1 / 5
The reference distribution a monitor compares live traffic against is the ___.
-
-
-
-
The baseline (often the training or a recent stable window) is what current feature/score distributions are compared to in order to detect drift.
2 / 5
An alert that fires repeatedly for the same underlying issue, training responders to ignore it, causes ___.
-
-
-
-
Alert fatigue happens when noisy or duplicate alerts desensitise the on-call team, risking that real incidents get missed.
3 / 5
Phrasing such as 'PSI on feature X exceeded 0.25 for 3 consecutive windows' describes the alert's ___.
-
-
-
-
A clear trigger condition states the metric, threshold, and duration so everyone understands exactly what caused the alert.
4 / 5
A short document telling the on-call engineer what to check and do when a drift alert fires is a ___.
-
-
-
-
A runbook gives step-by-step triage and remediation guidance, reducing time-to-resolution and reliance on tribal knowledge.
5 / 5
Routing an unresolved high-severity drift alert to a senior engineer or manager is called ___.
-
-
-
-
Escalation moves an alert up the chain when the first responder cannot resolve it within the expected time, ensuring serious drift gets attention.