Examining a dataset to understand its structure, content, completeness, and potential quality issues before using it.
Data profiling is the first step in any serious data quality effort. It involves analyzing each column to determine its data type, range of values, frequency distribution, null rate, and uniqueness. The result is a summary that tells you what the data actually contains — which may differ from what the documentation claims.
Example: Before integrating a new dataset of business licences, an analyst profiles it and discovers that 12% of postal code values are null, 3% contain invalid formats, and one column described as "integer" actually contains text values. This profile guides the cleaning work needed before the data can be used.
Profiling is not just a technical step — it builds understanding of the data's limitations and helps set realistic expectations for downstream analysis.