How many customer interviews do I need before building?

By Bonaventure Ogeto, Founder and CEO, Mctaba Labs
Published

The short answer

Plan for 10 to 15 interviews with one narrow type of customer, and stop when three conversations in a row tell you nothing new. If you are still surprised after 15, your group is too broad, so narrow it and continue. The number matters less than who you speak to and whether you ask about past behaviour.

Founders get this wrong in two opposite ways. The first kind speaks to three friends, hears three versions of "nice idea", and starts building. The second kind speaks to sixty people over eight months, fills a notebook, and never builds anything, because one more conversation always feels safer than a decision.

Both are avoiding the same thing: a clear rule for when you know enough. This article gives you one. Plan for 10 to 15 interviews with one narrow group, and stop when three in a row tell you nothing new.

Why is 10 to 15 usually enough?

Because you are not running a survey. You are not trying to prove that 62 percent of anything is true. You are looking for a pattern so strong that it shows up in a handful of conversations.

Customer discovery, the name usually given to this stage, asks a blunt question: does this group have this problem badly enough to act on it? If the answer is yes, you tend to hear it early and often. By the fifth conversation people are repeating each other. By the tenth you can finish their sentences. If the answer is no, you hear that quickly as well: polite interest, no workaround, no cost.

Small numbers work only under one condition. The people must be alike. Fifteen conversations spread across teachers, boda riders, pharmacists and landlords are fifteen separate studies of one person each. Fifteen conversations with one kind of customer are a real sample of that kind.

How do I know when to stop?

Use a rule you set in advance, and write it at the top of your notes.

Stop when three interviews in a row add nothing new, and you can predict what the next person will say before they say it.

"Nothing new" needs a definition, or you will fool yourself. After each conversation, ask whether it changed any of these:

  • how you would describe the problem
  • what people do about it today
  • what it costs them
  • who decides and who pays
  • how you would reach more people like this

If the answer is no five times, that interview added nothing. Three of those in a row and you are done, whether you are at interview 9 or interview 14.

There is a floor. Do not stop before eight, even if the first five agree. Early agreement is sometimes a sign that you interviewed five people from the same WhatsApp group who all know each other.

When do I need more than 15?

When there is more than one kind of person involved. The number is per group, not per idea.

Your situation Who to interview Rough plan
One type of consumer People with the problem 10 to 15
A small business where the owner decides, pays and uses Owners 10 to 15
A business where one person uses and another pays Users and payers, separately 8 to 10 of each
An institution such as a school, sacco or clinic The user, the person who approves, the person who signs 5 to 8 organisations, two or three roles in each
A marketplace with two sides Each side separately 10 to 15 per side

If you pass fifteen in a single group and are still being surprised, do not push on to fifty. Split the group. "Farmers" might turn out to be dairy farmers who sell to a cooperative and dairy farmers who sell at the gate, with different problems. Pick the half that felt the problem most and continue there.

What counts as an interview?

Not everything you would like to count.

Counts: fifteen to thirty minutes, with one person who has the problem, mostly about what they did the last time it came up, with notes written the same day.

Does not count:

  • A chat where you explained your idea and they reacted to it.
  • A group conversation, where the loudest person answers for everyone.
  • A comment thread or a poll.
  • Someone telling you what "people like them" generally do.
  • Anyone who does not have the problem but has opinions about it.

Be strict. Ten real interviews beat forty conversations of the other kind.

What should I ask, so the interviews can be compared?

Ask every person the same five questions, in whatever order the conversation allows. Their job is to fill the same five boxes every time.

  1. "When did this last happen? Walk me through it."
  2. "What did you do about it?"
  3. "What did that cost you, in money or in time?"
  4. "Who else was involved, and who decides what gets paid for?"
  5. "Who else do you know who deals with this?"

Do not describe your product. If they ask what you are building, say you are still trying to understand the problem and will come back to them. The fifth question matters more than it looks. It is how ten interviews become fifteen, and whether the person offers a name tells you how seriously they take the problem.

How do I find 10 to 15 people?

Go where the group already gathers, and ask for a conversation, not a favour.

  • WhatsApp and Facebook groups for the trade or the estate. Ask the admin first, then post one short message saying what you want to learn and that you are not selling.
  • Walk in at the quiet hour. A shop at 11am on a Tuesday, a clinic in mid-afternoon. Ask for ten minutes and offer to come back.
  • One insider. A single supplier, association official or respected member of the trade can introduce you to five people in a day.
  • Referrals from each interview. Question five above.

A message that works:

Good morning. I am trying to understand how people who rent out boda bodas in Kisumu collect the daily payment from riders. I am not selling anything. Could I ask you about it for 15 minutes on a call this week?

How do I read the results?

With a tally sheet, filled in after every conversation. It turns a pile of impressions into something you can count.

Here is an illustration. Say you spoke to twelve people in Kisumu who each rent out two or more boda bodas, about collecting the daily payment from riders. The numbers are invented to show the method.

What you are tallying Count out of 12
Raised the problem before you mentioned it 9
Could describe a specific recent case 10
Already has a workaround, such as an exercise book, a daily call or a relative who collects 11
Could put a cost on it in shillings or hours 7
Pays for something today to manage it 3
Decides and pays by themselves 12
Offered an introduction to another owner 6
Interviews that added something new 1 to 9 yes, 10 to 12 no

Read it from the top. Nine of twelve raising the problem unprompted is a strong sign that it is real. Eleven with a workaround confirms that they act on it. Only three paying for anything today is the weak point: the pain is real, but most owners solve it with their own time, so price will be the hard conversation. Twelve of twelve deciding alone means a short sale. And the last row shows the stopping rule was met at interview 12.

Compare that with a sheet where three of twelve raise the problem and two have a workaround. You do not need interview 13. You need a different problem or a different group.

Should I keep interviewing after I start building?

Yes, at a lower rate. The 10 to 15 answer one question, which is whether the problem deserves a product. They do not answer whether people will pay, which needs a real offer with a price, or whether your product solves it, which needs people using it.

Once the build begins, keep two customer conversations a week in your calendar. Change the subject from "tell me about the problem" to "show me how you used it yesterday".

The interviews also decide what the first version contains. The workaround that eleven of twelve people described is the thing your product replaces, so it defines the main screen. The cost they named sets the ceiling on your price. The person who pays tells you who the product must send its receipts and reports to. If you can hand a builder a tally sheet instead of a feature wish list, you will get a smaller product that fits the people who are going to use it.

Key takeaways

  • Ten to fifteen interviews with one narrow customer group is enough for most first products.
  • Stop when three interviews in a row add nothing new to your tally sheet.
  • Still hearing new things after fifteen means the group is too broad, not that you need fifty.
  • Count each buyer type separately. Two sides of a marketplace means two sets of interviews.
  • Tally the same five facts from every conversation so that you compare evidence, not impressions.
  • Interviews answer whether the problem is real. Only a payment test answers whether people will buy.

Frequently asked questions

Can I use a Google Form or a survey instead of interviews?
Not at this stage. A survey can only ask the questions you already thought of, and early on you do not know the right questions. Conversations surprise you, and surprises are the point. Surveys become useful later, when you want to check how common something is among people you cannot speak to one by one.
Should I pay people or give airtime for an interview?
A small thank you afterwards is fine, especially if you took someone away from their work. Do not advertise a payment beforehand, because you will attract people who want the airtime and not people who have the problem. Most people with a real problem are glad to talk about it for fifteen minutes.
Do conversations with friends and family count?
Only if they are truly in your customer group, and even then count them at half weight. They know you, they want you to succeed, and they will soften what they say. Use them to practise your questions and to get introductions to strangers who fit the group.
Is it better to interview in person, on a call or on WhatsApp?
In person at their place of work is best, because you see the exercise book, the phone and the workaround with your own eyes. A phone call is second. WhatsApp text is weakest for a first conversation, because people give short, tidy answers, but it works well for a follow-up question.
What if I cannot find even ten people to interview?
Treat that as a finding about the idea. If you cannot reach ten people with the problem when all you want is a conversation, reaching hundreds when you want their money will be harder. Either choose a group you have a route into, or find one person inside the group who will introduce you.

Sources

  1. The Mom Test: How to Talk to Customers and Learn If Your Business Is a Good Idea When Everyone Is Lying to You, Rob Fitzpatrick (book, 2013). Checked on 19 November 2026.
  2. The Four Steps to the Epiphany: Successful Strategies for Products That Win, Steve Blank (book, 2005). Checked on 19 November 2026.