How many participants does qualitative research need?

“How many people do we need to talk to?” is the first question every qualitative study faces, and the most common answers are folklore. The honest answer is: it depends on saturation, on segments, and on what each additional interview costs you. All three are worth understanding before you recruit anyone.

Saturation is the real stopping rule

You have interviewed enough people when new interviews stop teaching you new things. Researchers call that saturation, and it is the only principled stopping rule qualitative work has. The most cited empirical study of it, by Guest, Bunce and Johnson in 2006, coded sixty interviews and found that the large majority of themes had appeared by interview twelve, with the basic structure visible after six.

Saturation is measured against a specific population and a specific question. Change either, and the counter resets.

The five-user rule, used correctly

Jakob Nielsen’s famous five-user heuristic says five participants surface most usability problems. It is sound advice in its original habitat: evaluative testing of one interface with one audience, run iteratively. It does not transfer to discovery research, where you are mapping unknown territory rather than checking a known design against users.

Reasonable working bands, per audience segment:

• Usability evaluation of a single flow: 5 to 8 participants.
• Discovery interviews on a focused question: 8 to 15.
• Multi-segment work: apply the band per segment. Two distinct audiences at 8 each is 16, not 8.

The constraint was never curiosity

Most teams do not stop at five participants because saturation arrived. They stop because each live interview costs a moderator hour, a recruiting slot, a calendar fight, and a no-show risk. The marginal interview is expensive, so the sample shrinks to fit the calendar and the folklore gets invoked to justify it.

Async AI-moderated interviews change that arithmetic. When the moderator is software, the marginal interview costs money and zero researcher hours; in GhostClick the price is £25 per participant on top of £50 for the study, so a twelve-participant discovery study is £350 and no calendars. Whether you use that headroom for more participants or more segments, the decision returns to where it belongs: saturation, not scheduling. The trade-offs between the two modes are covered in async versus live interviews.

Watch completion, not just recruitment

A recruited participant is not a completed interview. Whatever mode you run, track where people drop out: a study that recruits fifteen and completes nine has a nine-person sample. Async formats help here too, since participants can pause and resume instead of abandoning, but the discipline is universal: plan your target on completed sessions, and over-recruit by a third.