ChatGPT prompt for surveys

Help me design a survey that produces usable evidence for a real decision.

Decision and objective: [DECISION AND OBJECTIVE]
Target respondents: [TARGET RESPONDENTS]
What is already known: [EXISTING EVIDENCE]
Topics to measure: [TOPICS]
Distribution and constraints: [DISTRIBUTION AND CONSTRAINTS]

First, state the decision this survey can inform and what it cannot establish. Turn each topic into a small set of measurable constructs. Recommend the shortest survey that can answer the decision question; do not add curiosity questions that will not change an action.

Create a survey with an opening consent or context statement appropriate to the audience, then group questions in a logical order: easy factual screening questions, current behavior or experience, specific evaluations, open explanation, and optional demographics last. Use one concept per question. Avoid leading language, double-barreled questions, absolutes such as “always,” undefined terms such as “regularly,” and answer choices that overlap or omit a likely option. Include “not applicable” and “prefer not to answer” only where they are meaningful.

For every closed question, provide the response scale, labels for anchors, and the intended analysis. Use balanced rating scales and label all points when the audience may interpret the scale differently. Include no more than two open-text questions, each with a reason it is needed. Add skip logic only where a prior answer makes a later question irrelevant.

Output: objective and decision map; respondent-facing survey; question-by-question rationale and analysis plan; bias and privacy risks; pilot-test checklist. Do not claim that a survey result proves cause, represents a population, or meets a benchmark without a sampling basis. Before answering, check every question against the one-concept rule and verify that each response option set is mutually exclusive and collectively useful. Ask up to 3 clarifying questions only if a required input is missing.

Design a confidential employee pulse survey about [TOPIC] for [EMPLOYEE GROUP]. The leadership decision it should inform is [DECISION]. Create 8–12 questions that separate experience, clarity, workload, and support where relevant. Use neutral wording, balanced scales, optional demographic questions only when sample size makes them safe to report, and an anonymous comment question. Explain how to protect confidentiality and report small groups. Do not promise anonymity if [DISTRIBUTION METHOD] can identify respondents. Ask up to 3 questions only if needed.

Create a customer discovery survey for [TARGET CUSTOMER] about [PROBLEM AREA]. Known context: [EXISTING EVIDENCE]. Write questions that uncover recent behavior, current workaround, trigger events, consequences, and willingness to change, without describing my proposed solution. Include screening and a consent-based invitation for follow-up interviews: [FOLLOW-UP PROCESS]. Output 10–15 questions, skip logic, and what each answer would help decide. Do not ask customers to predict hypothetical future behavior as the primary evidence. Ask up to 3 questions only if required.

Create an analysis plan for this survey: [SURVEY QUESTIONS]. My decision is [DECISION], target population is [TARGET POPULATION], and collected sample is [SAMPLE DETAILS]. Specify data cleaning rules, how to handle missing answers, key cuts or segments, open-text coding approach, and charts suited to each question. Flag where response bias, small samples, or nonrepresentative sampling limit conclusions. Do not invent findings or claim statistical significance without the data and appropriate method. Ask up to 3 questions only if essential.