Contact

GlossaryPublic records, privacy, and data quality

Data provenance

Data provenance records where information came from and how it changed. Learn how it helps trace a public-safety statistic back to its source.

Data provenance is the documented origin and processing history of information. It explains where a record or dataset came from, which versions were used, and what happened before the information reached its current form. For a published statistic, provenance connects the final number with the source records and the steps used to produce it.

Following the path to the result

The World Wide Web Consortium's provenance model describes the things, activities, and responsible people or organizations involved in producing information. The central idea is traceability: a reader can examine how an output was derived and who was responsible for its inputs and processing. W3C PROV Data Model.

A public-safety chart might begin with an agency download, select a date range, remove duplicate imports, group local codes, and calculate a rate using a separate population source. Each step can affect the final result and belongs in its documented history.

Example

Hypothetical: A dashboard reports 75 incidents for a month. Its notes identify an agency file downloaded on June 5, a filter for May occurrence dates, and a rule that counts distinct incident identifiers.

Another reader counts 82 rows in the current file. The provenance notes provide a way to investigate the difference: the file may have changed, rows may represent multiple offenses, or the filters may differ. The notes do not decide which explanation is correct, but they make the comparison possible.

Common confusion

Attribution names a source. Provenance goes further by describing the path from that source to the published result. A link to an agency homepage does not document the exact file, version, filters, and calculations used.

Provenance also differs from a guarantee of accuracy. A well-documented process can still contain an error. Its value is that the error can be located and the result assessed, rather than leaving the transformation hidden.

What useful provenance includes

For a summary, retain source identifiers, extraction dates, relevant versions, transformation rules, and the source of any denominator. When data is corrected, record what changed and why. When a derived category differs from the agency's original category, identify it as a later analytical grouping.

This documentation need not expose protected information. A public method can explain the source and calculation without publishing personal records. The aim is to give readers enough information to understand the result's origin and limits, and to reproduce it where the underlying data is available for that purpose.

Sources