Article
How Many Customer Interviews Do You Need?
Updated · Fonda editorial team ·How we write these pages
Short answer
Plan for 10 to 15 interviews for each customer group you are testing, and stop when two or three calls in a row tell you nothing new. Start with 5 or 6, review your notes, then continue in small batches. In one well-known study, 92% of the themes found across 60 interviews had already appeared in the first 12.
Your roadmap
- 1Explore with six calls
- 2Confirm with six more
- 3Check a second group
- 4Test willingness to pay
Key takeaways
- Start with 5 or 6 interviews in one customer group, then review before booking more.
- Expect a pattern by around 10 to 15 interviews in a narrow group. Broad or mixed groups need more.
- Each new customer group needs its own round of interviews.
- Stop when answers repeat, not when you hit a number.
- Interviews show what the problem is. A real offer, not more interviews, shows whether people will pay.
The short answer, and why there is no magic number
Customer interviews are qualitative research, so the goal is not a statistically valid sample. The goal is to hear the same problems, workarounds and reasons often enough that new calls stop surprising you. Researchers call this point saturation. The Nielsen Norman Group, a user research firm, says there is no golden number of interviews and suggests starting with a small sample of 5 or 6, analyzing as you go and adding people only while you are still learning new things.
For founders, that leads to a working default: 10 to 15 interviews per customer group, run in batches of about five, with a review after each batch. If your group is narrow, such as independent physical therapists in one city, the pattern may show up sooner. If it is broad, such as "small businesses", you will need more calls or a narrower group. The free customer interview kit includes a tally sheet built for a round of 10.
How many interviews by stage and goal
Scroll sideways to see the whole table.
| Stage or goal | Interviews to plan | What you are looking for | When to stop |
|---|---|---|---|
| Exploring the problem | 5 or 6 in one customer group | Which problems come up, and the words people use for them | When you can name the two or three problems people raise most |
| Confirming a pattern | 10 to 15 in total, same group | Whether the same problem, trigger and workaround keep showing up | When two or three calls in a row add nothing new |
| Each extra customer group | Another 5 to 10 per group, more if the group is broad | Whether this group has the problem in the same way and handles it the same way | Apply the same stop rule inside each group |
| Testing willingness to pay | 5 to 10 conversations that end in a real offer | Whether people agree to a price, a deposit, a pre-order or a paid pilot | When you have enough real yes and no answers to set a first price |
| Mapping every need in a product area | 20 to 30 | A near-complete list of needs before you design a full product | When new interviews stop adding needs to the list |
Starting points, not rules. The counts for one group and for mapping every need draw on the research below. The other rows are rules of thumb. All assume the people in each group are fairly similar.
Willingness to pay is the row where interviews help least. An answer to "how much would you pay?" is a guess about the future, and the guess costs the person nothing. Use interviews to learn what the problem costs today, then test price with a real offer, as described in how to get pre-orders.
What the research on saturation says
- 70%
of 114 themes found across 60 interviews appeared in the first 6
Guest, Bunce and Johnson, Field Methods, as reported in Guest, Namey and Chen, PLOS ONE (2020), 2006 - 92%
of those themes appeared within the first 12 interviews
Guest, Bunce and Johnson, Field Methods, as reported in Guest, Namey and Chen, PLOS ONE (2020), 2006 - 9 to 17
interviews reached saturation in the empirical studies reviewed, mostly with similar participants and narrow goals
Hennink and Kaiser, Social Science & Medicine, 2022 - 20 to 30
interviews uncovered an estimated 90% to 95% of customer needs, in research by Griffin and Hauser
Nielsen Norman Group, summarizing Griffin and Hauser, 2021
The best-known study is by Guest, Bunce and Johnson, published in the journal Field Methods in 2006. They analyzed 60 interviews to see how many it took before new themes stopped appearing. According to a 2020 paper by Guest and colleagues, 70% of the 114 themes turned up in the first 6 interviews and 92% within the first 12. A 2022 systematic review by Monique Hennink and Bonnie Kaiser found that studies using real interview data reached saturation within 9 to 17 interviews, particularly when participants were similar and the research goal was narrow.
Two cautions for founders. First, these studies mostly used fairly similar participants, while an early market is often a mix. The Nielsen Norman Group notes that broader goals and more varied participants need more interviews. Second, saturation means you have heard the themes, not that demand exists. Twelve people describing the same problem tells you the problem is real for that group. It does not tell you how many such people there are or whether they will pay.
Practitioners land in a similar range. Steve Blank, who teaches customer development, had teams in his Lean LaunchPad classes at Berkeley and Stanford talk to 10 customers before the class started, and their weekly slides reported more than 10 interviews a week.
The stop rule: stop when answers repeat
A number tells you how many calls to book. The stop rule tells you when you are done. After every batch of about five calls, list the themes you heard: problems, triggers, workarounds, money spent and the words people used. Then mark which themes are new.
Stop rule
Stop interviewing a customer group when the last 2 or 3 interviews added no new problem, workaround or buying reason, and you can roughly predict what the next person will say. If you are still surprised, keep going.
Do not stop early just because the first few calls were exciting. Before you call a round done, check that:
- You talked to strangers who fit your group, not only friends or past colleagues.
- You recruited from at least two channels, so one community's habits do not skew the answers.
- Each interview produced a specific story about a recent event, not general opinions.
- You kept the questions focused on the same few assumptions, so the answers can be compared.
If answers never repeat, more interviews will not fix it. Your group is probably too broad, or the problem differs from person to person. Narrow the group and start a new count.
What counts as a signal and what is just politeness
Most people want to be kind to someone who asks for help. Y Combinator's advice on talking to users is to discard compliments and fluff and ask about specifics from the person's past. Count an interview toward your pattern only for its signals, not its politeness.
Scroll sideways to see the whole table.
| Politeness (do not count) | Signal (count it) |
|---|---|
| "That's a great idea." | "I spent two evenings last month rebuilding our schedule by hand." |
| "I would definitely use that." | "We pay for a tool that does part of this, and I hate it." |
| "Yeah, that sounds annoying." | A specific story with a date, the people involved and a cost |
| "Let me know when it is ready." | "Can I try it next week with my real data?" |
| "I would pay something for that." | A deposit, a pre-order or a signed pilot |
| "Everyone has this problem." | An actual introduction to a colleague who deals with it |
| Nodding along to your description | Showing you a spreadsheet, template or workaround they built |
Example phrases written to illustrate the difference.
A quick test: could the person have said it without having the problem? Anyone can say "great idea". Only someone with the problem can tell you what it cost them last Tuesday. These customer interview questions steer toward stories, and the kit's tally sheet counts only signals.
A worked example
This is an illustration, not real data. Say you want to build a booking tool for independent dog groomers.
- 1
Explore with six calls.
You interview six groomers who run their own salons. Five mention no-shows, three mention messy scheduling by text and one mentions card payments. No-shows look like the main problem.
- 2
Confirm with six more.
In calls 7 to 12, four more groomers describe a recent no-show and what it cost them. Calls 10, 11 and 12 add no new problem or workaround, so for salon owners the pattern is clear at 12.
- 3
Check a second group.
Mobile groomers who drive to clients may differ, so you start a new count. After six calls, they care less about no-shows and more about route planning. That is a different problem, so you note it and stay focused on salons.
- 4
Test willingness to pay.
You go back to the eight salon owners who gave the most specific answers and offer a paid pilot at a set monthly price. Three pay a deposit, two say not now and three do not reply. That is a price signal no interview could give you.
In total that is 18 interviews and 8 offers. The number was never the goal. Each round stopped when it had answered its question.
Interviews are not surveys
Interview counts are about hearing patterns, so 12 interviews cannot tell you that a set share of dog groomers has a problem. If you need a percentage, write a short survey that uses the words people used in your interviews, and use the survey sample size calculator to work out how many responses you need for the margin of error you want.
The order matters: interviews first, to learn what to ask; a survey second, to measure how common the problem is; real offers third, to see who will pay. In Fonda, an AI co-founder for people starting a business, Interviews sit inside the Validate phase between Key risks and the Verdict, and you can ask Marty, your co-founder, questions at any step. It is free to start, and pricing explains paid use.
Frequently asked questions
Can I do fewer than 10 customer interviews?
Yes, if the group is narrow and answers repeat early. Five or six calls can show you the main themes. Treat a small round as a first look, and run more before you spend real money on building.
How many customer interviews should I do per week?
As many as you can review properly. Booking calls close together helps you compare them. Teams in Steve Blank's Lean LaunchPad classes reported more than 10 a week, but they were working on it as a full course. If you are building on the side, a batch of five followed by a review is a realistic unit.
Do I need more interviews for a B2B idea?
Often, yes. In a business, the user, the buyer and the person who approves spending may see the problem differently. Treat each role as its own group with its own count.
Do interviews with friends count?
Count them only if the friend truly fits your customer group, and weigh their enthusiasm lightly. Friends want to help you, which makes compliments more likely. Ask friends for introductions to people you do not know instead.
Should I keep interviewing after I start building?
Yes, with new questions. Once the problem is confirmed, interviews shift to how people use a first version and what stops them from buying. Regular contact keeps you from building for a customer you imagined.
Sources
- Nielsen Norman Group (Maria Rosala, 2021): interview sample size
- Guest, Namey and Chen (2020), A simple method to assess and report thematic saturation in qualitative research, PLOS ONE
- Hennink and Kaiser (2022), Sample sizes for saturation in qualitative research: a systematic review of empirical tests, Social Science & Medicine
- Steve Blank: New lessons learned from Berkeley and Stanford Lean LaunchPad classes (2014)
- Y Combinator: Startup School week 1 recap (Eric Migicovsky on how to talk to users)
Keep reading
- Free customer interview kit
- Customer interview questions
- How to do customer interviews
- Survey sample size calculator
- What is customer discovery?
- How to validate a startup idea
- The Best AI Tools for Founders in 2026 (Every Stage, Compared)
- How to Turn a Side Hustle into a Startup (The Right Way)
- How to Build a Landing Page That Tests Demand
- How to Build a Waitlist for a New Product
- How to Do a Competitor Analysis for a New Business