Check my data for bias before I draw conclusions from it
Checks who is missing from your data, how it was measured, and how it is shown, so your conclusion doesn't mislead.
The prompt
Check my data for bias and for ways it could lead me or my audience to a misleading conclusion.
Data and how it was collected: [PASTE or describe: source, who is included, period, sampling method, response rate]
Conclusion I'm likely to draw or have been asked to show: [e.g. "customers love the new feature"]
How it will be presented: [e.g. "bar chart in a monthly deck", "survey summary email"]
Audience: [WHO will see it, and what they will decide]
Check for:
- Selection and sampling bias: who is in the data and who is missing (e.g. only active customers, only survey responders, only closed deals, only one region or channel).
- Survivorship bias: failures, churned customers, rejected applicants, or cancelled projects dropped from the data.
- Response and measurement bias: leading survey questions, self-reported or manager-entered numbers, incentives to over- or under-report, and changes in how things were measured.
- Proxy bias: a metric standing in for something it doesn't really capture (e.g. logins as "engagement", tenure as "performance").
- Small bases: big percentages from tiny groups, or groups too small to compare.
- Presentation traps: truncated axes, cumulative charts that only go up, percent vs. percentage points, missing denominators, averages shown without the spread.
- Fairness: groups of people who could be disadvantaged by decisions based on this data, especially if they are under-represented in it.
Return:
1. A table: Risk | Evidence in my data | How it could mislead | Severity (high / medium / low) | How to check or correct.
2. The single biggest risk to my conclusion.
3. A safer way to phrase the conclusion, and a fairer way to present the chart or number.
4. Any extra data or breakdown that would settle the question.
Separate what you can see in the data from general risks I should rule out, and don't invent figures. If decisions will affect people (hiring, pay, pricing, lending, performance ratings), recommend a review by HR, legal, or the relevant specialist before acting.
Highlighted [brackets] are blanks to replace with your own details.
Fill in the blanks and use this promptMore Data & Analysis prompts
What to fill in
- e.g. "customers love the new feature"
- e.g. "bar chart in a monthly deck", "survey summary email"
- WHO will see it, and what they will decide
How this prompt is built
This prompt follows the Few-Shot structure (Few-shot prompting). A few-shot prompt includes one or more worked examples of the input and the output you want, so the model can copy the pattern instead of guessing it.
Related prompts
- Design an A/B testHypothesis, variants, metrics, sample size, and duration for a valid test.
- Break down my results into the business drivers behind themBuilds a driver tree from your results, sizes how much each driver contributed to the change, and names which ones to act on.
- Identify the leading and lagging indicators for my business or teamSorts your metrics into leading and lagging indicators, links each lead to the outcome it predicts, and fills gaps in early-warning measures.
- Find the 20% of activities driving 80% of my resultsRuns a Pareto analysis to show the few activities, customers, or products producing most of your results, and what to cut or double down on.