Find the Story Behind the Anomaly
Let ChatGPT investigate outliers, correlations and trend breaks in your data automatically
You can make a chart but the real questions come after: why did February spike, which products move together, what changed in Q3. Scrolling through rows or eyeballing a line graph wastes time and misses patterns you cannot see.
That is the gap this drop closes. It is a data & analysis skill built for ChatGPT, and it takes about ten minutes to set up the first time. After that it runs in under a minute.
Who should use it
Managers who already use ChatGPT for basic charts but need to spot what the numbers are hiding
How it works
The skill file does five things, in this order.
1. Upload your file and request statistical profiling. Ask ChatGPT to identify outliers, distribution shape and missing values before you pose any business question
2. Request correlation and cohort analysis. Tell it to flag which variables move together and which customer or product segments behave differently
3. Ask for change-point detection. Have it mark the dates or periods where trends reversed, volatility jumped or performance shifted
4. Get a ranked list of anomalies with context. Request a table showing the most extreme values, how far they deviate and what else happened that same period
5. Ask for hypothesis generation and next steps. Tell ChatGPT to suggest three possible explanations for each pattern and what additional data would confirm or rule them out
What comes back
A short investigation report listing the three biggest anomalies, likely causes and which datasets to pull next
The mistake to avoid
Always ask ChatGPT to separate statistical anomalies from business anomalies. A value can be an outlier mathematically but perfectly normal for your industry during holidays or fiscal year-end. Give it context about your calendar, promotions and known events so it flags only the surprises that matter.
Running it
Paste the prompt into the Instructions field of a new custom GPT, or use it directly in a chat. If the task involves files, turn on the code interpreter so calculations are executed rather than estimated.
Where this fits
On its own, one skill saves an hour a week. The compounding happens when three or four of them run in sequence on the same input, the same transcript that produces a scope of work also produces the follow-up email and the project brief. That is the point at which it stops being a prompt and starts being an internal tool. If you want that wired into the systems your team already uses, that is the work 67 Digital does.
In the file
I have uploaded a dataset. Before I ask business questions, I need you to conduct a statistical investigation.
First, profile each numeric column: identify outliers using the IQR method, note the distribution shape, flag missing values and calculate the coefficient of variation.
Second, run a correlation matrix for all numeric variables and flag any pairs with correlation above 0.7 or below -0.7. Explain what those relationships might mean in business terms.
Third, perform change-point detection on {time_column} for {metric_column}. Identify dates where the mean, variance or trend shifted significantly.
Fourth, list the top five anomalies ranked by statistical distance from normal. For each, show the value, the date or identifier, how many standard deviations it sits from the mean, and any other variables that were unusual in the same period.
Fifth, generate three hypotheses for each of the top three anomalies. For every hypothesis, specify what additional data or breakdown would help confirm or reject it.
Present your findings in a report with sections: Profile Summary, Correlations, Change Points, Anomaly Ranking, Hypotheses and Recommended Next Steps.
Context about my business: {business_context, seasonality, known events, fiscal calendar}.
Do not show code unless I ask. Focus on interpretation and actionable insight.