Qualitative research sample size: how many participants do you need

"How many participants do I need?" is the most common question in qualitative research, and the least satisfying to answer, because the honest response is: it depends on what you are trying to learn. Qualitative research sample size does not follow the statistical logic that governs surveys. It follows a different logic entirely, and once you understand it, the question largely answers itself.

Here is that logic, the rough numbers experienced researchers actually use, and the reasoning you can offer when a stakeholder asks how eight interviews could possibly be enough.

What actually determines qualitative research sample size

Statistical sampling asks: how many responses do I need for my estimate to generalise to the population within an acceptable margin of error? Qualitative work is not making that kind of claim. You are not estimating a proportion; you are mapping a territory of needs, behaviours, and mental models, then checking your map against reality. As our guide to qualitative and quantitative research puts it, qual gives you the why, not the how many.

So the real determinants are different:

The breadth of your question. "How do freelancers chase late invoices?" needs fewer people than "how do small businesses manage money?"
The diversity of your users. More distinct roles and contexts means more people to cover them.
The kind of claim you want to make. "This confusion exists, and here is its shape" needs far fewer people than "this is our users' most common frustration."
Your capacity to analyse. Twenty interviews you skim teach you less than eight you analyse properly. Every session you cannot give real attention to is recruitment budget spent on nothing.

Saturation: the real stopping rule

The concept doing the heavy lifting is saturation: the point at which new sessions stop yielding new insight. The first interview is all surprises. By the fifth, patterns start repeating. Somewhere between six and twelve, for a focused question in a reasonably similar group, you find yourself predicting answers before participants give them. That is saturation, and it is your signal to stop.

Two honest caveats. First, saturation is a judgement, not a measurement. You notice it in the doing, which is why analysing as you go beats interviewing all week and analysing on Friday: you cannot recognise the point where learning flattens if you have not been tracking what you have learned.

Second, saturation applies within a group of broadly similar people, and that is where samples genuinely grow. If new nurses and veteran nurses plausibly experience your product differently, that is two segments, and each needs enough participants to reach its own saturation, often five or six per segment as a floor. Not because qualitative work secretly needs statistical power, but because covering genuinely different contexts is the whole point. Be ruthless about which segment differences you truly believe matter, though: split by everything and no budget on earth will save you.

The five users rule is about usability, not discovery

Nielsen's famous argument that five users uncover most usability problems is probably the most quoted number in the field, and the most misapplied. It comes from usability testing: five people attempting the same tasks on the same interface will, between them, trip over most of the significant problems, and additional users mostly rediscover known issues. The efficient move is to test five, fix what you found, then test five more on the improved design.

That logic holds because a usability test is narrow: same tasks, same product, looking for breakage. Discovery research is the opposite: open questions across varied lives, where five people rarely reveal the shape of anything. Quote the five users rule for usability rounds. Do not let it quietly cap your discovery work, and do not let a sceptical stakeholder use it in either direction: it neither justifies five discovery interviews nor discredits a usability round of five.

Rough numbers that hold up in practice

Treat these as starting points to adjust, never as laws:

Usability testing · five per round, then iterate and test again
Discovery interviews, one segment · six to ten
Discovery interviews, multiple segments · five or six per segment
Diary studies · eight to twelve enrolled, expecting some drop off
Field visits and contextual studies · three to six deep visits

Whatever number you land on, recruit a little past it. No-shows and weak screener matches happen in every study, and a participant who turns out not to fit your criteria teaches you very little. Planning for eight and recruiting ten is cheaper than re-opening recruitment mid study.

If these numbers feel small, remember what they buy. Not a percentage with a confidence interval, but a defensible understanding of why people do what they do, which is the thing no survey can give you.

Enough is a judgement you can defend

The uncomfortable but liberating truth is that "how many" is the wrong first question. Decide what you need to learn, who differs enough to matter, and what claim you will make at the end. Then interview until new sessions stop teaching you, and be honest with yourself about when that happens. When someone asks how you knew eight was enough, the answer is not a formula. It is: we kept going until we stopped being surprised, and here is what we found.

People of Research is a community and newsletter for UX researchers, qualitative researchers, and anyone who has ever defended a sample of eight to a room that wanted eight hundred. Join us and get the good stuff in your inbox.