Analyze a small sample of customer research

Separate what happened from what you think it means

A handful of focused conversations can reveal language, workarounds, constraints, and questions worth testing. It cannot tell you what percentage of a broad market behaves the same way. Treat small-sample research as a way to develop and challenge explanations, not as a survey with a small margin of error.

Before reviewing notes, restate the research question and the people you intended to learn from. Keep a participant key that records only what you need for analysis: segment fit, relevant behavior, and session ID. Store contact details separately, restrict recordings, honor consent, and set a deletion date. GOV.UK’s research analysis guide recommends recording observations separately from interpretations; its notes and recording guidance covers consent and privacy considerations.

Make an evidence table

For each useful moment, write one row:

Keep quotes short and exact. Label paraphrases as paraphrases. Do not write “users need automation” in the observation column; that is an interpretation. Keep contradictory evidence and outliers visible rather than removing them because they do not fit the current idea.

Cluster carefully and count honestly

Group observations by recurring task, trigger, workaround, consequence, buyer, or barrier. Begin with what participants actually did, then ask what explanation could account for the pattern. Add alternative explanations. For example, “three people did not finish checkout” could mean unclear pricing, lack of purchase authority, missing trust information, or simply an irrelevant offer. Do not turn the first explanation into a finding without another check.

When reporting frequency, show the denominator and selection method: “4 of 7 invited independent consultants we interviewed described rebuilding the same report weekly.” This is not “57% of consultants.” The participants were recruited for a reason and the group is too selected to represent the population. Quantitative rates need a sampling and measurement design appropriate to the claim.

When you need to measure a bounded behavior across a larger, defined group, plan a survey with explicit recruitment and sample limits; see design a useful customer survey.

GOV.UK recommends timely analysis, extracting discrete observations, sorting themes, and turning agreed findings into actions. Its round-planning guidance is written for government service teams; a solo builder can adapt the evidence separation and follow-up loop without recreating a large workshop.

Write a finding that leads to a decision

Use this structure:

Finding: [specific segment] currently [observed behavior] when [trigger]. This suggests [interpretation], because [evidence]. We do not yet know [uncertainty]. Next we will [test or product decision].

Example: “Three of five interviewed studio owners exported invoice data each Friday because their accounting package could not match two client-specific fields. This suggests the mapping step is a recurring pain, but we do not yet know whether these owners would switch systems. Next, observe two more exports and test a priced import helper.”

After each round, decide to strengthen the evidence, narrow the segment, alter the offer, or stop. Recruit a different participant when a key alternative explanation remains. Use customer interviews, small usability tests, and the experiment brief to preserve a traceable path from question to next action.