The idea
Streaming results are often published before every event has arrived. A late event can therefore change an already visible total. The pipeline needs a policy for corrections, not just a timer that closes a processing window.
Design choice
Define the tolerated delay and whether older results are revised, marked provisional or left unchanged. Distinguish event time from processing time. A finalised business report may need a different correction process from a live dashboard.
Worked example
A daily dashboard marks yesterday's count as provisional for one day. Later corrections remain visible in a revision history instead of silently replacing a published figure.
Before you ship
- Set a correction window
- Mark provisional results clearly
- Track revisions to published totals
Examples are illustrative. Adapt the checks to your own environment.