How Many Customer Interviews Do You Actually Need?
For most early startup ideas, 10 to 15 focused customer interviews is enough to validate a problem, and you should stop when three or four interviews in a row stop surprising you. That “stop when it stops surprising you” rule matters more than the exact number, so the rest of this post is about how to hit it fast without wasting weeks.
A customer interview is just a conversation with someone who has the problem you think you’re solving, where you ask about their life and their past behavior instead of pitching your idea. If you’re new to the format, start with how to do customer interviews as a beginner and then come back here to figure out how many you actually need.
The real answer: stop when you stop learning
The number 10 to 15 is a starting estimate, not a rule. The thing you’re actually watching for is called saturation — the point where new interviews stop teaching you anything new. When your 12th interview says almost exactly what your 9th, 10th, and 11th said, you’ve hit it. Doing 20 more won’t change your decision; it’ll just make you feel busy.
Here’s the trap most first-time founders fall into. They either quit after two interviews because two friends said “cool idea,” or they do 40 interviews to avoid actually building anything. Both are ways of hiding. Two isn’t enough to see a pattern. Forty means you’re using research as a security blanket.
You’re looking for a pattern to repeat, then break. When you can predict roughly what the next person will say before they say it, you have your pattern. That usually happens somewhere between interview 8 and 15 for a specific, narrow problem.
How many interviews for validation, actually?
Because this depends on where you are, here’s a rough map you can use as a teen founder working a few hours a week.
| Your goal | Interviews to aim for | What “done” looks like |
|---|---|---|
| See if a problem is real | 8–12 | The same complaint shows up unprompted 5+ times |
| Compare two or three ideas | 5–7 per idea | One idea gets emotional reactions, the others get shrugs |
| Understand a segment deeply | 12–15 | You can finish people’s sentences |
| Test a specific solution | 6–10 | People ask when they can use it, or clearly don’t care |
A few things to notice. First, these ranges are small on purpose. You do not need a hundred interviews, and you almost certainly can’t get them while juggling school. Second, the number goes up the vaguer your target customer is. If you interview “students” in general, patterns take forever to appear because students are all different. If you interview “juniors at my school who play a varsity sport and are stressed about recruiting,” patterns show up in five conversations because those people are similar.
That’s the real lever: a tight customer definition cuts the number of interviews you need in half. Fix your target before you count.
What counts as a real interview (and what doesn’t)
Not all conversations count toward your number. A real customer interview:
- Is with someone who actually has the problem, not just anyone willing to talk.
- Digs into what they did last time the problem came up, not what they’d hypothetically do.
- Doesn’t pitch your idea until the very end, if at all.
- Leaves you with a specific story, not a “yeah that’d be useful.”
If your best friend says your idea is great, that does not count. Neither does your mom. This is exactly why my friends love it is not validation — people who care about you will lie to protect your feelings, and they’ll do it without even realizing. The fix is asking about their past, not your product, which is the whole point of the Mom Test: questions so grounded in real behavior that even your mom couldn’t accidentally lie.
So when you count to 10 or 15, count real interviews. Five sharp conversations with actual sufferers of the problem beat 30 polite ones with random classmates.
A simple process to get there fast
You don’t have unlimited time, so treat this like a one-week sprint. Here’s a schedule that fits around homework.
- Write your customer definition (30 minutes). One sentence: who has this problem and what makes them specific. If you can’t name a specific group, that’s your first problem to fix, not a reason to start interviewing everyone.
- List 15 people who fit (1 hour). Real names, not “some people on Reddit.” If you can’t find where they are, read where to find people to interview for your startup idea before you send a single message.
- Do 5 interviews (over 2–3 days). Fifteen minutes each. Ask about the last time the problem happened. Take notes on their exact words.
- Read your notes and look for repeats. After five, you’ll usually see two or three complaints starting to echo. Circle them.
- Do 5 more, aimed at your circled patterns. Now you’re confirming, not exploring. “You mentioned X earlier — walk me through the last time that happened.”
- Decide. If the pattern held across all ten, you have signal. If ten people gave ten unrelated answers, your problem is too fuzzy or too small — check how to tell if your startup idea is too small and consider narrowing before you continue.
Ten interviews at fifteen minutes each is two and a half hours of talking spread across a week. That’s genuinely doable even in season, and it’s the core of validating an idea before you build anything.
When you need more than 15 (and when you need fewer)
You’ll want to push past 15 in a few situations. If you’re serving two very different groups — say, both the students who’d use your app and the teachers who’d pay for it — treat those as separate interview sets and do 8 to 12 of each. If your first ten interviews are all over the place, that’s not a reason to do fifty more; it’s a signal your customer is too broad. Narrow, then re-interview a tighter group.
You can get away with fewer than ten in one case: when the reactions are extreme and consistent early. If your first six people all light up and ask “wait, can I use this now?” — or all six visibly don’t care — you’ve got your answer. Strong signal is strong signal. That’s when you move from talking to testing, using something like a fake door test or a concierge MVP where you deliver the solution by hand before you build the real thing.
One warning: don’t confuse polite interest with signal. “That sounds useful” is the sound of someone being nice. The signals that count are unprompted stories, people leaning in, and — best of all — someone trying to pay or sign up before you’ve built anything.
How to know your interviews are actually done
Run this quick check. You’re done when you can answer these without guessing:
- Who exactly has this problem, in one sentence?
- What do they do about it today, and why does that solution stink?
- How often does the problem happen, and how much does it cost them (time, money, or stress)?
- Would they change their behavior for a better option, based on something they’ve already done?
If you can answer all four with real quotes and stories behind them, stop interviewing. More conversations won’t sharpen those answers; building something and putting it in front of people will. If you can’t answer them after 15 interviews, the issue usually isn’t the number — it’s a customer definition that’s too broad or a problem that’s more of a mild annoyance than a real pain.
The goal was never a magic number. It’s reaching the moment where you can predict what the next person will say and you’re itching to build the fix. When you get there, you’re done.
Getting to that clean “stop” point is a lot faster with people asking you sharp questions each week. That’s the whole first week of the batch0 program: you pick a customer, run real interviews, and don’t move to building until your idea actually holds up. If you want structure and a live cohort to keep you honest, apply for free — you only pay if you get in.