Data
Profile a dataset
Nulls, ranges, duplicates and the columns that are quietly lying.
Fill it in
The file to profile.
Impossible dates, negative counts, placeholder values.
Which questions the data can actually answer.
Your prompt
Profile this dataset. Per column: type, null rate, distinct count, range, and the most common values. The most common value is the one that reveals the placeholder, because a date column where 1900-01-01 appears 4,000 times is not a date column. Flag what is impossible rather than merely unusual: negative quantities, dates in the future, ages over 130, and a country column with 43 spellings of one country. Those are the ones that survive into a report and get presented. Say what the data can actually SUPPORT. A column that is 60% null cannot carry a segmentation, and saying so now is cheaper than discovering it in the analysis.
Use Profile a datasetOpens with everything above already filled in.
Why this works
The most common value in a column is what reveals the placeholder: a date field where 1900-01-01 appears 4,000 times is not a date field. This flags what is impossible rather than merely unusual, and says which questions the data can actually support before anyone builds on it.