INDUSTRY FIELD GUIDE
Survey wording and weighting records for AI
Two surveys can report similar percentages while asking different questions or using different weights.
NAICS 541910 · Marketing Research and Public Opinion Polling
Two surveys can report similar percentages while asking different questions or using different weights. AAPOR disclosure standards include question wording, sampling and weighting context. A research firm can make an original exercise to test whether models carry those distinctions into a summary. AAPOR disclosure standards
Write two fictional survey waves about product packaging. Change the prompt in one, and supply weighted and unweighted tabulations with explicit denominators. Ask which comparison is valid and which limitations a reader needs to see. Score retrieval of wording, base and uncertainty separately.
Use invented responses and aggregates. Exclude panel identifiers, respondent-level records, private questionnaires and unpublished client studies. Document provenance and any permitted use of survey instruments.
Describe a synthetic survey evidence sample
Original Fiund exercise
Equal percentages can answer different packaging questions
A fabricated survey about a made-up package uses invented aggregate counts and illustrative weights. There are no respondents, panel records or client instruments behind the numbers.
| Record | Source | What it records |
|---|---|---|
| SW1 | Wave A wording and base | Question: 'Would you choose the recyclable box at the same price?' Unweighted yes 60 of 100; weighted yes 66 of weighted base 110. |
| SW2 | Wave B wording and base | Question: 'Would you pay more for the recyclable box?' Unweighted yes 54 of 90; weighted yes 45 of weighted base 100. |
| SW3 | Method note | Weights and samples differ between waves. No common-item bridge, sampling-error estimate or experiment supplied. |
Question for the model
Calculate the displayed weighted and unweighted shares, and assess whether these waves establish a change in preference for an identical proposition.
Show the reference answer and checks
SW1 gives 60% unweighted and 60% weighted. SW2 gives 60% unweighted and 45% weighted. Equal unweighted percentages conceal a different weighted result and, more fundamentally, different propositions: same-price choice versus willingness to pay more. SW3 supplies no bridge or uncertainty estimate. Report each wording and base alongside its percentage; do not present the 15-point weighted difference as an identified trend in an unchanged preference question.
- Selecting an unweighted numerator and a weighted denominator to calculate a hybrid share.
- Calling different question wording a time trend without a justified comparison design.
Fields to preserve
- Exact question and wave
- Defines the proposition measured.
- Numerator, denominator and weighting basis
- Makes each percentage reproducible.
- Sampling and comparability limitations
- Constrains the interpretation of differences.
Test another case
Add a Wave B same-price item with its own base of 40. Require item-specific comparison and prevent reuse of the pay-more question's denominator.
Keep examples synthetic and aggregate. Exclude respondent records and confidential questionnaires; illustrative percentages are not market findings.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.