Verified against ChatGPT · 2026-08-12
Write a Jobs-to-be-Done interview script that surfaces the real switching trigger without leading the witness
Writes a JTBD-style user interview script focused on the specific moment someone decided to switch tools, phrased to avoid leading questions that would just confirm what you already assume.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Write a Jobs-to-be-Done interview script for talking to people who recently switched to (or away from) the product below. The point of a JTBD interview is to find the real trigger moment and struggle that led to the switch — not to confirm assumptions I already have, so every question needs to be checked for leading language before it goes in the script.
PRODUCT AND SWITCH DIRECTION
Interviewing customers who switched away from a spreadsheet-based system to our inventory-management app.
MY CURRENT ASSUMPTION ABOUT WHY THEY SWITCHED
We assume they switched because spreadsheets couldn't handle multi-location stock syncing.
INTERVIEW LENGTH
30 minutes.
STRUCTURE THE SCRIPT IN THIS ORDER
1. A timeline reconstruction opener that asks the person to walk through the actual sequence of events leading up to the switch, starting well before the switch itself — a JTBD interview should establish what was happening in their life or workflow before they even started looking for an alternative, not open with "why did you switch," which invites a tidy retrospective justification rather than the messier real sequence.
2. Struggle-moment questions that probe for the specific moment something became actively painful enough to prompt looking for alternatives — ask what they were doing right before that moment, not what they generally disliked about the old tool in the abstract.
3. A forces-of-progress set covering what pushed them away from the old solution, what pulled them toward the new one, what anxieties they had about switching, and what habits of the old tool they had to overcome — these are four genuinely different forces and deserve separate questions, not one blended "why did you switch" question.
4. A check against my stated assumption — include one question late in the script, phrased neutrally, that would surface disconfirming evidence if my assumption is wrong, rather than a question that could only confirm it.
RULES AGAINST LEADING LANGUAGE
For every question, check whether it presupposes an answer ("wasn't it frustrating when...") or offers the interviewee a category to agree with ("was it mainly the price, or the features?") rather than asking them to generate the category themselves. Rewrite any question that fails this check before including it.
WHAT NOT TO DO
Do not include generic satisfaction-survey questions (rate your experience 1-10) — JTBD interviews are about reconstructing a causal story, not scoring sentiment. Fit the total question count to the stated interview length; do not write more questions than could actually fit.
OUTPUT FORMAT
The script in the four sections above, each question numbered, with a one-line note under any question that was rewritten explaining what leading language was removed.Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A JTBD interview script written without an explicit leading-language check tends to smuggle the interviewer's own hypothesis into the question wording — GPT-5.1 given only "write a JTBD script about why customers switched" will often produce questions shaped by whatever context clues it's given about the product, phrasing options as a menu ("was it price or features") that limits the interviewee to categories the researcher already had in mind, which defeats the entire premise of JTBD interviewing: finding the causal story the researcher didn't already assume. Building the script around timeline reconstruction rather than opening with "why did you switch" matters mechanically because a direct why-question invites a post-hoc, socially acceptable justification ("it had better features") rather than the actual messy sequence of events, while reconstructing the timeline forward from before the person even considered switching surfaces the real struggle moment, which is frequently different from the tidy reason a person would give if asked directly. Splitting the four forces of progress (push, pull, anxiety, habit) into separate questions rather than one blended question matters because these are genuinely distinct psychological forces in switching behavior, and a blended question lets an interviewee answer with whichever force is easiest to articulate, silently dropping the other three from the data entirely. The disconfirming-evidence question specifically counters confirmation bias in the researcher, not just the interviewee — a script built entirely from someone's existing hypothesis, however well-intentioned, will structurally tend to surface answers that confirm that hypothesis unless at least one question is deliberately built to have a real chance of contradicting it.
What you get back
Section 1, Q2: 'Take me back to a normal week before you started looking at other options — what did tracking inventory across locations actually look like day to day?' (Rewritten: original draft asked 'wasn't it hard to keep the spreadsheet updated across locations,' which presupposed the difficulty; the neutral version lets the interviewee describe the actual workflow without confirming a specific pain point in advance.)
Verified against
ChatGPT GPT-5.1 · 2026-08-12
Changelog
- 2026-08-12 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — branding & design, built and maintained for you — that's Scult's day job.
EXPLORE BRANDING & DESIGN
