How to Run a Branch Customer Survey Without Letting Large Outlets Dominate

When a chain sends one customer survey to every outlet and then combines all responses, the largest branches usually dominate the result. That may be appropriate if the question is “What did the average respondent experience?” It is not appropriate if leadership wants to compare branch quality or understand whether a smaller region has a serious problem.

Decide what the headline represents

There are at least three legitimate questions: the average experience across all responses, the average branch experience, or the experience of a defined customer population. Each requires a different weighting decision. Do not choose weights after seeing which version makes the score look best.

If the business wants each branch to count equally, calculate branch-level scores and then summarise them using a declared rule. If it wants the experience of the average transaction, larger branches may appropriately contribute more. If response rates vary widely, report them alongside scores; a high score from a very small response count is not equally stable.

Example: two outlets, different volumes

Branch A receives 1,000 transactions and 100 survey responses with an average score of 4.2. Branch B receives 100 transactions and 30 responses with an average of 3.5. The combined response average and the average of the two branch scores answer different questions. Neither should be presented as the universal “chain satisfaction” score without explaining the method.

Avoid sampling traps

• Do not survey only customers who completed a successful transaction if the aim is to understand failed visits.
• Do not let managers hand-pick respondents.
• Keep the invitation process consistent, while respecting channel preferences and applicable privacy rules.
• Check whether one language or device type is underrepresented.
• Separate staff-entered complaints from customer-submitted survey responses.

Compare carefully

Show score, response count, invitation count, response rate and uncertainty or stability indicator. Use a minimum reporting threshold where a tiny sample could expose an individual or create volatile rankings. A branch with few responses should be labelled “insufficient evidence,” not automatically ranked last.

Turn results into action

Look at the specific customer task behind the score—booking, waiting, payment, finding the outlet or receiving a service. A chain-wide average may tell leaders that something changed, but branch-level comments and process data help identify why. Re-run the survey after a targeted intervention using a comparable method, and avoid claiming causation if other changes occurred simultaneously.

Source notes

• Internal editorial framework: Proposed survey design workflow; validate against relevant contracts, policies and applicable law..