“How many people do we need to interview?” is the most common question in every qual research brief. It is also the question that reveals the most about how well the team understands qualitative methodology.
The honest answer is: it depends on your research question, your method, and how many distinct segments you are studying. But there is a principled way to arrive at the right number, and it has nothing to do with statistical significance.
Why Qualitative Sample Size Works Differently
Quantitative research uses sample size to achieve statistical power: enough respondents that your results can be generalised to the broader population within a defined margin of error. A survey that aims to represent India’s urban consumers might need 1,000 to 2,000 respondents.
Qualitative research has a completely different goal. You are not trying to measure how many people feel something. You are trying to understand how and why they feel it. The depth of each data point matters far more than the number of data points.
This is why adding the 30th IDI respondent to a study that reached saturation at 12 does not add value. It adds cost, time, and more of the same.
Thematic Saturation: The Real Benchmark
The principle that drives qualitative sample sizing is thematic saturation. You have reached saturation when new interviews stop producing new themes or insights. The last several interviews you conduct are confirming what you already know rather than adding to it.
In most well-designed studies with a clear research scope, saturation occurs earlier than researchers expect. This is not a failure. It is the method working correctly. The goal was never volume. The goal was understanding.
Research on saturation points consistently finds that for a single, focused research question with a relatively homogenous respondent group, 6 to 10 in-depth interviews is often sufficient. Nielsen Norman Group’s landmark research on usability testing found that 5 respondents reveal 85% of usability problems. The principle transfers broadly to qualitative research.
Sample Size by Method
Method | Typical Range | Per Segment | Notes |
In-depth interviews | 8 to 15 | 8 to 12 | Lower end for homogenous groups, higher for diverse populations |
Focus groups | 3 to 6 groups | 2 groups minimum | Each group: 6 to 8 participants |
AI-moderated interviews | 20 to 100+ | 15 to 30 | Lower cost enables larger samples for richer thematic coverage |
Ethnographic observation | 5 to 10 | 5 minimum | Extended observation reduces the number needed |
Dyadic interviews | 8 to 12 pairs | 6 to 8 pairs | Pairs of connected respondents (e.g. couples, colleagues) |
How Segmentation Multiplies Your Sample
The single biggest driver of sample size in qualitative research is segmentation. Each segment you want to understand separately requires its own saturation point.
If you are studying purchase behaviour across three customer segments (new customers, lapsed customers, and advocates), you need enough interviews within each segment to reach saturation for that group. A sample that is diverse in aggregate but thin within each segment will not produce reliable segment-level insight.
A common error is to treat a total sample of 20 IDIs as adequate for a study with four segments. That gives you five respondents per segment, which is unlikely to reach saturation for any of them.
The right framing is: what is my minimum per-segment sample, and how many segments do I have? Multiply those, and you have your starting point. Then consider whether any segments are more internally diverse than others and add accordingly.
When to Stop: Practical Signals
Beyond the theoretical concept of saturation, there are practical signals that tell you when you have enough data:
- You are hearing the same stories with different names attached to them
- New interviews are extending existing themes rather than introducing new ones
- You can predict with reasonable accuracy what a new respondent will say based on their profile
- Your analysis team is no longer updating the theme map after each new transcript
If you are three interviews into a study and already seeing strong thematic patterns, that is not a reason to stop. It is a reason to revisit whether your respondent screening is tight enough. Early saturation sometimes signals that you are talking to too homogenous a group.
AI Moderation Changes the Calculus
Traditional qualitative research operates under a cost constraint that shapes sample sizes. A 15-IDI study takes three to four weeks and costs significantly more than a survey of the same scope. This cost pushes teams to under-sample.
AI-moderated interviews break this constraint. Since the per-interview cost is dramatically lower with an AI moderator, it is now practical to run 40 to 60 AI-moderated IDIs at a cost that previously bought you 10 to 15 human-moderated ones.
This does not mean you should always run 60 interviews. It means you can make the sample size decision based on research quality rather than budget. If your study genuinely benefits from 40 interviews across four segments, you can now afford to do it properly.
Larger AI-moderated samples also produce more stable thematic analysis. With 50 interviews, you can be more confident that a theme appearing in 12 respondents is real rather than an artefact of a small, potentially unrepresentative sample.
Quick Reference: Sample Size by Study Type
Study type | Segments | Recommended total |
Brand perception (single segment) | 1 | 10 to 12 IDIs |
Customer journey (multi-segment) | 2 to 3 | 8 to 10 IDIs per segment |
Concept test (exploratory) | 1 to 2 | 8 to 12 IDIs |
Usage and attitude (broad scope) | 3 to 4 | 10 to 12 IDIs per segment |
AI-moderated discovery study | 2 to 4 | 20 to 30 per segment (if budget allows) |
Once you have your sample, see How to Write a Discussion Guide for AI Interviews to ensure you get the depth your sample can deliver. And for guidance on what to do with the data, see How to Analyse Qualitative Data from AI Interviews.