Skip to content
AITroveRead. Build. Understand.
Make this comfortable

Survey prompts: define the population and invitation frame

Last updated: 4 Oct 202611 min read
tutorial
AdvancedBy AITrove Editorial

A survey prompt should name the decision, target population, eligibility rule, invitation frame, recruitment channel, field period, and unit of analysis. A contact export is not the population merely because it is convenient. The model can identify gaps in a proposed frame and draft a recruitment brief, but it cannot make an opt-in group representative by wording the invitation more carefully. Store one invitation ID per eligible person and decide how duplicates, bounced messages, and unknown eligibility will be counted before fieldwork. Do not let a later summary silently substitute respondents for everyone the product serves.

Operational case

Meridian Parcel wants to understand pickup-window preferences among customers who used pickup in the last 30 days. Its contact list contains 240 unique invitation records, all from customers who enabled email notices. Customers who use pickup but lack email notices are outside that frame. The study memo names that gap. The assistant may draft a neutral invitation for those 240 contacts, yet the final report cannot claim to describe every Meridian customer or every pickup user. A duplicated address is resolved before invitation so the person is not counted twice.

Output
Target: recent pickup users
Frame: 240 unique email-notice contacts
Excluded: pickup users without email notices
Unit: one eligible person
Claim boundary: invited frame, not all customers

Performance and review cost

Deduplicating N contact records by stable participant key is O(N) expected time and O(N) memory. Checking frame coverage may require a separate population count and review of excluded channels. The extra intake work protects the meaning of every later percentage. Prompt text cannot repair coverage gaps after responses arrive; keep the frame manifest and recruitment version attached to the analysis.

Common Mistakes

  • Do not call the email-contact frame the entire customer population.
  • Do not count duplicate invitation addresses as separate people.
  • Do not invent eligibility for an unverified contact.

Connected lessons

prompt engineering
survey research
Storage details