A supplier presents an on-time fulfillment percentage with a neat chart. Before discussing whether the percentage is good enough, ask for the count of orders that could not be classified.
Missing timestamps are not automatically late shipments. They are also not evidence of on-time performance. A useful review preserves that distinction.
Establish what the report measures
Shopify's order analytics documentation distinguishes fulfilled orders from delivered orders and defines its time-to-fulfill metric as a median. These are different measures; a median duration is not an on-time compliance percentage. Shopify order analytics definitions.
For a supplier's custom report, request the start event, end event, deadline, timezone and exclusion rules. The contractual measure might start after an authorized hold clears, for example. Record that definition before interpreting the data.
A 100-order example with ten unknown outcomes
We constructed a synthetic cohort of 100 eligible orders to demonstrate the arithmetic. It is not a warehouse audit or a sample of any provider's results.
- 84 have evidence of completion within the defined deadline.
- Six have evidence of completion after that deadline.
- Ten lack enough evidence to classify.
Among the 90 classified orders, 84 divided by 90 gives 93.3% on time. Across all 100 eligible orders, only 84% are confirmed on time.
If every unknown later proves late, the final cohort rate is 84%. If all ten prove on time, it is 94%. The interval is an arithmetic bound under these assumptions, not a statistical confidence interval or a prediction.
A report showing only 93.3% hides the unresolved portion. A report showing 84% as the definitive final performance would also overstate what is known. Present both the confirmed result and the unknown count.
Ask for the missing-record explanation
Have the provider identify why evidence is unavailable. Are the ten records concentrated in one carrier, warehouse, shift, service or integration period? A common cause may be more informative than the headline percentage.
Ask for the original timestamp or supporting operational record where available. Preserve the initial extract and the explanation for any later correction. An unexplained replacement spreadsheet makes it difficult to distinguish recovered evidence from changed definitions.
Choose records before seeing the outcomes
For your own review, define an eligible period and order population first. Request the complete cohort where feasible. If sampling is necessary, document the selection rule and avoid accepting only the supplier's best examples. This article does not prescribe a statistically representative sample size.
Include exceptions under the agreed rules, and report their counts. Do not remove troublesome records merely because they make comparison harder.
Use a decision that acknowledges uncertainty
Our proposed review sheet has four fields: eligible orders, confirmed on time, confirmed late, and unresolved. Attach the metric definition and evidence source. If the unresolved share could change your procurement decision, request further evidence before relying on the score.
A provider need not have perfect data to be worth considering. It should be able to show where the data stops and how it investigates the gap. That is a more useful due-diligence conversation than accepting a polished percentage at face value.
Related reading
Sources
Accessed September 29, 2026. Recheck platform and carrier terms before publication.