Event time and arrival time are different
An item may be completed at 11:50 p.m. on one day and reach the reporting system at 12:10 a.m. the next. An event-time report assigns it to the first day. An arrival-time report assigns it to the second. Both can be useful, but they answer different questions.
Use event time to describe when operational activity occurred, if the source records it reliably. Use arrival time when examining ingestion or processing activity. Retain both where available so a delay in reporting does not automatically look like a delay in the work.
Make boundaries unambiguous
A reporting window needs a start, an end, and a time zone. A practical rule is to include the start and exclude the next boundary: from Monday at 00:00 up to, but not including, the following Monday at 00:00. This avoids counting a boundary event twice.
Store and exchange timestamps with a clear offset or time-zone convention. Convert them deliberately for the operational calendar. A daylight-saving transition can produce a local day that is not exactly 24 hours, so avoid assuming that every local boundary is a fixed-duration interval.
Distinguish activity from a snapshot
“Items received this week” is activity over an interval. “Items open at Friday close” is a state at a point in time. Summing daily open-item snapshots does not produce a count of unique open items for the week; it produces an accumulation of daily states.
Label snapshot charts with the observation time. If historical state must be reconstructed, establish whether the event history is complete enough to do so. A current status field alone may not reveal what the status was last month.
Decide how late updates affect history
Late records can revise a previously published period. Choose whether a report remains provisional for a stated interval, is recalculated whenever new records arrive, or is frozen under a documented rule. No single approach suits every operational decision.
Show freshness and revision status separately. “Updated today” does not mean every historical event has arrived. If a comparison uses provisional data, make that limitation visible beside the chart so the latest period is not mistaken for a complete one.
A practical review checklist
For each time-based measure, write down the event timestamp, reporting zone, boundary rule, source delay behavior, and historical revision policy. Test records just before and after a boundary, a late-arriving record, and a record with a missing timestamp.
These examples describe reporting design choices. The appropriate rule depends on actual source behavior and operational needs, which should be confirmed before a time-based measure is used for a consequential decision.