Is there a right number of user interviews?
No fixed number is correct for every study, and anyone who gives one without asking about your question is guessing. What exists are rules of thumb that work well in practice. They are starting points, not proofs.
- Five to eight interviews: a good start for one question and one audience. Themes begin to repeat.
- Twelve to twenty interviews: a safer range when your audience is varied or the decision is expensive.
- Twenty to fifty or more: useful when you want to compare segments or see which themes are widespread.
Qualitative interviews tell you what exists and why. They do not tell you exact percentages. If you need a precise share of customers who feel something, you need a survey with a proper sample, and interviews can help you write the right survey questions.
What factors decide how many interviews I need?
Five things move the number up or down.
- Variety in your audience. If all participants use the product the same way, themes repeat early. If you have different roles, company sizes or use cases, each group needs its own interviews.
- Specificity of the question. "Why do trial users not finish setup" needs fewer than "How do people in our market choose tools".
- Cost of being wrong. A button label can be settled quickly. A new pricing model or product line deserves more evidence.
- Quality of each interview. Short, vague conversations give less than a few deep ones, so more shallow interviews do not make up for a weak guide.
- Your ability to analyze. Fifty transcripts are only useful if somebody can read them or if the findings point you back to the quotes.
Start from the decision. Ask what you would do differently if you heard A rather than B, and how sure you need to be.
How do I know when to stop?
Researchers use the idea of saturation: you stop when new interviews no longer add new themes. You can apply it without fancy methods.
- After each interview, write down the themes you heard.
- Mark which ones are new and which repeat.
- When three or four interviews in a row add nothing new to the list, you are close.
- Check that the interviews covered each segment, because saturation in one group says nothing about another.
Saturation is a judgment, not a measurement. Be wary of stopping because you are tired. Look for disconfirming cases, meaning a customer who behaves differently from the pattern, and talk to a few more before you stop if the pattern is thin.
The questions you ask matter as much as the count. Choose from the customer interview questions and keep them open.
How many interviews for each kind of study?
These ranges are practical suggestions, not research findings. Adjust them to your audience.
- Discovery of a new problem: eight to twelve, across at least two kinds of customer.
- Usability of one flow: five to eight per user type, with fixes between rounds.
- Concept or message testing: ten to twenty, so you hear a range of reactions.
- Churn interviews: ten to twenty of people who actually left, spread over reasons and plan types.
- Win and loss interviews: eight to fifteen of each, to see what repeats.
- Segment comparison: at least eight to twelve per segment.
- Large pattern check: thirty to a hundred or more, when you want to see how widely a theme from your first round appears.
The last item is where moderated work by hand hits a wall. A hundred live interviews means a hundred calendar slots. Our page on user interview software explains the alternative of sending one link.
Is it better to run interviews all at once or in rounds?
Rounds usually win. Start with five to eight interviews, read them, change the guide, then continue. You learn which questions work and which confuse people, and later interviews are better for it.
A sensible plan for a product team:
- Round one, week one: five to eight interviews to find the themes.
- Round two, week two: revise the guide and run ten more, including segments you missed.
- Round three, if needed: a larger set to test how common the main themes are.
Stop whenever the decision is clear. Doing more interviews after you already know what to do is a waste.
Rounds also protect you from a mistake in the guide. If a question is leading, you want to find out after six interviews, not sixty. A pilot with one friendly participant helps, and our discussion guide template includes a checklist.
How do I weigh the cost of more interviews?
Count the real cost per interview: recruiting or inviting, scheduling, moderating, notes, transcripts and analysis. With a live moderator, the cost grows with every conversation. With an AI moderator, the cost per interview falls as you use more of your plan limit.
A worked example from our plans. The Starter plan is $99 per month for 100 chat interviews, so $99 divided by 100 is about 99 cents per interview at the Starter limit. The Team plan is $249 for 400 chat interviews, about 62 cents per interview at the Team limit. Annual billing takes 50 percent off. A limit used in a period does not come back, so the per interview figure assumes you use the allowance. Voice minutes are counted separately. You can compare the options on AI moderator pricing.
Whatever method you use, the cost of too few interviews is a wrong decision, and the cost of too many is wasted time. Aim for enough to see the pattern with some margin.
Who should the interviews include?
Aim for the people your decision affects, not the people easiest to reach. Include recent customers, long-time customers and people who left. Include a few who use the product in unusual ways.
Moderators does not recruit participants. You invite your own customers or contacts with a link, so you know who is in the study. If you can only reach friendly users, treat the findings as limited and say so.
- Write who you want before you invite.
- Invite more people than you need, because not everyone replies.
- Keep notes on who answered, so you can see gaps.
- Compare the sample to the customer base to check for obvious bias.