Verified against ChatGPT · 2026-08-13
Design an engagement survey aimed at one specific attrition problem, not a generic annual checkbox
Builds a targeted employee survey structured around a specific, named business problem this organization is trying to diagnose, so the questions actually produce evidence for a decision instead of generic satisfaction scores nobody acts on.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Design an employee survey aimed specifically at the problem below — not a generic annual engagement survey, a diagnostic instrument built to produce evidence for a real decision.
THE SPECIFIC PROBLEM YOU'RE TRYING TO DIAGNOSE
Voluntary attrition on the customer support team has been 34% over the last 12 months, well above the company average of 12%, and leadership needs to know why before approving a retention budget.
WHO THIS SURVEY GOES TO
All 22 current customer support agents, across 3 team leads.
DECISION THIS SURVEY NEEDS TO INFORM
Whether to approve a compensation adjustment, a schedule-flexibility change, or a management-training investment, or some combination — the survey needs to point at which one matters most.
WHAT DATA YOU ALREADY HAVE
Exit interviews from the last 6 departures all mentioned inflexible scheduling as a factor, but none mentioned compensation directly.
CONSTRAINTS (length, anonymity, timing)
Must be anonymous, under 10 minutes to complete, needs to go out before next quarter's budget planning cycle.
Build the survey backward from the decision it needs to inform — every question should exist because its answer would change what that decision looks like, not because it's a standard engagement-survey topic. If existing data already tells you something (exit interview themes, attrition numbers by team or tenure), do not ask a question that would just re-confirm what you already know; use those existing findings to sharpen a more specific follow-up question instead. Mix a small number of scaled questions for trend-tracking with a larger share of specific, situational questions that ask about a concrete recent experience ("think of the last time you considered leaving — what was the specific trigger") rather than abstract satisfaction ratings, since concrete recall questions produce far more actionable detail than an abstract "how satisfied are you" score. Group questions so that a segment cut (by tenure, team, or role level, as relevant to the specific problem) would actually be possible without breaking anonymity, and flag if the given population is small enough that any segment cut risks identifying individual respondents.
WHAT NOT TO DO
Do not include generic filler questions common in engagement survey templates ("I would recommend this company as a great place to work") unless they specifically bear on the stated decision — every question in this survey has to earn its place by being use-case-relevant, not by being a standard question everyone asks.
OUTPUT FORMAT
1. The survey questions themselves, each tagged with which specific decision-relevant angle it addresses.
2. A short note on segment-cut risk given the stated population size.
3. One paragraph on what existing data made a generic question unnecessary and what more specific question replaced it.Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
The core mechanistic weakness of a generic annual engagement survey is that its questions are chosen to be broadly reusable across every company and team rather than to inform any specific decision, which produces data that is comparable year-over-year but rarely actionable for a particular live problem — building the survey backward from a named decision instead forces every question to justify its own inclusion by whether its answer would actually change that decision, which is a fundamentally different and much stricter design constraint. Using existing data (here, exit interviews already pointing at scheduling, not compensation) to eliminate redundant questions and sharpen follow-ups matters because asking a question you already know the general answer to wastes limited survey length and respondent attention on confirmation rather than the more specific detail still needed to act — if scheduling is already implicated, the useful question is not "is scheduling a problem" but "which specific scheduling constraint is the biggest issue," which existing data alone can't answer. Concrete situational recall questions ("think of the last time you considered leaving") outperform abstract satisfaction scales because they engage episodic memory of an actual event rather than asking someone to generate an abstract self-assessment on the spot, which tends to regress toward a socially neutral middle score — this is a well-established survey-methodology finding and is exactly why a satisfaction scale alone produces flat, low-signal results on a small team. Flagging segment-cut anonymity risk explicitly matters because a 22-person population split by 3 team leads means some segments could be as small as 6-8 people, small enough that a distinctive comment is plausibly attributable to a specific respondent, and a survey that promises anonymity but doesn't structurally protect it will be answered less honestly by respondents who correctly suspect they could be identified.
What you get back
Q: 'Think about the last time you seriously considered leaving in the past 6 months. What was the specific trigger?' (decision angle: distinguishes compensation vs. scheduling vs. management driver). Segment-cut note: with only 22 respondents across 3 leads, cutting by individual team risks groups as small as 6-8, so report findings by tenure band instead of by team lead to protect anonymity. Existing-data note: since exit interviews already implicate scheduling, the survey skips a general 'is scheduling flexible enough' question and instead asks which specific schedule constraint (weekend rotation, shift start time, PTO approval lag) is the biggest issue.
Verified against
ChatGPT GPT-5.1 · 2026-08-13
Changelog
- 2026-08-13 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
