Sample Size in Qualitative Research: How Many Is Enough?

When the people you study are similar and the question is narrow, studies that tested saturation reached it within 9 to 17 interviews or 4 to 8 focus groups. That is the finding of a review of 23 such tests (Hennink and Kaiser, 2022). Broader questions or mixed groups need more.

Saturation means the point where new data stops adding new insight. "Similar" means similar on whatever matters to your question, such as city, product use or life stage, not simply age.

The 16 to 24 figure comes from a single test, not a review, so treat it as a signal that understanding meaning takes more interviews than listing themes, not as a fixed rule.

Our recommendation: if you plan to compare segments, such as two cities or users and non-users, size each segment from this table as if it were its own study.

Start from 9 to 17 interviews or 4 to 8 focus groups, adjust for how broad your aim is and how mixed your group is, add a no-show buffer, and keep a stopping log to justify the final number. The six steps are:

1. Write the aim in one sentence and name the group. Test: if the sentence needs "and" to join two questions or two kinds of people, split it into two studies or two segments.
2. Take the starting number from the table above.
3. Adjust it. Move toward the top of the range, or beyond it, if the aim is broad, participants differ in ways that matter, your guide is loose, or you need meaning rather than a list of themes. Stay at the bottom if the aim is narrow, participants are tightly screened and the guide is structured.
4. Add the recruitment buffer. See the no-show table below.
5. Keep a stopping log. Record new codes after every interview and stop when the rule below is met.
6. Write the justification. Template: "We planned [N] interviews based on [source], recruited [N plus buffer] to allow for no-shows, and judged saturation by [rule]. The rule was met at interview [n]."

The rule comes from one study: sixty in-depth interviews with women in West African countries, where saturation occurred within the first twelve interviews and the basic elements of the broad, cross-cutting themes were present by six (Guest, Bunce and Johnson, 2006). It measured how fast themes appeared, in one fairly uniform group.

Later tests show the number moves with what you count:

Twelve works when you want to hear the range of themes in a similar group. It fails when you need to understand why, when the group is mixed, or when your guide is open and your coding is built from scratch.

A study needs fewer interviews when each interview carries more relevant information. This idea is called information power, and it rests on five things: the study aim, how specific the sample is, how much established theory the study builds on, the quality of the dialogue, and the analysis method (Malterud, Siersma and Guassora, 2016).

In plain terms, you need fewer people when:

• the question is narrow;
• everyone you recruit has direct experience of exactly what you are asking about;
• the interviewer draws out detail rather than short answers;
• you are testing ideas you already hold, not building them from scratch.

Phenomenological studies, which describe how a small group lives through one experience, fit this logic: depth per person is high, so the group is small and tightly defined. The tested ranges above come from thematic studies. For phenomenology, we recommend setting the number by information power and letting the stopping log decide when to end.

A sample of 12 has a 71.8% chance of hearing a theme held by 10% of your target group at least once. It takes 29 interviews to reach 95%. This is our calculation, not a published result.

Method: the chance of hearing a theme at least once is 1 − (1 − p)ⁿ, where p is the share of the group who hold it and n is the number of interviews. It assumes each interview is an independent, random draw from the group, which recruiting by quota or screener does not strictly satisfy.

Twelve interviews will almost certainly surface what most people think. They will often miss what one person in ten thinks. Qualitative research does not estimate how common a theme is, so hearing it once is a start, not proof.

Our recommendation: name the rarest theme you cannot afford to miss, read its row, and either raise the number or recruit deliberately for that type of person. Choosing people because they have a feature relevant to your question is called purposive sampling.

Log the new codes each interview adds and stop when the latest few add almost nothing. The method has three parts: a base size, a run length and a new-information threshold (Guest, Namey and Chen, 2020).

In their tests, base sizes of 4, 5 or 6 interviews, run lengths of 2 or 3, and thresholds of 5% or less, or none at all, were compared. Base size had almost no effect, so 4 is workable.

Our adaptation: after the first 4 interviews, take the latest 2 interviews and divide the number of new codes they added by the total codes found before them. Stop when the result is 5% or less. If your examiner or client wants a stricter standard, require 3 interviews in a row that add no new codes.

Example log. These numbers are invented to show the arithmetic:

Then run a meaning check, because a code count only shows that the themes have appeared. Pass: you can explain each main theme with examples from different participants, and the last two interviews did not change how you would describe any of them. Fail: if either interview shifted your explanation, keep going.

Recruit the number you need divided by one minus your expected no-show rate, rounded up. For 12 completed interviews and a 15% no-show rate, that is 12 ÷ 0.85, so 15 bookings. The rates below are planning scenarios, not measured figures.

Replace people who do not turn up with people who match the same screening criteria, or your last interviews will drift from the group you planned. Our guides on recruiting participants for a study in India and how to conduct focus groups in India cover screening and scheduling.

We recruit participants for paid qualitative studies across India, including metros, Tier 1 and Tier 2 cities, and urban, semi-urban and rural locations. Recruitment covers screening and coordination, and language needs are assessed project by project.

Send us your brief with the method, audience, geography, language, planned sample and timeline. We assess feasibility and reply with a tailored proposal and quote, because pricing is set per project. For a first project we are open to a small pilot or introductory arrangement, depending on scope.