Founders ask me this question like there's a badge waiting on the other side of it. Do 100 interviews and unlock the achievement. Do 50 and get the participation trophy. I understand the appeal of a number. A number feels like progress you can report to a board, a spouse, or yourself at two in the morning when you're wondering if any of this is real.
Here's the uncomfortable answer: the number matters less than what you hear once you get there. I have watched founders run 40 interviews and learn nothing, because every question was really just their pitch wearing a question mark. I have watched other founders learn everything they needed in 12 conversations, because they asked about behavior instead of opinion and actually stopped talking long enough to hear the answer.
I coach founders through customer discovery and non-dilutive funding, and I run I-Corps style programs where this exact question comes up in week one, every time. So here's the real answer, with real thresholds attached, instead of the "it depends" dodge that helps nobody make a decision this week.
The real answer is saturation, not a number
Qualitative researchers call it saturation. You have done enough interviews when new conversations stop producing new information and start confirming what you already know. That is a pattern, not a headcount, and it is the actual finish line, even though it is a lot less satisfying to write on a slide than "we talked to 100 customers."
What saturation actually sounds like
You will know you are close when the fifth consecutive person describes the same core frustration, in roughly the same words, with no new wrinkle added. That repetition is the signal. It means you have found a real, shared problem instead of five unrelated complaints that happen to rhyme if you squint.
The three thresholds worth knowing
Ten to fifteen conversations, focused purely on the problem, is enough to tell you whether you are chasing something real or chasing your own assumptions. Twenty to thirty conversations gets you to what most qualitative researchers consider reliable saturation for a single customer segment, and B2B founders should lean toward the higher end of that range because buying decisions in that world involve more people and more variation between accounts. A hundred conversations is not a magic number either, but it is roughly where most founders stop being surprised by what they hear, which is its own kind of signal worth paying attention to.
None of those thresholds replace listening for repetition. They are guardrails, not a finish line you cross and then stop thinking.
Why founders undercount on purpose
Founders stop early for a specific, human reason: interviews that go well feel like validation, and validation feels like permission to go build the thing you already wanted to build. Five encouraging conversations feel like enough evidence when what you actually want is to stop interviewing and start coding. That is not research discipline. That is impatience wearing research's clothes.
The founders who avoid this trap treat every enthusiastic answer with more suspicion, not less. Polite enthusiasm is cheap. A customer describing, unprompted, the actual dollar cost of their current workaround is expensive to fake and worth far more than ten people saying your idea sounds cool.
What good looks like at each stage
Early discovery needs interviews that surface the problem in the customer's own words, without you mentioning your solution at all. Validation needs interviews that test whether your specific approach beats their current alternative, with real detail about switching costs and who else has to approve the decision. Later-stage interviews, closer to willingness to pay, need to include a real ask: would you commit budget, a pilot, or a signature to this today, not hypothetically next quarter.
Each stage needs a different number of conversations because each stage is answering a different question. Confusing the stages is how founders end up with 60 interviews that all answer the same easy question and zero that answer the hard one.
The mistake that burns interviews without producing evidence
The single most common way founders waste a good interview count: they pitch instead of ask. "Would you use a product that does X" produces polite agreement, not evidence. "Walk me through the last time you actually ran into this problem" produces a real story, with real detail, that you can act on. The customer development method Steve Blank built this discipline around exists specifically to keep founders asking about the past instead of guessing about the future.
Fifteen well-run interviews that stick to real behavior beat sixty interviews that quietly turned into a focus group for your own idea. Quality of question changes what a given number of conversations is actually worth, which is why I keep coming back to saturation instead of a magic count. For the field guide on what those questions should actually sound like, see your first 100 customer conversations. And once the evidence is solid, the next problem is usually turning it into a pitch that survives a real conversation.
Customer discovery isn't optional, and it isn't a box you check once before moving on to the parts of building a company that feel more productive. The founder who builds without the commercial instinct discovery gives you almost never recovers that ground later, no matter how good the engineering turns out to be.
Frequently asked questions
Is 5 customer interviews enough?
Five is enough to notice a pattern worth chasing, not enough to trust it. Treat five interviews as a reason to schedule ten more, never as validation you can build on. Founders who stop at five almost always mistake early enthusiasm for evidence.
How many customer interviews before I write code?
At minimum 10 to 15 focused on the problem, not the solution. If those conversations describe the same core pain in similar language, you have enough signal to sketch a prototype. If they describe different problems, keep interviewing before you build anything.
What counts as a real customer interview versus a casual conversation?
A real interview follows a structured guide, focuses on past behavior instead of hypothetical opinions, and gets documented immediately afterward. A hallway chat with a friend who says your idea sounds great is not research, no matter how good it feels.
Can I interview friends and family for customer discovery?
Avoid it when possible. People who care about you will soften criticism and inflate enthusiasm, which corrupts your data before you have collected any. Interview strangers who match your target customer profile, even when that is slower and less comfortable.
When do I stop interviewing and start building?
Stop counting interviews and start watching for repetition. When five consecutive conversations raise the same core problem with no new dimension, you have reached saturation for that segment. That signal matters more than any fixed number.
Do B2B startups need more interviews than consumer startups?
Usually yes. B2B buying involves more stakeholders, longer sales cycles, and more variation between accounts, so most B2B teams need 20 to 30 interviews to reach the same confidence a consumer startup might reach in half that.
Count the interviews if it helps you stay accountable. Just do not confuse the count with the evidence. The number gets you in the room. Listening for repetition is what actually tells you when to leave it and go build something.