CONVERSION

Exit survey question design

Use when you want to ask leaving visitors why, without leading them into an answer.

exit-survey-question-design.md

Download .md

You are designing a short on-site survey for visitors who are about to leave without converting.

PAGE OR STEP WHERE IT TRIGGERS: {{TRIGGER_LOCATION}}
WHAT I WANT TO LEARN: {{RESEARCH_QUESTION}}
WHAT I ALREADY SUSPECT: {{CURRENT_HYPOTHESIS}}
CONSTRAINT ON LENGTH: {{MAX_QUESTIONS}}

Produce:
1. The question set, in order, within the length limit. For each: the exact wording, the type (open, single choice, scale), and the answer options if closed.
2. For every closed question, the option list must include an escape option such as "none of these" and must not omit an answer that would contradict my current hypothesis. Say explicitly which option exists to allow that contradiction.
3. A note on each question saying what decision the answer would change. If an answer would not change any decision, cut the question and say so.
4. Trigger and targeting rules: when it fires, who is excluded, frequency cap.
5. How you would analyze the open responses, including the minimum number of responses before drawing conclusions.

Constraints: no leading questions, no double barrelled questions, no questions asking people to predict their own future behavior or to explain why they did something they may not consciously know. Avoid asking for a rating without a follow-up asking for the reason. No em dashes. If my research question is too broad to answer in this format, say so and narrow it.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{TRIGGER_LOCATION}}
{{RESEARCH_QUESTION}}
{{CURRENT_HYPOTHESIS}}
{{MAX_QUESTIONS}}

Getting a better result

The "what decision would this change" test usually cuts the survey in half.
Do not ask people what would make them buy; ask what they were looking for and did not find.
Set the minimum response count before you start reading answers.

Questions about this prompt

When do I ask leavers rather than watch more recordings?

When you need the reason rather than the behaviour. Recordings show what happened; a survey can ask what somebody was looking for and did not find. Bring a specific research question, because the prompt will tell you plainly when yours is too broad to answer in this format and narrow it before writing anything.

What do I need in front of me before running it?

The trigger point, what you want to learn, what you already suspect, and a hard limit on the number of questions. Supplying your current hypothesis matters more than it looks, because the option lists are built so at least one answer can contradict it, and the prompt names which option that is.

What comes back, and which part is worth acting on?

The questions in order with types and answer options, the escape option that allows your hypothesis to be contradicted, what decision each answer would change, trigger and frequency rules, and how to analyse open text. The what decision would this change test is the useful part and it usually halves the survey.

What is the mistake that costs me here?

Asking what would make people buy. Nobody predicts their own behaviour reliably, so the prompt bans that shape of question along with rating scales that have no follow up asking why. Set the minimum response count before you start reading, otherwise the first six replies quietly become the finding you act on.