How to do it
Screen harder than you interview
The quality of the answer depends almost entirely on who is in the room. One screening question — 'when did this last happen to you?' — removes most of the people who would have been politely useless. Somebody who cannot remember the last occurrence does not have the problem often enough to buy a solution.
Count repetition, not interviews
The number that matters is how many times you have heard the same specific thing, not how many calls you have taken. Keep a tally of distinct workarounds mentioned. When a whole conversation adds nothing to that list, you are near saturation.
Stop when two consecutive interviews add nothing new
That is the practical saturation test. It usually happens between five and eight for a tight segment, and much later for a loose one — which is itself the finding: if you are at fifteen and still hearing new problems, your segment is too broad to build for.
Re-open the count when you change segment
Saturation is per segment. Eight interviews with agency owners tells you nothing about in-house teams. Founders who mix segments reach a false saturation and then build something that fits neither.
Worked example
Saturation is per segment, and that is the part people miss
Eight conversations with agency owners tells you about agency owners. It tells you nothing about in-house teams, even if both have the same job title on paper. Founders who mix segments reach a false saturation — the conversations stop producing new information because they have stopped being comparable.
The practical rule: if you change who you are talking to, restart the count. If you are fifteen interviews in and still hearing new problems, the segment is too broad to build for, and that is the finding.
Why surveys are not a shortcut
A survey scales the number of responses and destroys the thing that made the conversation useful. You cannot ask a follow-up, you cannot hear the hesitation, and every question is necessarily hypothetical — which is exactly the class of question that produces unreliable answers.
Fifty survey responses saying people would use your product is weaker evidence than one person describing what they did last Tuesday. Use surveys to size something you already understand, never to discover it.
What goes wrong
- Running a survey instead, which replaces depth with volume and hypotheticals with more hypotheticals
- Interviewing anyone willing rather than anyone qualified
- Pitching during the interview, which converts the rest of the call into politeness
- Mixing two segments and averaging their answers
- Stopping at two or three because the first ones went well
Is your number plausible?
Each check catches a different way the arithmetic can be right and the answer still wrong.
- If you cannot name three specific workarounds you heard more than once, you have not reached saturation
- If every interviewee liked the idea, check whether you described it — you were supposed to be asking about their week
- If the same person could not tell you when it last happened, their interview should not be in the count
Related answers
What if I can only get three interviews?
Three is enough to start hearing repetition and not enough to trust it. Treat what you heard as hypotheses to check rather than findings, and keep looking — if three is genuinely the ceiling, that itself says something about reachability.
Should I record the calls?
Ask, and most people say yes. The value is not the recording but the transcript search: when you are at interview six and someone mentions a spreadsheet, being able to find the other two mentions verbatim is what turns an impression into evidence.
Can I interview people who are not in my target segment?
Early on, yes — it helps you find the segment. Once you are counting toward saturation, no; keep those conversations in a separate pile or they will contaminate the count.