Backfill

Backfill is the addition of records or values for an earlier period to a dataset that already exists. In public-safety reporting, it can make a past month's total increase when previously missing records become available. That increase does not represent newly occurring events at the time of the update. The new information becomes available later than the period it describes.

Why older totals can change

A chart has at least two relevant dates: when the activity happened and when the information became available. Backfill connects those dates. A report published in June may add records about April activity, so the April bar changes when the chart is refreshed.

Seattle provides a concrete example of delayed availability: its crime-data documentation says reports awaiting approval do not appear in the public release until a subsequent update. That release process can add previously absent records to an earlier period. This is a local example, not a timetable for other agencies. Seattle Police Department dataset documentation.

Example

Hypothetical: On May 2, a dashboard contains 80 incidents with April occurrence dates. On May 9, the agency supplies five additional April records. The refreshed April count becomes 85.

The five-record increase describes an update to the available April history. It does not mean five incidents occurred on May 9. If the dashboard instead groups by report date, some records may fall into a different month; the date field determines the interpretation.

Common confusion

Backfill is one kind of revision. A correction to an address or a change in an offense category also revises data, but it may add no previously missing record. An estimate is different again: it supplies a calculated value where direct observations are incomplete. Adding actual records and estimating missing activity should be labeled separately.

The term also appears in staffing, where it can refer to filling an uncovered position or shift. Here it refers specifically to historical data.

Reading a changing historical chart

Compare the same period using the same date field and filters. Keep the extraction date with any downloaded figures. If a historical total changes, look for a revision note describing the affected dates and whether records were added, removed, or recoded.

A useful update note might say, “April now includes five reports received after the previous release.” That tells readers more than “data updated.” Without a stated reason, a higher historical total alone cannot establish whether the change came from backfill, corrections, or a different counting method.

Related terms

• Revised data
• Reporting lag
• Data completeness

Sources

• City of Seattle: SPD Crime Data documentation, hosted by Data.gov — Explains approval-related delays and changes during the reporting process; supports the local example of historical records appearing later.