Before writing a customer survey, I would write the decision it is supposed to improve. “Understand our customers better” is a reasonable ambition, but it does not tell anyone which questions belong in the survey or what happens after the answers arrive.
“Decide whether to fund a self-service onboarding option” is different. It identifies an investment, a practical uncertainty, and someone who needs to act. That is where a survey can earn its cost: reducing the chance of making an expensive decision on a weak assumption.
Start with the action the answers could change
List the realistic options. You might improve documentation, add guided setup, keep the existing process, or run a smaller pilot. Then ask which missing information could change the choice.
If every possible answer would lead to the same plan, stop. Either the decision has already been made or a survey is not the information you need. Sometimes a handful of observed onboarding sessions will reveal more than a broad preference questionnaire.
Include staff time spent designing the survey, recruiting respondents, providing incentives, analyzing answers, and reporting back. For a recurring survey, add tool costs and the work of maintaining questions, invitation lists, and reporting. Customers contribute time too. A longer questionnaire should earn that request with a clear use for each question.
Ask about a specific recent experience
For the onboarding decision, start with behavior: “During your first week using the product, which setup step, if any, did you need help completing?” Follow with the source of help and whether the task was completed.
That is more useful than asking whether customers would appreciate a better onboarding experience. Almost anyone can agree with an appealing promise. The decision needs information about the problem, its frequency, and its consequences.
Pew Research Center’s question-writing guidance explains why specific language, one concept per question, and careful response options matter. Apply those principles literally. Ask about speed and clarity separately. Include “did not try” where appropriate, so you do not mistake lack of experience for satisfaction.
Make room for an answer you did not anticipate
Use a brief open question before presenting your preferred explanations: “What, if anything, made setup difficult?” Then use focused follow-ups for the decision you need to make.
Read the answers without forcing every comment into the solution the team already wants to build. A request for better instructions might conceal a product error. A request for a phone call might reflect a high-risk decision that written content cannot reasonably resolve.
Test the questionnaire with a few suitable people before sending it widely. Ask what they think each question means. This is a practical check on whether you will receive interpretable answers, not an opportunity to coach them toward your intended response.
Check who is missing
A survey sent to current active customers can miss the people who abandoned setup. A newsletter poll reaches the people willing to open and respond. Those are useful groups, but they are not automatically representative of everyone affected by the decision.
The Census Bureau identifies nonresponse as a potential source of bias. For a business survey, compare respondents with the invited population using information you legitimately hold, such as customer tenure or onboarding completion. Report meaningful differences. More responses from the same narrow group do not resolve a missing group.
Translate the finding into a testable investment
Hypothetical example: a survey costing $1,500 suggests that unclear setup instructions are a recurring obstacle. The business considers a $6,000 documentation improvement, including staff production time and planned maintenance during the evaluation period. Each additional retained customer would provide $300 in contribution during that period after variable service costs.
The improvement needs 20 incremental retained customers to cover its $6,000 cost. Including the survey raises the combined investment to $7,500 and the break-even threshold to 25 customers. That makes the decision concrete. Survey agreement that documentation would be helpful does not establish that those 25 customers will stay.
Use the finding to design a limited pilot. Compare completion, retention, and support demand with a credible baseline or control. Count lower support spending only if it actually falls, or describe released capacity separately. Do not add savings already included in the per-customer contribution figure.
Close the decision
Continue if the pilot supports enough incremental contribution to justify the full investment. Revise if the obstacle is real but the proposed response does not help. Stop if the likely benefit cannot cover the cost, or if another intervention solves the problem more effectively.
Keep the survey result and the outcome distinct in the final report. One tells you what respondents said. The other tests what changed when the business acted. I want both, because a useful customer insight should improve a decision, and the decision still needs to earn its return.
For the numbers behind the decision, see how to calculate marketing ROI. Put the approach to work with the free marketing resources, or work with me on your marketing strategy and implementation.
