Why it matters
Latency matters because teams need a precise shared meaning for the delay between starting a request and observing a response milestone. Vague language turns incidents into arguments about words instead of fixes.
When everyone uses the same definition, alerts, status updates, and post-incident reviews stay aligned.
How it works
In practice, the delay between starting a request and observing a response milestone shows up as a concrete signal you can measure or communicate. Operators define what good looks like, watch for deviations, and record what happened when expectations break.
The useful version of latency is operational: it changes who gets notified, what customers see, or which metric a team reviews after an incident.
Practical example
Imagine a team operating around p95 latency of 240ms for /v1/search. When observed behavior stops matching the definition of latency, the team treats that change as a reliability event with a clear owner and next step.
Common misconception
Latency is the same as downtime
That reading usually collapses distinct ideas into one slogan. Keep latency tied to observable behavior so the definition stays useful under pressure.
How Fajita handles this
Watch latency with response-time thresholds so slowdowns page someone before full failure.