पाठशाला Pathshala · ग्राहक Grāhak, The customer · Lesson 01 · Start

The customer interview, done properly

Most founders have already done twenty customer interviews and learned nothing from them, because they pitched, asked about the future and talked. Here is the protocol that produces evidence instead.

Pathshala, The Founder Library · 10 October 2026 · 7 min read

Twenty people told you they loved the idea and would definitely use it. Three months later none of them has. Nothing went wrong with the people. The interviews were designed to produce that result, and almost every first-time founder designs them the same way.

This lesson is the protocol. It borrows its spine from Eric Migicovsky’s Startup School lecture on how to talk to users and its rules of evidence from Rob Fitzpatrick’s The Mom Test, and it adds the parts that matter in an Indian market: who actually decides, how to get in the room, and what a number looks like when a customer gives you one.

Why most interviews produce nothing

Migicovsky names three mistakes, and they account for nearly all the wasted hours. Pitching: the founder describes the idea and the rest of the conversation is the customer reacting to it, politely. Hypotheticals: “would you use this?” and “would you pay for this?” produce answers about an imagined future that the customer has no stake in and will not remember. Talking too much: the founder fills every silence, and silence is where the customer was about to say the useful thing.

Fitzpatrick’s three rules are the cure, and they fit on an index card: talk about their life instead of your idea; ask about specifics in the past instead of generics or opinions about the future; talk less and listen more. The test in the book’s title is that a good question is one your mother could not lie to you about, because it asks what she did, not what she thinks of you.

There is a deeper reason the pitch ruins the interview. Indian business culture is courteous. A trader in Surat or an HR manager in Pune will not tell a young founder across the table that the idea is weak; they will say it is interesting and that they will think about it. Compliments are the default social response, not data. The only way to escape them is to never give the customer anything to compliment.

Who to talk to, and how many

Define the segment in one sentence before the first call, narrowly enough that two people in it would recognise each other’s problems. “Small businesses” is not a segment. “Owners of 20 to 50 person companies in Pune who run payroll themselves on a spreadsheet” is. A narrow segment makes the pattern visible in ten conversations; a broad one hides it in a hundred. Paul Graham’s advice in How to Get Startup Ideas, to look for problems you have yourself, is also the shortest route to a first segment: if you are in it, the first interview is with you, and the second is with the person at the next desk.

Migicovsky’s route in is unglamorous and works: test the questions on yourself, then on friends and colleagues through warm introductions; one or two people is enough to begin. From there, every interview ends with “who else should I talk to?” and the segment introduces you to itself. Industry events are the cheap way to meet twenty of them in a day; a YC company he describes found its customers by walking into fire stations. In India the equivalent is the trade association meeting, the wholesale market before it opens, the chartered accountant whose clients are all in your segment.

How many? Ten to twenty in a tightly defined segment, and you stop when the answers to “tell me about the last time” stop surprising you, usually around the twelfth. A single interview rarely needs more than fifteen minutes if the five questions are asked and nothing else. Respect that; the shorter the ask, the more people say yes.

The five questions

Migicovsky’s five, in the order they work. What is the hardest part about doing the thing you are trying to solve? Let them name it; do not offer a list. Tell me about the last time you encountered this problem. This is the hinge. Get the date, the setting, the sequence of what happened and what they did. Why was this hard? Their reason, not yours. What, if anything, have you done to try to solve it? The workaround: a hire, a spreadsheet, a WhatsApp group, a tool they pay for. What do you not love about the solutions you have already tried? The gap, in their words.

Then the questions that decide whether this person can become a customer: how often does this happen, how many hours or rupees does it cost, and who in their organisation would decide to spend money on it. Ask for the number and wait. “A lot” is not a number. “Two days at the end of every month” is.

The protocol, end to end

Everything above compresses into a list that fits on a phone. Use it before, during and after each conversation, and reset it for each round.

Reading what you heard

Three kinds of statement should be set aside the moment they appear. Compliments (“that is a great idea”) are the social reflex and carry no information; Fitzpatrick’s rule is that, apart from industry experts who have built similar businesses, most opinions are worthless. Fluff is anything in the future or the general tense: always, usually, never, would. Anchor it back to a specific occasion in the past and the actions already taken. Ideas (“you should add…”) are to be noted and not acted on; ask why they want it and what they do today without it. Behind every idea a customer offers is a job they are trying to get done; Clayton Christensen and his co-authors describe in Know Your Customers’ “Jobs to Be Done” how people hire a product to do a job, and that the job is never simply about function. Write down the job, not the feature.

The statement that does carry information is a commitment, and Fitzpatrick identifies three currencies for it: time, reputation and money. A second meeting with a colleague is time. An introduction to their accountant or their boss is reputation. A deposit, a signed letter of intent or a request for an invoice is money. An interview that ends in a compliment and no commitment is a polite no, however warm it felt.

The customer cannot lie to you about what they did last month. They can lie to you, kindly, about anything you ask them to imagine.

A worked example: payroll in Pune

A founder wants to build payroll software for small companies and has fifteen minutes with the owner of a thirty-person design agency in Pune. The wrong interview opens the laptop: “We are building a simple payroll tool for companies like yours, would this be useful?” The owner says yes, it looks clean, send me the link. Nothing has been learned and the founder leaves believing it has.

The right interview asks what the hardest part of paying thirty people is. The owner says it is not the salary calculation; a spreadsheet does that. It is the last week of the month, when the accountant needs attendance and leave data that lives on three people’s phones, and the provident fund and professional tax deposits have to be reconciled before the deadline. The last time was eight days ago; two salaries went out late because one manager was travelling and the leave data arrived on the thirty-first. It happens every month. The workaround is a part-time accountant at ₹12,000 a month and a WhatsApp group called “Attendance” that nobody updates on time. What she does not love is that the accountant is paid to chase, not to account. She decides on spend up to about ₹25,000 a month herself.

Now the founder has a scored row: recent, monthly, costed at ₹12,000 a month plus two late salaries, worked around, and the person decides. The product has also moved: it is not “simple payroll”, it is collecting attendance and leave from managers before the accountant needs it. And the owner has agreed to introduce the founder to the accountant and to two other agency owners. That is time and reputation committed. Ten more interviews like it, and the company has a segment, a price ceiling and a first feature, before a line of code.

A weekly ritual

Book five interviews a week until the segment stops surprising you, and keep three habits. First, write every interview up within thirty minutes, with quotes marked as quotes, because by the next morning memory will have improved the answers. Second, keep the ledger: one row per conversation, scored on recency, frequency, cost, workaround and authority, so that twenty interviews become a table rather than an impression. Third, on Friday, read the week’s rows together and change one question for next week. The customer interview is a skill, and it improves at the rate you review it.

The protocol is a method, not a guarantee; the customers decide. The sources below are short and worth reading in full before the first interview.