March cohort, 48 salons:
| Month since first payment | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|
| Paying | 48 | 40 | 35 | 32 | 30 | 28 |
| Share | 100% | 83% | 73% | 67% | 63% | 58% |
By month six it is at 58%, and the curve is still falling by two or three salons a month. Is that a plateau or not? You cannot tell from the overall curve. It needs splitting.
The service knows how many stylists each salon has: the number arrives in the masters_total property of the master_added event from the event plan and sits in the database. It is known before the first payment, and you can act on it: target ads and talk to the salon about its own job. A stylist who works alone also signs up as a salon.
| Month since first payment | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|
| Two stylists or more | 30 | 28 | 26 | 25 | 25 | 24 |
| One stylist | 18 | 12 | 9 | 7 | 5 | 4 |
For salons with two or more stylists the curve levels off at around 80% by month four: between months four and six, one salon out of 25 left. Solo stylists leave every month, and by month six four of the 18 remain. They are what drags the overall curve down.
One cohort is not yet a rule. In the April and May cohorts, salons with two or more stylists also hold at around 80% by month six, while fewer than a third of solo stylists remain.
Outside benchmarks exist, but handle them with care. In 2020 Lenny Rachitsky and Casey Winters asked 20 experienced growth practitioners what counts as good retention after six months. For SMB and mid-market SaaS (customers there are companies of roughly 100 to 1,000 employees, far bigger than a salon), where the count is companies that have started a paid subscription, good came out at around 60% and great at around 80%. The definition is close, but not the same: theirs is "active", the service's is "paid", and their "six months later" corresponds to our seventh month since first payment, one month beyond the table. And these are the opinions of twenty people, so decisions are made on your own earlier cohorts.
The Sean Ellis survey
The curve shows what salons do. Why they do it, a survey can suggest. Sean Ellis proposed asking a single question: "How would you feel if you could no longer use this product?" The answer options are "very disappointed", "somewhat disappointed", "not disappointed" and "I no longer use it" (the survey as described in the Real Startup Book).
Ellis compared the answers of nearly a hundred startups: companies that struggled to find growth almost always had fewer than 40% answering "very disappointed", and companies with strong traction had more (Rahul Vohra's account of Superhuman in First Round Review). The Real Startup Book calls that 40% a heuristic: a guide, not a law. It also says whom to ask: people who have used the core of the product, used it at least twice, and used it in the last two weeks. Thirty answers give only a direction; you treat the number as a signal from 100 upwards.
The service sent the survey to the owners of 900 salons that paid for October and took at least two online bookings in the last two weeks. 210 replied:
| Salons | Replied | "Very disappointed" | Share |
|---|---|---|---|
| Two stylists or more | 185 | 96 | 52% |
| One stylist | 25 | 3 | 12% |
| Total | 210 | 99 | 47% |
47% overall is above the heuristic, but it is an average of two very different groups: 52% for salons with a team, 12% for solo stylists. Only 25 solo stylists replied, below the minimum of 30, so 12% is a hint rather than a direction. The hint matches the retention curve, and the match counts for more than the exact percentage.
Remember who is missing from the survey. It went only to salons that stayed and use the product, and roughly one in four of them answered, perhaps the happiest ones. Salons that left are not in it at all; only the curve speaks for them. And words are not money: "very disappointed" is a statement, renewing a subscription is an action. When they disagree, believe the renewal.
Why salons leave
The third source is talking to the ones who left. The analyst got through to 20 solo stylists who stopped paying in spring and summer:
- 12 said their clients message them on Instagram anyway, and a stylist knows her own free time without a calendar;
- 5 said 12,000 ₸ a month is too much for one person;
- 3 had gone to work at a salon or moved away.
No button will fix that. The service takes load off the receptionist, and a solo stylist has none: for twelve of the twenty, the job simply is not there.
What to do with the answer
The answer to the investor fits in three lines: "PMF is there for salons with two or more stylists: in three spring cohorts in a row about 80% of them are still paying in month six, and 52% of the owners of such salons who replied would be very disappointed without the service. Solo stylists do not have it. We will not double the budget blindly; first we will find out who Instagram is bringing in."
For the marketer this means three steps:
- Break sign-ups down by channel and number of stylists: how many solo stylists there are among October's 320 and where they came from. Until those numbers exist, there is nothing to say about the quality of October's doubling.
- Write ads for salons with a team: about the receptionist who cannot keep up with evening messages, not about "easy booking for stylists".
- Count payback on the salons that stay, and on contribution rather than revenue. Unit economics will show how much you can pay for a salon from the segment where PMF exists.
Solo stylists are not necessarily lost: perhaps they need a different product, cheaper and simpler. But that is a separate hypothesis with its own pilot. They have no place in the general funnel.
Exercise: PMF for your own product
You need a product with a subscription or repeat purchases: a client's service, an app, a shop with regular customers. If you have no client, use the booking service's numbers from this lesson.
A similar case to work through. A partner of the service runs a "three months free" promotion in a Telegram community of beauty professionals. In a week there are 500 sign-ups, and by the end of the month 40 of those salons take online bookings, while the other 460 never turned booking on. The partner is drafting a press release: "The service has found product-market fit." What do you tell him?
There is no PMF to see here. 500 sign-ups show that three free months interest some people, and the 460 salons that never even turned booking on show that the product has not been useful to them yet. There is no repeated value at all so far: this cohort's first paid month is three months away. If nothing changes, a second promotion like this will bring hundreds more salons that do not use the product. Before you invite new ones, find out who those 40 are: how many stylists they have, where they came from and whether they are still taking bookings a month later. And call ten of the salons that never turned booking on. The press release can wait until this cohort's first payment.
Video: finding Product-Market Fit
An introductory overview of the concept.