Most people look at data and describe what they see. This prompt makes AI do something harder — it looks for what's missing, what's inconsistent, and what the surface numbers are designed to obscure. You paste in a dataset, a table, a report, or even a summary, and the prompt runs a structured forensic pass: anomalies, gaps, misleading aggregates, hidden segments, and the questions the data can't answer but should.
It works because it forces the model to interrogate the data rather than narrate it. Instead of getting a summary that confirms what you already think, you get a list of things worth challenging — before you make a decision, write a report, or hand numbers up to a client.
The Prompt
You are a data forensics analyst. Your job is not to summarize this data — it is to interrogate it.
Here is the data:
[PASTE YOUR DATA, TABLE, REPORT, OR SUMMARY HERE]
Run the following analysis and return each section clearly labeled:
1. ANOMALIES
Identify any numbers, ratios, or trends that are statistically unusual, suspiciously round, or inconsistent with the rest of the dataset. Flag them and explain why they stand out.
2. MISSING DATA & GAPS
What data is absent that should logically be present? What time periods, segments, categories, or metrics are unaccounted for? List each gap and explain what decision risk it creates.
3. MISLEADING AGGREGATES
Are any totals, averages, or percentages masking important variation underneath? If you averaged two very different groups together, call it out. Surface any cases where the aggregate hides more than it reveals.
4. HIDDEN SEGMENTS
Are there subgroups or cohorts buried in this data that behave differently from the overall trend? Identify them and explain what would change if they were separated out.
5. CAUSATION TRAPS
What correlations in this data are likely being mistaken for causation? What alternative explanations exist for the patterns shown?
6. QUESTIONS THIS DATA CANNOT ANSWER
List 3–5 specific questions a decision-maker would reasonably ask that this data is not equipped to answer. Be precise — name the missing variable or context.
7. VERDICT
In 2–3 sentences: what is the single biggest risk of making a decision based on this data as-is, and what is the one piece of additional data that would most change the picture?When to use it
Use this any time you are about to make a significant decision from data — or when someone hands you numbers and expects you to just trust them. It is especially useful in these situations:
- You received a client report or marketing analytics dashboard and need to give feedback or recommendations without being misled by cherry-picked metrics.
- You are reviewing your own ad performance data and want to find out why results look flat even though one campaign appears to be working.
- A vendor, agency, or team member sent you a results summary and you want to pressure-test it before signing off or renewing.
- You are building a proposal or pitch and want to audit the market data or case study numbers you plan to cite for weaknesses a skeptical buyer might spot.
- You pulled a sales or lead data export and you want to find the hidden segment — the customer type, traffic source, or funnel step — that is quietly dragging down your averages.