The generative versus evaluative distinction is the one that most teams miss. Teams that skip generative research — the phase where you try to understand the user's world without a specific solution in mind — often spend months building the wrong things with high quality. The pattern is that evaluative research, which tests whether a specific solution works, is more comfortable because it validates decisions already made. Generative research is uncomfortable because it requires genuine openness to finding out you were wrong about the problem. The teams that build the right things consistently do generative research before they have a solution hypothesis, and evaluative research after.
Why five interviews are enough for qualitative patterns
The counterintuitive finding from user research methodology is that you reach thematic saturation quickly with qualitative interviews. With five to eight interviews, you are hearing roughly the same themes again with new examples. This is not because users are all the same but because the problems that affect 70% or more of your target population emerge quickly. The implication is that running 20 interviews before taking action is usually unnecessary. The incremental insight from interview nine through twenty is often marginal compared to the insight from one through eight, and the time would be better spent acting on what you already know.
The most common qualitative research mistake in product teams
Asking users what they want, rather than observing and asking about what they do. Users are not good at predicting their own behavior or designing solutions. Asking "would you use a feature that does X?" produces optimistic responses that do not predict actual usage. Asking "walk me through the last time you tried to do X" produces behavioral data that does predict what the product actually needs to do. The shift from solution-focused questions to behavior-focused questions is small in wording and enormous in the quality of insight it produces. Most teams know this rule and still violate it under time pressure.