Data profiling

Data profiling is an X-ray of a dataset. For each field it measures how many values are missing, which formats appear, how many distinct values there are, minimums and maximums, duplicates and odd cases.

It is done before three key moments: when defining quality rules (so they are realistic), when migrating or integrating systems (so errors are not carried over) and when preparing data for AI (to spot gaps or biases before training).

Many tools do it automatically, but you can also start with a spreadsheet. What matters is turning findings into issues with an owner and a deadline, not a report that gets filed away.