Product market fit is the point at which a product has found a market that wants to keep it. Marc Andreessen put the phrase into common use in June 2007, when he called it the only thing that matters for an early company. Platform and ecosystem strategy closed the last of the choices about how to win, so a strategy at this stage holds a segment, a promise, a business model, a price and a route to the customer. Every one of those choices is a claim about people the team has not yet persuaded. Product market fit is the first measurement that puts those claims in front of those people and reports what came back.
The sections below give the survey question Sean Ellis published in 2009 and the answer bands that come out of it. They then cover the method Rahul Vohra used at Superhuman in 2017 and the pockets of fit an aggregate score hides. The page ends with the limit on what any fit measurement says about the future.
The Sean Ellis survey question
Sean Ellis published one question and one threshold in 2009. He had led growth at a series of early companies, and he had noticed that the companies which kept growing all gave the same answer. The question asks a current user how they would feel if the product were taken away from them, and it offers three answers.
- Very disappointed.
- Somewhat disappointed.
- Not disappointed.
Only the first answer counts towards the score, and the score is the share of respondents who choose it. Some versions of the survey add a fourth option for people who have stopped using the product, and those answers are removed before the share is counted. Ellis put the line at forty per cent. He had seen that companies above it went on to grow, and that companies below it stalled however much they spent on acquisition.
The answer bands and what each one means
The forty per cent line cuts the possible results into four bands, and each band points at a different decision, so the number on its own is never the end of the reading.
| Result | What the number says | The next decision |
|---|---|---|
| Forty per cent or more across everybody surveyed | The product holds the market the strategy defined | Spend on reaching more of that market |
| Under forty per cent overall, with one segment above forty | Fit exists in a pocket the average is hiding | Narrow the strategy to that segment and measure again inside it |
| Under forty per cent overall, with no segment above forty | The product has not yet found a group that would miss it | Return to the segment and the value proposition before spending on growth |
| Under forty per cent, with too few answers to segment | The survey cannot answer the question yet | Collect more answers from current users before reading the score |
Only the forty per cent figure comes from Ellis. Every row beneath the first one is a reading of what a low score leaves open, and that reading is what makes the number useful. One figure on its own cannot separate a product with no market from a product with one market it has not yet found.
The Superhuman segmentation method
Superhuman is the case worth following in detail. The method used there found a market that the aggregate score had hidden. Rahul Vohra ran the survey in the summer of 2017, and the score came back at 22 per cent, which is a little over half of what Ellis asked for. Vohra stopped reading the aggregate and split the answers by who had given them. The very disappointed respondents clustered in one kind of work. They were founders, executives, managers and people in business development who handle a hundred to two hundred emails a day and send fifteen to forty of them. Narrowing the survey population to that group lifted the score by ten points, before a line of the product had changed.
Vohra then split the roadmap in two. Half of the work went into what the very disappointed group already valued, which was speed, keyboard shortcuts and automation. The other half went into what held the somewhat disappointed group back, which was a mobile application, integrations, attachment handling and search. Three quarters later the score stood at 58 per cent.
The method underneath those numbers is a loop Superhuman ran every quarter. The team surveys its users and splits the answers to find the group that would be most disappointed. It then reads what that group values, and what the group beside it says is missing. The next quarter of the roadmap is divided between the two.
Pockets of fit inside an aggregate score
The loop works because an aggregate score is an average over everybody a team has reached. If one segment depends on a product and three others merely tolerate it, the score comes back middling, and the figure gives no hint that the first segment is there at all. A team can split the answers by role, by volume of use, by company size or by the job somebody was doing when they signed up. That is what turns a score into a direction.
Geoffrey Moore made the same argument about markets in 1991, when Crossing the Chasm proposed taking one beachhead segment completely before moving to the next. A fit score read segment by segment is that argument in the form of a measurement. Reading a score that way carries a cost of its own. If a team narrows to the segment where fit already exists, it is choosing to serve fewer people on purpose, and a strategy that names the narrower segment is the honest form of that choice.
Fit as evidence about the present
A fit score is evidence about the users a product already has. The survey asks how people the team reached this quarter would feel about losing something that exists today, so the answer describes the present. It says nothing about the two years the strategy has committed to.
A strategy commits to claims a survey cannot reach. Three of them decide the outcome, which are whether the segment will still be that size in three years, whether the channel will stay open and whether a competitor will hold its price. There is no customer to ask about any of those three. Writing the claims down, so that each one can be ranked and tested on its own, is the next piece of work.