Careers
Part of Customer research: how journey maps expose backstage constraints
How business teams run customer research well
Business teams run customer research well by matching each method to the claim it can support, then closing the loop with customers once the evidence lands.
What to take away
- Write the decision and the claim before you recruit. If no customer-facing choice can change, stop.
- Match the method to the claim: interviews explain why, surveys estimate how many, records show what happened.
- Pilot recruitment, consent, tasks, analysis and deletion as one journey, not as separate steps.
- Keep raw records and coding notes so a second researcher can reach your conclusion from the same files.
- Retire evidence when the product, the market or the customer group it described has moved.
Customer research is a chain: a decision, a claim, a method that can support that claim, and a change someone owns. Break one link and the study becomes an expensive opinion.
Start from the decision, not the question
Write down who will act, by when, and what they will do differently for each plausible result. Include the result you expect and the one you fear. A study that cannot change a customer-facing choice is a reading exercise.
Audit what you already hold
Support tickets, cancellation reasons, sales objections and product usage often answer the first question for free. Check the population each source covers, its date, and what it cannot say. Then scope the gap.
Match method to claim
Pick the method by the sentence you need to defend.
Method that carries it
- "Customers find this confusing"
- Moderated sessions
- "18% of buyers abandoned at step 3"
- Behavioral records
- "This wording lifts completion"
- Controlled experiment
- "Support wait times drive churn"
- Survey with a defined frame
What it cannot do
- "Customers find this confusing"
- Estimate how many
- "18% of buyers abandoned at step 3"
- Explain the reason
- "This wording lifts completion"
- Generalize beyond the tested group
- "Support wait times drive churn"
- Prove cause on its own
Qualitative work explains mechanism. Surveys estimate prevalence within a stated frame. Records show captured action. Experiments test a specified change. Combine methods only when each answers a separate question.
Keep the purpose honest
The Federal Trade Commission's guide to protecting personal information tells businesses to know what they hold, keep only what they need, protect it, and plan for incidents. Tell participants what you will record, who will see it, how long you keep it, and how they withdraw. Keep research separate from promotion.
Pilot the whole journey
Run recruitment through deletion once before fielding. Test the screener, the consent language, the tasks, the recording setup, the export and the coding plan. Fix defects, then freeze the version you field.
Make analysis traceable
Keep originals, define codes before you count, and log exclusions. Write customer words, observation, interpretation and recommendation as separate lines. A quote shows one person said it. It never shows how many would.
Close the customer loop
Give every finding an owner, a change, a guardrail and a review date. Record why you rejected the findings you rejected; that record is where most customer research mistakes get caught early. When the review date arrives, compare what happened with what you predicted.
Worked example: a checkout complaint
A team hears that checkout confuses buyers. Interviews explain the confusion but cannot size it. Records show where people stop. An experiment tests one fix against a control. Each step answers a different sentence, and none of them substitutes for the others.
Assign the practice
The NIST Privacy Framework starting guide describes a voluntary process for identifying and managing privacy risk. Use it to name who owns each dataset and response. It is not legal clearance.
The CISA business-system logging guidance explains how event records support security review. Keep access, change and correction logs for research files, without treating a log as proof that a metric is right.
W3C's accessibility, usability, and inclusion resource explains how usability methods and user involvement complement accessibility standards rather than replacing them. Plan for buyers, former customers, non-customers and people the current experience excludes.
For the wider method set behind these steps, see primary market research best practices.
Common questions
When should evidence be retired?
Retire or refresh it after material change in customers, product, pricing, policy, market or method. Evidence about a checkout that no longer exists cannot guide the current one. Date every finding and set a review point when you file it.
Can quotations prove prevalence?
No. A quotation shows that something was said, not how often it happens. Prevalence needs a defined population, a sample drawn from it, a measurement and a stated uncertainty. Report quotes as illustration and put the number beside them from a source that can carry it.
What if the findings are inconclusive?
Inconclusive results still have value: they stop an unsupported decision. Record what the study could not settle and what would settle it. Then decide whether a different method, a larger sample or no further spend is the honest next step.
Who should own the change after the study?
The person who owns the customer-facing decision, not the researcher. Researchers supply evidence and limits; owners commit to the change, the guardrail and the review date. Without a named owner, findings stay in a slide deck.
