Product-market fit is not a feeling, and it is not a round of applause at a demo day. It is a set of behaviours you can measure. Most first-time founders confuse being liked with being needed. Friends and mentors saying "nice idea" tells you almost nothing. What tells you the truth is whether people keep coming back, whether they would be upset if your product disappeared, and whether the numbers are getting better cohort after cohort. This lesson gives you three concrete instruments and the early signals that tell you that you are close.
Retention cohorts: the truest signal
Retention is the single hardest metric to fake, which is exactly why it matters most. A cohort is simply a set of users grouped by when they first signed up or paid: your January signups, your February signups, and so on. You then track what percentage of each cohort is still active in week one, week two, month one, month two, and beyond.
Plot this and you get a curve. In a product without fit, the curve keeps sliding toward zero: everyone eventually leaves. In a product with fit, the curve flattens, meaning that after early drop-off, a stable core keeps using you month after month. The rarest and best pattern is a "smile" curve, where lapsed users return as you improve. Aim to see flattening by roughly month three to six. Define "active" honestly for your product: for a payments or commerce app it might be a repeat transaction, not just an app open.
The Sean Ellis 40 percent test
Growth expert Sean Ellis popularised this test around 2009 after working with companies like Dropbox and Eventbrite. You survey people who have genuinely used your product and ask one question: "How would you feel if you could no longer use this product?" The options are "very disappointed", "somewhat disappointed", and "not disappointed".
The benchmark: if 40 percent or more say "very disappointed", you likely have product-market fit. Below that, you probably do not, yet. The real value is in the follow-up questions: ask what type of person would most benefit, what the main benefit is, and how you could improve. Then focus relentlessly on the segment that already loves you. Superhuman famously used exactly this method and raised its score from 22 percent to 58 percent over several quarters by doubling down on its most enthusiastic users. Survey only recent, real users, and be careful with tiny samples: ten replies is a hint, not proof.
NPS: useful, but not a fit metric
Net Promoter Score asks, "How likely are you to recommend us to a friend or colleague?" on a scale of 0 to 10. Scores of 9 to 10 are promoters, 7 to 8 are passives, and 0 to 6 are detractors. Your NPS is simply the percentage of promoters minus the percentage of detractors, giving a number between minus 100 and plus 100.
NPS is a good pulse on sentiment and word-of-mouth, and word-of-mouth matters enormously in India where WhatsApp forwards and community referrals drive early growth. But treat it as a supporting metric, not proof of fit. People can say they would recommend you and still not return. Retention and the 40 percent test measure behaviour and need; NPS measures goodwill. Track it over time, and read the free-text comments more than the number itself.
Leading indicators that you are close
These metrics take weeks to move, so watch for earlier signs while you build:
- Unprompted pull. Users message you asking for access, features, or higher limits without being chased.
- Organic word-of-mouth. New users arrive saying a friend or colleague sent them, not from your ads.
- Usage deepening. The same users do more each week: more transactions, more logins, more invites.
- Reluctance to churn. When you try to end a pilot or trial, users push back and ask to keep it.
- Willingness to pay. People pay, renew, or upgrade rather than only using the free tier.
Practically, set up a simple cohort table in a spreadsheet from day one, run the Sean Ellis survey over email or WhatsApp every quarter, and glance at NPS alongside. Do not chase vanity numbers like total signups or app installs. If your cohort curve is flattening, your "very disappointed" score is climbing toward 40 percent, and users are pulling the product from you, you are close. Everything else is noise.