Consulting & Strategy prompt of the day

Survey readout: cut the crosstabs down to the few that change the recommendation

The survey came back with forty crosstabs and the deck is due Thursday. Find the cuts that actually change the answer and park the rest.

You are a consulting research lead turning raw survey results into a readout. I will paste the questionnaire, the sample description, and the tables.

Produce:

A) SAMPLE CHECK — who answered, the sample size per segment, and which segments are too small to report on their own. Name the minimum base you are using and apply it consistently.

B) DECISION-RELEVANT CUTS — the 3-5 findings that would change the engagement's recommendation, each stated as a so-what sentence with the exact figures, base sizes, and question number behind it.

C) INTERESTING BUT NOT DECISIVE — findings that are true but would not change the answer, listed in one line each for the appendix.

D) CAVEATS — question wording that could bias answers, segments with low bases, and any gap between what respondents say and what behavioral data would show.

E) NEXT CUT — the one additional analysis that would most sharpen the recommendation.

Inputs: [PASTE QUESTIONNAIRE] · [SAMPLE DESCRIPTION + FIELD DATES] · [PASTE RESULT TABLES] · [THE RECOMMENDATION THIS SURVEY SHOULD INFORM]

Rules: Do not invent percentages, base sizes, or significance — report only what the tables show, and write "not testable from these tables" where a difference cannot be judged. Verify every figure in the readout against the source tables before it reaches a slide. Keep respondent-level data and any personal identifiers out of consumer AI tools; aggregated tables only.

Why it works — Survey decks bloat because every crosstab looks like a finding. Sorting results into decision-relevant versus merely interesting forces the team to say which numbers would actually move the recommendation, and the sample check stops a twelve-person segment from becoming a headline. The caveats section is what a skeptical client will raise anyway, so it is cheaper to write it first.