Where the number comes from, and where judgment takes over
Most students hear "twelve interviews" and assume the number is arbitrary. It is not. The interview masterclass article will teach you how to run a good interview. This article answers a different question: why this many?
The honest answer combines three things. First, there is real qualitative research behind the number — and it converges within a narrow band. Second, there is practical math that makes 12 work specifically for early-stage founders. Third, there is a curve of learning that only reveals itself if you stay in the work past the point most students quit. Skip any of the three and the assignment makes less sense than it should.
For decades, qualitative researchers asked the question you are asking: how many interviews are enough to actually understand something? In 2006, Greg Guest and his colleagues at Family Health International ran a study of 60 in-depth interviews in West Africa and tracked when new themes stopped appearing in the data. They found that roughly 70 percent of themes emerged within the first six interviews. By interview twelve, the figure was over 90 percent. Adding more interviews after that produced diminishing returns.
This was not just one study. In 2022, Monique Hennink and Bonnie Kaiser published a systematic review of 23 empirical studies that tested the same question across very different contexts. Their finding: for relatively homogeneous populations with a narrowly defined research question, saturation typically occurs between 9 and 17 interviews. The most recent integrative review, by Wutich, Beresford, and Bernard in 2024, places theme saturation at around 9 interviews, meaning saturation at 24, and full theoretical saturation at 20 to 30+.
Twelve sits in the middle of the most replicated band. It is not a magic number. It is the rough point where, statistically, you have heard most of what your specific customer segment has to tell you.
The number is grounded in empirical research, not invented for the syllabus.
The research is about saturation in academic studies. Founder reality is different. A founder is not just trying to identify themes — they are trying to make a decision. The 12-interview target is also the smallest number that lets you do three useful things at once.
You can split your sample into two customer segments and still have six interviews per segment. Six is the lower bound for noticing a pattern within a group. If you are building a product that might serve two distinct types of buyer — say, working professionals and small business owners — twelve gives you the minimum to compare them.
You build in a buffer for the conversations that go nowhere. Realistically, two or three of your twelve interviews will produce thin data — the person was distracted, the connection was weak, the fit was wrong. Twelve gives you nine or ten conversations of usable depth, which is closer to the saturation point for one segment.
You force yourself past the moment where most students quit. The curve of customer interview learning is not linear. There is a discouraging middle — usually around interview five or six — where the conversations start sounding similar and you wonder if there is anything left to learn. Pushing past that point is where the second-order insights appear. Students who stop at six rarely get there.
Watch what actually happens across twelve interviews and a pattern emerges that surprises most founders.
Interviews one through three are usually about your nerves. You are still figuring out how to ask. You are too eager. You over-explain your idea. You walk away with thin notes and a vague sense that the person was being polite.
Interviews four through six are where you start hearing real things. Your questions get sharper. People start telling you about their lives instead of reacting to your pitch. You begin to notice patterns — but you also start hearing contradictions, and your confidence drops. This is normal. It is also the most common place where students give up.
Interviews seven through nine are where the noise resolves into signal. You start hearing the same objection from people who have nothing else in common. You hear the same unmet need described in different words. The contradictions turn out to be segmentation — different types of customer with different problems.
Interviews ten through twelve are where you stop hearing new things, and that is the point. Saturation is not a feeling of victory. It is a quiet recognition that your sample has told you what it knows. The Guest et al. research calls this "code saturation" — you are still learning, but at a sharply diminishing rate.
Most students quit at interview five or six, the exact point where the work was about to start paying off.
The 9–17 range from the Hennink and Kaiser review applies to homogeneous populations with narrow aims. ASEAN founders frequently work in conditions where neither holds.
If your venture spans more than one country, the research changes. Hagaman and Wutich's 2017 study on cross-cultural metatheme saturation found that 20 to 40 interviews were needed when the study aimed to identify patterns across cultural sites. A founder building a product for both Thailand and Vietnam is doing exactly this kind of cross-site research, even if they do not call it that.
The same applies if you are serving two genuinely different customer types — say, students and corporate trainers — or two very different geographies, urban Bangkok and rural Northeast Thailand. In those cases, twelve is your floor, not your target. The serious answer is six to twelve per segment, which means twenty-four for two segments and more for three.
This is one reason why the GVP assignment is twelve and the GVS phase, when you start executing, expects you to keep going. Twelve is what it takes to write a defensible problem definition. Real product-market fit work usually requires triple that, spread across whatever segments and markets you actually serve.
This is where the math hands the work back to you.
The research can tell you that twelve is a defensible minimum. It cannot tell you that your twelve are the right twelve. A founder who interviews twelve people from the same building, the same WhatsApp group, the same university year, has not done twelve interviews — they have done one interview twelve times. The variation in your sample matters more than the number in your sample. This is the part the academic literature calls "information power" — it is not just how many people you talk to, but how different they are from each other within the group you are trying to understand.
Judgment also decides when to stop earlier than twelve. If by interview six you are getting consistent, surprising, segment-defining answers, and your time is running out, the academically defensible move is to write up what you have and move on. Six rich interviews from carefully selected sources can produce more decision-grade insight than twelve shallow ones from convenient ones. The reverse is also true: if at interview twelve you are still hearing genuinely new things, the research and your common sense both say keep going.
Twelve is the floor. The ceiling is whenever you stop hearing new things — which is rarely as soon as you hope.
Here is the part of the assignment that pays for itself for years afterward, and the reason 12 is also a number worth thinking about as a long-term investment.
Twelve customers who agreed to be interviewed by you, gave you their time, and answered your questions are not just data points. They are twelve people who know what you are working on. If you stay in touch — a brief update three months later, a thank-you when something they said helped, an invitation to try the first version — most of them will become something more useful than any market report.
Some will become your first paying customers. Some will refer you to friends in their industry. Some will give you a second round of feedback that is more honest than the first because they have seen you act on the first round. A small percentage will become evangelists who tell their network about you without being asked.
This is also why the assignment specifies real conversations, not surveys. A survey gives you data. An interview gives you a relationship. The relationship is the asset. The data is the byproduct.
You are not being asked to do twelve interviews because someone arbitrarily picked the number. You are being asked because the qualitative research literature converges on roughly that figure for the kind of work you are doing, because the practical math of comparing customer segments and building in buffer requires close to that number, and because the learning curve of interviewing only resolves into useful signal somewhere past interview seven.
Twelve is the floor of a good answer to a hard question. It is not the ceiling. It is also not the goal. The goal is to know your customer well enough that you can describe them, their problem, and their world to a stranger and have that stranger say "I know exactly who you mean."
When you can do that, you have done the work. Twelve interviews is usually how long it takes.
Minimum of 12 structured conversations with potential customers — people who actually have the problem you are trying to solve.
Sample design first. Before you start, decide whether you are validating one customer segment or two. If two, you are aiming for six per segment, not twelve total — that is twenty-four. Be honest about which you are doing.
Diversity within the segment matters more than quantity. Twelve people from the same WhatsApp group is not twelve interviews. It is one interview twelve times. Cast a real net.
Track the curve. After every five interviews, write a short note: what is showing up consistently? What is new since last time? When new themes stop appearing, you are approaching saturation.
Stay in touch. The twelve customers you interview are your first twelve advocates — if you keep them posted as you build. Treat the assignment as the start of a relationship, not a transaction.
You are not interviewing customers to prove your idea is good. You are interviewing them to find out what is actually true.
Kasper-Tanakrit Sermsuksan is the Founder of SEA Bridge, Dean of SEA Bridge Institute of Entrepreneurship, and a Visiting Lecturer at Chulalongkorn University, Faculty of Engineering (Computer Engineering). Learn more →
