Investigate a decision with appropriate evidence
Market research should reduce a defined uncertainty relevant to a decision. Clarify what needs to be learned before selecting a method or collecting information. Existing official statistics, internal records, prior studies, and external publications can identify gaps, but their date, population, definitions, and collection methods must be checked. Internal sales are not the whole market, and a respected old report is not automatically a current estimate after conditions have changed.
Match the method to the question. Qualitative interviews can explore experiences, language, motivations, and mechanisms in depth. Focus groups can reveal interaction and contrasting views, but agreement in a room is not a population approval rate. Quantitative estimates require an appropriate target population, sampling frame, selection process, and measurement. A large volunteer fan survey can remain systematically unrepresentative. Increasing the number of people in an incomplete frame does not restore excluded groups.
Question design affects what the answers mean. Avoid leading praise, double-barreled items, vague frequency terms, and forced substantive opinions from people without relevant experience. Pretest wording and response scales. In multilingual Swiss research, check conceptual equivalence instead of assuming literal translation produces comparable measures. Protect participants' information and explain the intended use appropriately. Recognize social desirability, nonresponse, and interviewer confirmation bias as possible influences rather than treating every recorded answer as an unmediated fact about the market.
Interpret summaries carefully. Means, medians, and distributions describe different aspects of a dataset. A small group of high purchases can move an average without changing the typical customer's behavior. Distinguish percentage points from relative change. Correlation alone does not establish causation, and a before-and-after result can be mixed with seasonality or other simultaneous changes. A margin of error depends on design assumptions and does not automatically cover every bias. Failure to detect a difference is not universal proof of equivalence. Present observed evidence, interpretation, assumptions, limitations, and recommendations separately. Document coding and selection rules so another reviewer can understand how the result was reached. Research becomes useful when findings return to the original decision and support a justified next step.