पाठशाला Pathshala · ग्राहक Grāhak, The customer · Lesson 05 · Start

Segmentation that changes what you build

Men aged 25 to 34 in metro cities do not buy protein powder. People who joined a gym in January do. How to split a market by behaviour and buying trigger, and pick the segment to serve first.

Pathshala, The Founder Library · 11 October 2026 · 8 min read

Men aged 25 to 34 in metro cities do not buy protein powder. People who joined a gym in January and finished the tub a friend gave them do. The first sentence is a demographic and tells you how to buy advertising. The second is a segment, because it tells you what to build, where to be and what to say. This lesson is how to split a market so that the split changes the product.

It covers why demographics fail the founder, how to split by behaviour and then by trigger, how narrow is narrow enough, a calculator that shows what a conversation is worth under each cut, a scorecard for choosing the first segment, and a worked example in Pune.

Why demographics fail the founder

Clayton Christensen, Scott Cook and Taddy Hall open Marketing Malpractice with the number that indicts the usual method: some thirty thousand new consumer products are launched each year and more than ninety per cent fail, after companies have spent heavily on understanding their customers. The authors’ diagnosis is that marketers segment by attributes, age, income, city, company size, and attributes correlate with purchases without causing them. Nobody buys a milkshake because they are a 35-year-old commuter. They buy it because the drive is long and breakfast will not last until ten, and the same person buys the same shake for a different reason at four in the afternoon. The job, not the customer, is the unit of analysis.

For a founder the failure is practical. A demographic segment is large, which feels safe, and it is cold, because at any moment almost nobody in it is looking to buy. It tells you nothing about what to build because its members do not share a problem; they share a census row. And it leads to the generic message, “for restaurants”, “for SMEs”, “for working professionals”, that nobody in a courteous market will tell you is boring.

Split by behaviour, then by trigger

Two cuts replace the demographic one. Behaviour is what the person does about the problem today: the workaround, how often the problem recurs, what they spend on it, and who in the household or company does the work. Two restaurant owners with identical revenue are in different segments if one has a purchase manager and the other sends a WhatsApp message to a vegetable vendor at six every morning. The [interview protocol](/library/the-customer-interview-done-properly) collects all four behaviours in fifteen minutes, which is why twenty interviews produce a segmentation and a survey of a thousand does not.

Trigger is the event that moves a person from having the problem to shopping for a solution: the [hiring moment](/library/jobs-to-be-done-job-customer-hires-you-for) of the previous lesson. The cook left. The second outlet opened. The twentieth employee joined and the spreadsheet broke. A GST notice arrived. A customer paid ninety days late for the third time. A segment is the set of people who share both the behaviour and the trigger, and the trigger is what makes it warm: at any time a known share of them are in the market and can be found by the event itself, which is usually visible, a job post, a new listing, a shop board going up.

Test the split with three questions. Do the members name the same alternative when asked what they do today? Is the decision made by the same role, the owner, the HR head, the mother? Do they raise the same first objection? If the answer to any of the three is no, you have two segments wearing one label, and the product that serves both will serve neither.

How narrow is narrow enough

Geoffrey Moore, whose Crossing the Chasm made the beachhead the standard first move, gives the test in three clauses: the first segment should be big enough to matter, small enough to lead, and a good fit with your crown jewels, the things your company is unusually good at. The pragmatic buyers who make up most of any market need a compelling reason to buy and a high degree of trust, and both are only available to a company that has visibly solved the whole problem for people exactly like them.

Paul Graham’s Do Things That Don’t Scale gives the reassurance founders need to go that narrow: Facebook launched with a potential market of a few thousand people at Harvard; Microsoft began with a few thousand hobbyists; the question is not how big the first market is but how big the company could get if the founders do the right things. A segment is narrow enough when ten conversations in it show a pattern and a founder can name the room its members are already in. It is too narrow when the arithmetic below says year one cannot pay a small team even at full penetration.

The same discipline works inside your own user base. When Rahul Vohra measured Superhuman’s product-market fit, 22 per cent of users said they would be very disappointed to lose the product, against the 40 per cent benchmark he took from Sean Ellis. He tagged every respondent by what they did for a living, kept only the very disappointed, found that founders, managers, executives and business-development people dominated, and recomputed for them alone: 33 per cent. He then built a detailed profile of that high-expectation customer and spent three quarters building for her. The score reached 58 per cent. The segment was chosen by behaviour, how people felt about losing the product, and it changed what was built.

The segment scorecard

When two or three candidate segments survive the arithmetic, score each from one to five on five criteria and pick the one with no ones. Pain intensity: the rupee or hour figure from the interviews, not the adjective. Trigger frequency: how many members enter the buying state in a year; a segment where the trigger fires once a decade is a lifetime of cold calls. Reachability: whether there is a room they are already in, a WhatsApp group, an association, a distributor, a market street, or whether each one must be found alone. Ability to pay and who signs: the alternative’s cost sets the ceiling and the signer sets the sales cycle. Your advantage: the [founder-market fit](/library/founder-market-fit-unfair-advantage-test) that lets you serve this segment better than the next founder.

Do not multiply the scores; a weighted sum hides the one that kills the segment. Read the row and ask whether you would be glad to spend two years with these particular people. The answer is data too.

A segment is not a group of people who look alike. It is a group of people who would all fire the same thing on the same day for the same reason.

A worked example: restaurants in Pune

A founder has built purchasing software for restaurants and describes the market as “restaurants in Pune with ₹1 crore to ₹5 crore in annual sales”, a list of perhaps three thousand names from the delivery apps. Twenty interviews later the behaviours split cleanly. Single-outlet owners stand in their own kitchen and have no purchasing problem they will pay to solve; they can see the stock. Groups with five or more outlets have a central kitchen and a purchase manager, and the problem they describe is supplier contracts, a different job. The eleven owners who light up are those with two to four outlets, buying daily from mandi vendors on credit over WhatsApp, with a manager at each outlet and no purchase team. Their trigger is specific and recent: they opened the second outlet within the last six months, and on that day they stopped being able to see what was bought.

The segment is now owner-run restaurant groups in Pune with two to four outlets that added an outlet in the last six months. Perhaps three hundred fit the behaviour and, say, forty per cent met the trigger this year. At a close rate of one in four among those reached while the trigger is live and ₹36,000 a year, that is about thirty customers and ₹10.8 lakh in year one, against a conversation worth ₹9,000 rather than the ₹1,800 the demographic cut yields. Small, and the right size for a first year, because of what it changes. The product is no longer “inventory for restaurants” but “see both kitchens’ purchases on one phone before the vendor’s bill arrives”. The channel is no longer the delivery-app list but the interior contractors and kitchen-equipment dealers who fit out second outlets, and the licensing offices where new premises are registered. The message is the trigger itself: just opened your second outlet? Pricing follows the alternative, which is the owner’s own evenings and a trusted manager, and the roadmap’s next item is the fifth-outlet group with its central kitchen, which is the next pin in Moore’s bowling alley rather than a feature on this one.

What changes when the segment changes, and when to revisit it

A segment has earned its place when it changes five things, and if it changes fewer than three it was a demographic in disguise. The product: which feature is first and which is cut. The channel: the room where the trigger is visible. The message: the trigger named in the customer’s words. The price: anchored to this segment’s alternative. The roadmap order: the adjacent segment that shares the product but not the trigger, served second.

Revisit the segment after every twenty interviews or every quarter, whichever comes first, and ask three questions. Are the new customers still arriving through the trigger, or has the trigger stopped predicting purchase? Has the close rate among triggered buyers held, or is the message drifting generic? Has a second behaviour appeared in the interviews often enough to be a segment of its own? When the honest answer to the first is no, the segment has been exhausted or misread, and the arithmetic above is where the next one is chosen. Write the segment sentence at the top of the [ideal customer profile](/library/ideal-customer-profile-on-one-page), which is the next lesson, and date it.

The calculator holds the founder’s estimates and nothing else; replace them with interview data as it arrives. The sources are short, and the Superhuman essay is the best worked example of behavioural segmentation in print.