Data completeness

Data completeness describes how much of the expected information is present. It can concern whole records, required fields within those records, or coverage of the intended period and area. A completeness claim needs a defined scope: complete for which records, which fields, and which use?

Two different checks

One check asks whether all expected records arrived. Another asks whether the records contain the needed details. The UK Government Data Quality Framework includes both records and essential values in its completeness definition and distinguishes completeness from accuracy. Government Data Quality Framework.

A call dataset might include all submitted incidents but lack arrival times for some responses. It could support a count of those incidents while supporting response-time analysis for only a subset. Calling the whole dataset simply “complete” would obscure that difference.

Example

Hypothetical: A department expects 200 incident records for a reporting period and receives all 200. Of the 160 responses eligible for an arrival-time measure, 144 have a usable arrival timestamp.

Record delivery is complete against the department's expected list. Arrival-time completeness for the eligible responses is 144 ÷ 160 × 100 = 90%. The remaining 40 incidents were outside that measure's defined scope, so including them in its denominator would answer a different question.

Common confusion

A filled field can contain the wrong value. Completeness therefore does not establish accuracy. A placeholder such as “unknown” may satisfy a file's formatting requirements while still leaving the information unavailable for analysis.

Not applicable is different from missing. An event with no transport may have no applicable hospital-arrival time. Whether a field is expected depends on the event and the published definition. Counting all blanks as identical can misstate the problem.

What readers should look for

A useful completeness statement names the expected set and explains how it was established. “All records received from the source were processed” is narrower than “all incidents in the jurisdiction are included.” The first statement does not verify what the source omitted.

For a calculated measure, look for the number eligible, the number included, and the reasons for exclusion. Check whether missing values are concentrated in a particular month, agency, or type of call; a high overall percentage can conceal an important gap.

If the expected number is unknown, an exact completeness percentage cannot be established merely by counting the available rows. The publication should describe that uncertainty directly.

Related terms

• Coverage gap
• Missing data
• Not applicable
• Data quality

Sources

• UK Government: Government Data Quality Framework, completeness dimension — Distinguishes the presence of records and essential fields from their accuracy. The eligibility calculation is hypothetical.