Incident model feedback
How to adjust a monitor's alert sensitivity using the feedback mechanism on an incident.
Last updated
How to adjust a monitor's alert sensitivity using the feedback mechanism on an incident.
For monitors that use Smart Training, you can adjust alert sensitivity directly from an incident. Sensitivity controls how wide or narrow the model's confidence interval is — a wider interval means fewer alerts; a narrower interval means more.
Sensitivity is set on a scale of –5 to +5:
–5
0.99
Widest — least sensitive, fewest alerts
0 (default)
0.90
Balanced
+5
0.80
Narrowest — most sensitive, most alerts
Each step on the slider changes the confidence interval by 0.02.
Navigate to Data Quality > Incidents and open the incident you want to provide feedback on.
Click the thumbs up (👍) icon if the alert was accurate. Click the thumbs down (👎) icon if it was a false positive or missed something.
For monitors with Smart Training enabled, clicking thumbs down reveals the sensitivity slider. Adjust the slider and submit — the new sensitivity takes effect on the next scan.
Sensitivity adjustment is available for Freshness, Volume, and Field Health monitors with Smart Training (Auto threshold) enabled. It is not available for Custom SQL monitors or monitors using manual thresholds.
Too many false positives → move the slider left (toward –5) to widen the confidence interval.
Missing anomalies you expect to catch → move the slider right (toward +5) to narrow the interval.
For newly created monitors, wait until at least a few weeks of scan history have accumulated before adjusting sensitivity — early scans may not yet reflect the true variance in your data.
Last updated