A successful API response may contain yesterday’s weather
A monitoring dashboard displays a temperature value with no visible error. The measurement is several hours old because a sensor gateway stopped sending updates. Freshness monitoring makes the age of observations visible and actionable.
Define expected update behaviour
Different instruments and datasets have different reporting intervals and acceptable delays.
Separate event time and delivery time
A reading may be measured at one moment and published later. Preserve both concepts where relevant.
Monitor stale values
Repeated numbers, old timestamps and missing periods can indicate faults, though stable real conditions are also possible.
Show quality to consumers
APIs should provide appropriate timestamps, status information and quality flags.
Avoid misleading fallbacks
Cached data can be useful if clearly labelled, but should not silently appear live.
Assign recovery ownership
Freshness alerts need investigation, repair and verification of resumed trustworthy data.
Break-test: a gateway reconnects and uploads old observations
Can the API preserve their original times rather than displaying them as new measurements?
Connected knowledge
See API Post Release Monitoring, Sensor Failure Detection and NEA.