The final survey memo should identify the population of interest, actual invitation frame, field dates, instrument version, response dispositions, item wording, denominator, and limitations. Separate descriptive sample results from population claims and product hypotheses. The model can draft prose from a locked results table, but it should not recompute figures in free text or broaden 'eligible complete respondents' into 'customers.' A reviewer must trace each quoted comment and numerical statement to its response and item record. Release should stop if branch logic, wording versions, or privacy checks remain unresolved.
Make this comfortable
Survey prompts: release findings with explicit scope
Operational case
Meridian's draft headline says '60% of customers want narrower windows.' The approved table only shows 39 of 65 eligible complete respondents selecting that option. The invitation frame had 240 email-notice contacts, and 84 forms were submitted. The reviewer rewrites the claim to describe the eligible respondents and includes the exact Q7 wording and frame limitation. The team treats a narrower-window control as a design hypothesis, not a proven preference across the full customer base.
Claim: 39 of 65 eligible complete respondents selected narrower
Item: Q7-v2, pre-confirmation window preference
Frame: invited email-notice contacts only
Not supported: 60% of all customers
Decision: test optional narrow-window controlPerformance and review cost
Reconciling C claims against an indexed results ledger is O(C) lookups plus review of cited records. The review cost grows when a report hides question versions or branch exclusions, so keep those fields alongside each figure. A concise claim with a visible denominator can be verified faster than a persuasive paragraph that asks readers to guess who answered.
Common Mistakes
- Do not replace 'eligible respondents' with 'customers.'
- Do not detach a percentage from its item wording and denominator.
- Do not call a proposed control a measured improvement.
Connected lessons
- Production prompt engineering
- Prompt Engineering
- Report prompts: reconcile every claim with the packet
- Interview prompts: turn bounded findings into testable decisions
- Survey prompts: define the population and invitation frame
- Survey prompts: check wording against the intended construct
- Survey prompts: validate answer options and skip paths
- Survey prompts: pilot the instrument before fielding
- Survey prompts: reconcile invitations, submissions, and item denominators
- Survey prompts: code open text without exposing respondents
- Project: review Meridian pickup survey findings
- Survey prompt decisions
prompt engineering
survey research