Estudios de Mercado I · Tema 5 · Question wording, biases & pretesting (5.3 · 5.5 · 5.7)
Question checker
Estudios de Mercado I · Tema 5 — Question checker worksheet
A leading question asked of 10,000 people is still leading. Train your eye on faulty questions first, then scan your own draft, tick the launch checklist and run an AI pretest. The scanner flags what may be wrong — you decide whether it is.
What this practical is for
To train your eye to spot faulty questions (5.3) and response biases (5.5), then pretest your own questionnaire before fieldwork (5.6–5.7) — because a bigger sample never fixes a bad question.
What you will do
Step 1 — practise on 15 faulty questions. Aim for 12 or more, and read the rewrite for each one.10 min
Step 2 — came from the builder? Your questionnaire, project topic and analysis plan are already here. If not, paste your questionnaire and write the topic. For every flag, decide: real fault or false alarm?15 min
Step 3 — tick the launch checklist honestly. Every unticked box is a task still to do.5 min
Step 4 — add your analysis plan and run the AI pretest (three passes). Compare it with your own diagnosis.10 min
Step 5 — pilot with 3–5 real people, time it on a phone, and record every change with its reason. Download the PDF.20 min + pilot
What you hand in
The PDF: the flags you accepted or rejected, with reasons.
The checklist.
Your changes, final number of questions and timed length.
How it is marked
Rejecting a false alarm with a good reason counts as much as fixing a real fault.
The faults you fixed that the scanner missed.
Evidence of the pilot — what real people stumbled on.
Before → the questionnaire builder. After → pilot it, fix it, and launch it on your platform.
Working over several days? Everything saves itself in this browser. On a shared computer, or to switch computer, use 💾 Save my work at the bottom of the page and, next time, 📂 Open saved work.
1 Practise — what is wrong with this question?
Pick the fault. You get the reason and a clean rewrite straight away. The test for every question: could two reasonable people read it differently? and can someone answer honestly without being nudged?
2 Scan your own draft
Paste the questionnaire: one question per line, with its options on the lines below starting with - (or on the same line after a colon, separated by /). The builder's “Part B” output is already in this format.
3 The launch checklist (5.6)
Do not send it until every box is ticked. These are the checks no scanner can make for you.
4 AI pretest — three passes (5.7)
The prompt below already includes your draft and the scanner's flags. Add your analysis plan so the AI can check scale ↔ analysis fit too. Paste it into a new chat.
Then do what no model can: pilot it on 3–5 real people like your respondents and watch where they hesitate. A hesitation is a question to rewrite. AI use must be declared in your report.
5 What you changed
You are the methodologist. Which flags were real, which you rejected (and why), what the AI and the pilot changed, and what you cut.
Saved in this browser
To continue another day: your work saves itself in this browser on this computer. On a shared or university computer, or to switch computer, press 💾 Save my work — it downloads a small file. Next time, open this page and press 📂 Open saved work to pick up exactly where you left off. The PDF is what you hand in; the saved file is only for you.