What to give
- Research question: [e.g. why do customers cancel in the first month]
- Who will take part and how they are recruited: [e.g. an email to all active users with a Rs 200 voucher]
- Method and the full question list or task script: [paste]
- Who collects the data: [e.g. the product manager who built the feature]
- How results will be analysed: [e.g. percentages by plan type]
Check each of these
- Selection: does the recruitment reach only the easy, happy or loyal people.
- Survivorship: are the people who left, failed or never started missing.
- Leading wording: do any questions suggest the answer they hope for.
- Order effects: does an early question colour the later ones.
- Social desirability: are people asked things they may lie about to look good.
- Interviewer influence: does the person asking have a stake in the answer.
- Incentive pull: could the reward attract people who only want the reward.
- Analysis choices: are groups, cut-offs or exclusions decided before seeing the data.
Output format
| Bias | Risk: high, medium, low or none | Where it shows in the plan | Fix before fieldwork |
Then the three changes that matter most, in order, and a rewritten version of any question flagged as leading.
Guardrails
- Quote the exact line from the plan for every risk you flag.
- Suggest fixes that fit the stated budget and timeline.






