A startup can conduct many customer conversations and still learn very little. The problem is often the structure of the interview: founders explain their idea, ask whether it sounds useful, and leave with encouraging opinions that are difficult to act on.
A more useful approach investigates a real task, examines recent behavior, and connects observations to a specific product decision. The process can remain lightweight while becoming much more disciplined.
Begin with a decision you need to make
Before recruiting participants, write down the uncertainty the research should address. For example: do independent repair businesses struggle more with booking jobs or keeping customers informed afterward?
This is more focused than asking whether businesses “need better software.” It also provides a basis for deciding whom to interview.
The UK Government Service Manual recommends choosing research methods around the questions a team needs to answer and feeding findings into planning and prioritization. Although written for public services, that approach can be adapted to early product discovery.
Ask about a recent example
An illustrative opening question might be: “Walk me through the last job that needed several customer updates.” Follow the sequence from the initial trigger to the final outcome.
Ask what tools were used, where information was copied, who became involved, and what happened when the process broke down. If the participant can safely demonstrate a workflow, observe it rather than relying only on a general description.
Government interview guidance recommends open, neutral questions, follow-up questions, and attention to stories and real examples. It also advises preparing a discussion guide while allowing the conversation to develop naturally.
Avoid turning the interview into a product demonstration halfway through. If you want feedback on a prototype, make that a clearly separate activity.
Separate observation from interpretation
After each session, record what the participant actually said or did before deciding what it means.
For example, “The participant copied an appointment time into three systems” is an observation. “The market needs an integrated scheduling platform” is an interpretation that requires more evidence.
The Service Manual’s analysis guidance recommends extracting observations, grouping related material, developing findings, and deciding actions. It also encourages involving observers in the analysis to reduce the influence of any one person’s interpretation.
A small startup can apply this with a shared document. Keep the original observation beside the proposed finding so another team member can inspect the reasoning.
Choose the next smallest test
Suppose several participants describe the same communication problem. A reasonable next experiment might be a simple message template or a manually delivered service that tests whether the proposed improvement helps.
Specify what you want to learn from that experiment. Will users return to it? Does it remove a repeated step? Can the team deliver it consistently? These are suggested product-learning questions, not universal success thresholds.
Also record contradictory evidence. A participant whose process already works may reveal an important difference in business size, customer type, or workflow.
End each research cycle with a decision: continue investigating, test a particular intervention, narrow the target group, or stop pursuing the idea. Interviews become valuable when they change what the startup does next.
