Turn customer interview notes into a decision

A customer conversation is not a vote for your idea. Its value is the detail it gives you about a real situation: what happened, what the person tried, what changed, and what remains uncertain. This guide shows how to turn notes from a small number of exploratory interviews into a decision without presenting them as market statistics.

Use this after you have spoken with someone who recently faced the problem you are exploring. If you still need to find suitable participants, start with the customer interview screener. For the conversation itself, use the customer interview worksheet.

1. Write the decision before rereading the notes

Begin with the next decision you actually need to make. Examples include whether to keep investigating a task, which user group to interview next, whether to prototype a step, or whether to stop a feature idea. Do not begin with “prove that my product is good.” That framing encourages you to treat supportive comments as stronger evidence than contradictory behavior.

Write the decision in one sentence and list the assumptions that could change it. A simple record is enough:

• Decision due: [what I need to decide, and by when]
• Current assumption: [who has the problem, in what situation, and why it matters]
• Evidence that would weaken it: [an observation that would make another explanation more likely]
• What this review cannot establish: [for example, market size, retention or willingness to pay across a population]

This keeps a conversation useful even when the conclusion is “we do not know yet.” A clear unresolved question is more valuable than a confident answer built from vague praise.

2. Separate the account from your explanation

Make three columns before summarizing:

A sentence such as “people want an easier dashboard” usually combines a proposed solution with an interpretation. Try to recover the event underneath it: what the person was doing, what went wrong, what they used instead and what happened afterward. If the conversation did not establish those details, record them as unknown rather than filling the gap yourself.

3. Compare situations, not compliments

For each conversation, record a few comparable fields: participant role, recent trigger, task, current alternative, consequence, frequency as described by that person, who chooses or pays, and what they did next. A compact table makes it easier to see where accounts differ without forcing them into a single story.

Look for repeated situations and repeated workarounds, not repeated adjectives. Three people calling an idea “interesting” may simply mean the introduction was understandable. Two people describing the same manual step is a more specific observation, but it still does not tell you how common the problem is in a larger market. An exploratory sample helps you decide what to investigate next; it is not a representative survey.

Keep meaningful differences visible. A buyer may care about approval and risk while a daily user cares about speed. A new business and an established team may describe the same task differently. Do not average those groups together just to make the notes look consistent. Split them into segments when the role, trigger, current alternative or consequence changes the decision.

4. Search for the explanation that could prove you wrong

When a pattern seems promising, write at least one alternative explanation. If someone spends time on a workaround, perhaps the task is painful; perhaps a policy requires the workaround; perhaps a one-off deadline caused it; or perhaps the participant was trying to be helpful during your interview. These explanations lead to different next steps.

Look deliberately for a counterexample: a person who matches your target description but does not experience the problem, a person who solved it another way, or a case where the proposed improvement would create a new cost. A useful review includes evidence that weakens the preferred explanation as well as evidence that supports it.

Do not change your success criterion after seeing the notes. If the evidence is mixed, the result is mixed. You can narrow the audience, change the hypothesis, or run another test, but record why the new question follows from what you observed.

5. Choose the next test from the uncertainty

Match your next action to the question that remains:

• If you do not know whether the problem is recent, recruit people who performed the task within a defined period and ask them to reconstruct the event.
• If you do not know where the task breaks down, watch a person complete it with their current tools before showing a prototype.
• If you do not know whether a workflow is understandable, give a participant one realistic task with a sketch or sample output and avoid coaching.
• If you do not know whether anyone will take a next step, make a clear, truthful offer that you can fulfill and record the actual action. Do not substitute a fake checkout or an unfulfilled promise.
• If you do not know who can approve a purchase, ask how a comparable decision was made last time and who participated.

Set the scope, time limit, evidence to record and decision rule before you run the test. There is no universal number of interviews or pass threshold. Choose a modest exploratory scope suitable for the cost and reversibility of the decision, and do not describe the result as statistical validation.

Worked example — fictional

A fictional founder is considering a scheduling product for independent repair shops. One shop owner says, “I would definitely use that.” That sentence is not enough to choose a feature. The founder asks about the last missed appointment. The owner describes writing a reminder on paper, then getting pulled into a repair before making the call. A technician, however, says the bigger issue is that customers do not confirm whether a part has arrived.

The notes now contain two roles and two possible problems. “The shop needs an automated reminder app” is an interpretation, not an established fact. The founder records an unknown: which situation happens often enough to justify a change, and who would be responsible for it? The next test is to observe the existing confirmation process at two shops that match the intended audience, including a shop that says missed bookings are not a major issue. Those visits would help choose what to investigate; they would not establish demand across all repair businesses.

One-page evidence review

1. Decision: What choice will this review inform?
2. Best-supported observation: What specific recent event did a participant describe or demonstrate?
3. Interpretation: What do you think it means, and what is another explanation?
4. Variation: Which roles or situations did not fit the pattern?
5. Unknown: What can you not claim yet?
6. Next test: Who or what will you observe, what will you record, and when will you review it?

Respect participants and protect their information

Tell people why you are speaking with them and how you will use their comments. Ask permission before recording, publishing a quotation or sharing identifiable details. Store only notes you need, remove names from working documents where possible, and do not paste confidential customer records into this site's browser tools. See our privacy notice for information about the tools here.

Where this method stops

Interviews are useful for understanding context, language and past behavior. They cannot by themselves establish market size, reliable conversion, long-term retention or profitability. Self-reported frequency and willingness can be inaccurate, and people who agree to speak with you may differ from people who do not. Use interviews to make the next question sharper, then select a suitable behavioral or operational test for the claim you need to examine.