Customer Interviews Are Evidence—Once You Stop Treating Them as Quote Hunts

Account managers nominate responsive champions. The interview guide asks what they love. Marketing extracts three polished lines. The result may be sincere, but it says almost nothing about stalled adoption, quiet dissatisfaction, failed evaluation, or the people who left.

Start with the decision the research must inform. Build a recruitment matrix around the differences that could change that decision: new and mature users, strong and weak outcomes, active and churned accounts, simple and complex implementations, different roles, and people who declined an offer. The matrix is a coverage plan, not a claim of statistical representativeness.

Begin with a concrete episode: “Tell me about the last time this became urgent.” Walk through trigger, people involved, information sought, alternatives, decision, implementation, and consequence. Ask what happened next and what artifact existed—a calendar invite, spreadsheet, ticket, or approval note.

Avoid selling during the interview, stacking assumptions into the question, or rewarding praise with enthusiasm. A useful interviewer can sit with an awkward pause and ask “What makes you say that?” without turning the conversation into a cross-examination.

Explain the research purpose, participant burden, recording, who can access raw material, retention, withdrawal route, and how findings will be used. Ask separately before recording, attributing a quotation, naming the company, publishing a story, or contacting the person again.

Incentives should compensate time without buying a favorable answer. AAPOR’s survey standards emphasize quality controls across sampling, instruments, interviewing, coding, and analysis. Those disciplines travel well to interview operations even though qualitative interviews answer different questions from probability surveys.

Code observations with interview cohort, question context, strength of evidence, contradiction, and a link to the approved source excerpt. Write a short finding as claim, evidence, counterevidence, boundary, and proposed action. Do not let an AI summary become the only surviving record.

Stop when additional interviews add little decision-relevant information across the planned contrasts—not when a round number is reached. Report who was missing. A qualitative study can reveal language, sequences, constraints, and mechanisms; it cannot tell you that “72% of buyers” think something unless a suitable quantitative study did that work.