Market Research and Customer Insight
Ask the right question first
Research follows a sequence: define the decision and question, choose a design, collect, analyze, and report with recommendations. Always consider secondary research first - internal CRM data, regulatory filings, Census, industry benchmarks. It is fast and cheap but may be dated or defined differently than you need. Primary research is tailored but slower and costlier.
Qualitative methods (focus groups, depth interviews, ethnographic observation, mystery shopping) explain why but do not generalize. Quantitative methods (surveys, experiments, tracking studies) measure how many/how much and support statistical inference. Strong programs use qualitative work to build hypotheses and quantitative work to test them.
On sampling: probability designs (simple random, systematic, stratified, cluster) support inference to the population; non-probability designs (convenience, judgment, quota, snowball) do not. Report sampling error and watch for non-sampling error - bad sampling frames, nonresponse bias, poorly worded items. Reliability is consistency; validity is measuring what you intended. Avoid leading and double-barreled questions, order effects, and social-desirability bias on sensitive financial topics such as debt, income, and credit problems.
Common metrics include CSAT, Net Promoter Score (share of 9–10 promoters minus share of 0–6 detractors), brand awareness and consideration tracking, and A/B tests judged on pre-specified metrics with adequate sample size.
Worked example: 500 respondents - 60% promoters, 25% passives, 15% detractors. NPS = 60 − 15 = +45. Passives are excluded from the score but not from the analysis.
Rule of thumb: a random sample of about 1,000 adults gives roughly ±3 percentage points at 95% confidence in a large population. Precision improves with the square root of sample size, so quadrupling n only halves the margin of error.