Straight Fact · desk research and BespokeData methodology · 18 Nov 2026
Define on-time before measuring it
3 min read · Free article
On-time performance sounds like a simple percentage until people disagree about the promise, the event or the eligible population. An order dispatched on schedule can still arrive late. A revised date can make a delayed order appear punctual. Agree the question before extracting the data, and keep the definition beside the result. Otherwise two apparently conflicting figures may both be calculated correctly while describing different parts of the process.
The UK Government quality framework's fit-for-purpose approach is relevant to choosing usable records [Q1]. ASQ's process-mapping guidance provides a starting point for distinguishing events and hand-offs [A1]. Our practical application is to define receipt, dispatch and delivery separately, with the owner and source for each timestamp. A process map helps locate evidence; it does not establish that every recorded event is accurate.
In a hypothetical month, 500 distinct eligible orders have an original promised dispatch date in the period. All have usable timestamps, and 400 were dispatched by that promise. On-time dispatch is 400 divided by 500, or 80%. Now suppose 20 of the 100 late orders had their promises revised and would qualify as on time under the revised dates. That alternative definition gives 420 divided by 500, or 84%. The four-percentage-point difference is a definition change, not an improvement in operations.
Specify the reporting timezone, whether comparisons use calendar or working time, and how cancellations and partial shipments are handled. Say whether eligibility follows promises due in the month or orders received in it. Report missing timestamps as unclassifiable, not automatically on time or silently excluded. Preserve revision history where original customer commitments matter. These choices should be owned by the business, not inferred differently by each report builder.
A practical metric register records name, decision, numerator, denominator, grain, period, timezone, exclusions, source and owner. Give the rule a version and change date. If the definition changes, recalculate previous periods where possible or mark a break in comparability. Ask another authorised reviewer to reproduce one result from the same records. That check often reveals joins or duplicate shipments that a polished percentage conceals.
The example is hypothetical and says nothing about an actual company's punctuality. Reliable definitions do not establish causality or guarantee improvement. They make the next investigation more credible. Use the full operational assessment guide to connect a checked baseline with process hypotheses, a bounded pilot and named action owners instead of treating a monthly percentage as a complete diagnosis.
Dated method & source register
Read and reviewed 18 Nov 2026. Desk research and original BespokeData recommendations and hypothetical examples; no primary study, independent expert review or Atora generation.
- [Q1] UK Government · Data Quality Framework
Fit-for-purpose quality dimensions; public-sector context.
- [A1] ASQ · Flowchart
Process mapping reference; no claim that a map establishes causality.