Customer Discovery Interviews: A Practical Guide

Imagine a founder who wants to build an admin-support service for small shops. The owners say the idea sounds useful. But which task would it help them finish? Customer discovery interviews ask someone about work they recently did. What steps did they take? What got in the way?

What can customer discovery interviews tell you?

An interview is a talk to learn about one relevant person's experience. It is not a sales call. It is not a demand estimate for a market. Your sample is small and self-selected, so it cannot tell you how many people share a problem.

UK service-design guidance recommends open, neutral questions and real examples (Government Digital Service, 2017). This is practice guidance, not proof of sales. Memory is a report, not verified behavior.

Keep the report apart from what you think it means. Suppose a shop owner says she spends two hours each Friday reconciling orders. That is her account. If you write "admin is a major pain point," that is your own view. This account gives you detail. It does not measure demand. Describe first, then decide what to check next.

Who should you speak to first?

Start with people who face the work, not people who might. Include current customers, people using a workaround, and people who chose not to solve the problem. Official Australian business guidance names current customers, competitors' customers, and nonbuyers as research sources, alongside other information (Australian Government, n.d.). It describes general research scope. It does not show that a small sample predicts who will buy.

Work often involves more than one person. The user who does the job may not be the approver who signs off on spending, so list both roles separately. One proposed question to open that gap: "When you needed this solved, who else had to agree?"

Before you invite people, write two lines. First, your single learning question. An example: how do small shop owners handle weekly order reconciliation? Second, who you need to speak to. An example: has reconciled orders in the last month. Then note who your sample may miss. If you only ask people in your network, say so. Friends can help you rehearse questions. They may not be the customers you need to hear from.

To find people, you could ask a business contact to pass on your invitation. Ask them to pass on your note, and let the other person decide. Do not share anyone's details without permission. Mark exact quotes clearly. Label your own account of their words as a paraphrase.

How do you ask without steering the answer?

A short discussion guide for a service-business founder might run like this: When did this last come up? Walk me through what you did, step by step. Who else was involved? How do you handle it today, including any workaround? What did it cost in time, money, or delay? How did you decide?

No method proves that a person is telling you the truth.

What should you explain before the interview?

Cover purpose, data, use and sharing, recording, voluntary choice and withdrawal, and retention, then record the agreement and honour it (Government Digital Service, 2018). This bounded UK guidance sits in a government research setting, so check its local legal adequacy. Minimize private data, keep notes secure, and take specialist advice for sensitive topics or minors.

How do you turn mixed answers into useful notes?

Separate what was said from what you think it means, then keep both. Official practice guidance recommends distinguishing recorded observations from interpretation. It then suggests grouping the material, stating findings, and deciding what to do next (Government Digital Service, 2016). That is a suggested method, not proof of unbiased analysis. Its purpose is to stop interpretation from quietly replacing evidence.

Two shop owners both call weekly admin frustrating. One describes two hours reconciling orders each Friday. The other says the issue is rare and an assistant handles it. This hypothetical contrast is reported behavior, not independently observed fact. The accounts suggest different needs. They do not establish market size or that either would buy.

Here is an original worksheet you can copy, with two rows from that example.

P1's two hours is an estimate, not a stopwatch reading. P2's "rare" is vague and needs a number. Neither account proves purchase intent. When you compare accounts, do not turn four conversations into a percentage for the whole market. Look for patterns in the details: same steps, same timing, same workaround. Keep contradictions and contrary accounts visible. An unusual account can still point to an assumption you have not tested.

What should you test after the conversation?

Choose the unknown that matters most. Plan a small test around it. Use the worksheet's last column to choose. You might refine the audience, run a follow-up with the same person, observe one real workflow with permission, or offer a tightly bounded paid pilot of one narrow task.

Each option produces different evidence. A follow-up deepens one account. Observation checks whether described steps match reality. A small offer tests whether someone will pay for a defined task. None guarantees anything. Interview praise, a repeated problem, one pilot payment, or a fixed number of conversations does not equal product-market fit. Weigh price, reach, and whether you can deliver.

For help with costs, see building a service price you can explain. Keep the loop small and honest: describe, interpret, preserve contradictions, then test the weakest link.

For more on applying these habits to real decisions, see Mherie Vic business and life insights. If you want help framing a discovery plan, you can discuss advisory fit.

This article is for informational and educational purposes only and does not constitute financial, legal, tax, medical, or professional advice. Individual results vary.