Operations
Part of Customer research: how journey maps expose backstage constraints
How to Research Customer Research: Method, Coverage, Rules
Match a customer research method to the question, then check group coverage, the rules that attach to the sample, and the conditions that end the study.
What to take away
- Pick the method from the decision, the groups it affects and the evidence that would change your mind.
- Plan on typical ranges5 to 8 participants per usability round, 12 to 20 interviews per segment, 300 to 1,000 completes for a survey.
- Treat participant counts as typical ranges, and read the formula behind any survey size before you accept a quote.
- Name the groups the sample must reach, and say which ones the claims do not cover.
- Write stop conditions before recruiting, then report a failed study instead of extending it.
Match the method to the research question
In customer research, the method follows the question, not the other way round. Start from the decision, the affected groups and the evidence that would change your mind. A fashionable method is the wrong one when it cannot answer that question.
Use the table as a planning aid. The participant counts are typical ranges, not standards.
Method that fits
- Why did renewals fall?
- Interviews, diary study
- How common is the problem?
- Survey with a known frame
- Can people finish the task?
- Moderated usability test
- Which feature matters most?
- Conjoint trade-off exercise
Typical participants
- Why did renewals fall?
- 12 to 20 per segment
- How common is the problem?
- 300 to 1,000 completes
- Can people finish the task?
- 5 to 8 per round
- Which feature matters most?
- 300 or more completes
Failure mode to watch
- Why did renewals fall?
- Only reachable customers get recruited
- How common is the problem?
- The frame excludes affected buyers
- Can people finish the task?
- Tasks written to confirm the design
- Which feature matters most?
- Attributes nobody would trade away
A focus group surfaces language and disagreement. It does not give you a rate. A survey gives a rate only when the sampling frame supports one, so frame, quotas and non-response belong in the plan.
A worked example: interview versus survey
Suppose renewals fell for two quarters running. Interviews with 12 to 20 lapsed customers can explain the mechanism: a price change, a missing integration, a support gap. They cannot tell you how many accounts share that reason.
A survey of 300 to 1,000 customers drawn from the billing frame can estimate prevalence, with the caveat that non-response tilts the result. Run the interviews first, then size the survey to the frame.
Before either starts, settle who to invite and what to tell them. Recruitment and consent questions come before the count, because they decide who can be reached.
What the counts can and cannot do
The counts in the table are starting points, not standards. Sample size follows the question: margin of error, confidence level and expected completion rate set the number, and the formulas behind sample size determination are worth reading before you accept a quote.
More completes do not fix a frame that never included the group you need. A large sample of the wrong people is still the wrong sample, and a smaller one drawn from the right frame can carry the claim.
Interviews and usability tests are counted per segment or per round, so each added segment brings its own interviews. Survey reports should state sample, frame, quotas and response, and the AAPOR standards article covers what US survey research teams disclose.
Coverage of customer groups
Ask which roles can approve, block or pay, and whether the plan reaches them. Then check the edges.
- Former customers and failed applicants
- Non-customers who chose a rival
- People excluded by device, language, location or disability
- Groups too small to generalize from, and where you will say so
Record sample, frame, quotas, response and attrition in the write-up. State which groups the claims do not cover.
Which rules attach to the sample
In the United States, commercial consumer research usually sits outside 45 CFR 46, the rule that governs federally funded human subjects work and requires IRB review. Touch health information or federal funding and expect an IRB.
California studies add CCPA and CPRA duties. Canadian panels fall under PIPEDA, which sets the ground rules for private-sector organizations collecting, using and disclosing personal information, and its principles cover consent, retention and access. When one sample crosses the border, the US and Canada privacy rules put the two regimes side by side.
Stop conditions and the written recommendation
Close with a method recommendation: the decisions it suits, the populations, the safeguards, the minimum evidence and the limits. Add skill needs and the conditions that end the study early.
Stop when the evidence crosses your action threshold, or when the study cannot reach the population it needs. Report that outcome instead of extending the field. Compare the final choice against market research strategy best practices that tie a decision to its evidence.
Common questions
How many participants do I need?
Use the typical ranges above as a start: 5 to 8 per usability round, 12 to 20 interviews per segment, 300 or more survey completes for a prevalence estimate. Adjust for frame, quotas and expected non-response.
Why run interviews before the survey?
Interviews explain the mechanism, and a survey then estimates how many customers share it. Running them in that order lets the survey be sized to the frame.
Does commercial consumer research need IRB review?
Usually it sits outside the federal rule. Touch health information or federal funding and expect an IRB.
How do I keep a study from drifting past its question?
Write the action threshold and stop conditions before recruitment opens. When the evidence crosses the threshold, or the study cannot reach its population, stop and report.
