How Many Customer Interviews Is Enough? What the Research Says, and When to Stop

•7 min read

The short answer: plan for 10 to 15 interviews per customer segment, and stop when two or three conversations in a row add nothing new to the themes you plan to act on. The two peer-reviewed studies people cite found that the range of themes showed up within 9 to 12 interviews, and that a thorough understanding of each theme took 16 to 24. Those are health-research figures, not product figures, but they line up with what I see in discovery work: by the tenth real conversation you have heard the main patterns. The open questions are whether you understand them, and whether your "ten" is ten people from one segment or three people from each of four.

What the two studies measured

Most "how many interviews" advice traces back to two papers. Both measured saturation, the point where new interviews stop producing new themes.

Guest, Bunce and Johnson (Field Methods, 2006). Sixty in-depth interviews with women in two West African countries. Saturation occurred within the first twelve interviews, and the basic elements for metathemes were present as early as six.

Hennink, Kaiser and Marconi (Qualitative Health Research, 2016). Twenty-five in-depth interviews. They split saturation in two. Code saturation, where the range of thematic issues was identified, arrived at nine interviews. Meaning saturation, a richly textured understanding of those issues, took 16 to 24.

Two caveats before you paste these into a planning doc. Both are health studies of one fairly uniform population, not founders interviewing a mixed list of prospects. And both measured when themes stopped appearing in their own data, which is not a rule for every interview program. The verbatim sentences, the DOIs and the caveats are on our customer discovery statistics page.

The gap between Hennink's two numbers is the useful part. Hearing every theme takes fewer interviews than understanding each one. If your question is "what are the top problems?", nine to twelve is in range. If your question is "why does this problem happen, what do they do instead, and what would make them switch?", budget for the high end.

Count per segment, not per project

Fifteen interviews spread across founders, agency owners and enterprise PMs is five interviews per segment. You will reach saturation on nothing. Each group has its own problems, workarounds and language, and a theme that is loud in one is noise in another.

Define the segment before you count. A segment is a group of people who share the problem, the context it shows up in, and roughly the way they would buy. If you cannot say who is in and who is out in one sentence, you have a list, not a segment.

Then plan per segment. Ten to fifteen per segment is a reasonable budget for most B2B discovery. Three segments means 30 to 45 conversations, and most teams cannot do that in one round. Pick the segment you would bet the next quarter on and go deep there first.

A stopping rule you can run with a spreadsheet

Saturation is easy to claim and hard to see without a system. This rule needs one sheet.

  1. After each interview, list the themes it supports. Themes, not quotes: "manual reporting eats Friday afternoons," "nobody owns the handoff to finance."
  2. Mark which themes are new. The first few interviews will be mostly new. By the eighth or ninth, a new theme should be rare.
  3. Stop when two or three consecutive interviews add no new theme to the set you plan to act on. One more interview after that is cheap insurance.

Two failure modes show up with this rule. If every interview is still surprising you at number twelve, you have a crowd, not a segment, and more interviews will not fix it. Go back and tighten who you are talking to. If you hit saturation by interview five, check your questions before you celebrate. Leading questions produce fast, clean, worthless agreement.

I said this in Stop Vibe-Coding Your Way to Nowhere and it still holds. We have seen people talk to 30 customers and learn nothing because they asked leading questions and heard what they wanted to hear, and we have seen teams talk to eight people and change direction because they asked open questions and listened. If you are hearing the same patterns after a handful of real conversations, you might have enough. If every conversation surprises you, keep going.

Count people, not quotes

A theme backed by twelve quotes sounds solid. If eight of those quotes came from one talkative customer, it is a theme backed by five people. One person who repeats a point five times is one data point.

This matters more once AI is doing the clustering, because AI output arrives looking finished whether or not it is right; how to analyze customer interview transcripts with AI walks through the spot-check. Put the count of distinct people next to every theme, and make sure none of those people is your own interviewer.

A plan that fits a real calendar

For one segment:

  • Round one, six to eight interviews. Open questions about the problem, the last time it happened, and what they did instead. Synthesize after the round, not after each call, so the first interviews do not anchor the rest.
  • Round two, four to seven interviews. Test the round-one themes without leading. Ask about the problem again; do not ask "do you agree that reporting is painful?" If round two mostly confirms round one, stop. If it reshuffles the themes, you learned something about the segment and probably need a third round.
  • A thin trickle after that. Two conversations a month per segment keeps the theme map current and catches a market shift before the roadmap does.

Ten to fifteen is a planning number. The stopping rule decides when you are done.

Where UpSight fits

UpSight clusters the evidence from your interviews into themes, and every theme links to the verbatim quotes behind it with speaker and timestamp, so "is this five people or one loud one" is a click rather than a re-read. You can import calls from Zoom, Google Meet, Microsoft Teams, Fathom, Granola and Otter instead of re-recording, and when a new interview lands you can see whether it added a theme or confirmed one.

Pricing is published: Free at $0, Pro at $29 per month, Team at $39 per user per month. The Free plan covers a first round.

FAQ

Is five customer interviews enough?

Five is enough to find out whether you are asking the right people the right questions. It is not enough to act on. In Guest's study the basic elements of the big themes were present by six interviews, but saturation on the full set came at twelve. Treat five as a pilot round.

How many customer interviews do I need for B2B SaaS discovery?

Ten to fifteen per segment is a reasonable budget. Multiply by the number of segments you are deciding between, then pick the one you would bet on and do it first. Fifteen interviews across three segments is five per segment, which is not enough for any of them.

Should I interview 50 or 100 customers?

Only if you have that many segments, or you are measuring how widespread a theme is after you already understand it. Past saturation, extra interviews cost time without changing the themes. A short survey to a larger group is the cheaper way to size a theme once you know what the themes are.

How many interviews before AI-generated themes are reliable?

The same number as for human-generated themes: enough that the themes stop changing. The extra step with AI is verification. Check that each theme's quotes say what the theme claims and that the people count is real. Themes from three interviews are hypotheses whether a person or a model wrote them.

How many lapsed customers should I talk to about why they left?

Start with 10 to 15. That is usually enough to hear the main reasons repeat. Win-back starts with a question covers how to run those conversations.

Next step

Pick one segment, schedule six interviews, and write down the themes after each one. If you want the counting and the receipts handled for you, see how UpSight runs customer discovery.

Try this on three of your own calls.

Free plan, no card. Every answer shows the quote behind it.

Start Free Trial

Founders and product leads: see the customer discovery platform. Consultants: see customer discovery for consultants. Already have recordings? See customer call analysis.