Metric Anomaly Engine reference profile.
Is this business metric's movement a real anomaly that requires investigation, and how urgent is it?
Overview
Surfaces revenue, growth, and reliability problems before they show up in a quarterly review.
Who owns this decision
VP Data, Head of Analytics
What this Specialist Model actually does
Decision: Is this business metric's movement a real anomaly that requires investigation, and how urgent is it?
The final taxonomy records 6 supporting or external components, but it doesn't promote an Oumi-owned specialist model for this engine.
Who uses this Specialist Model
- VP Data
- Head of Analytics
- Director of Business Operations
- VP Product
Where the specialist model sits
The final taxonomy records 6 supporting or external components, but it doesn't promote an Oumi-owned specialist model for this engine.
Time-series statistical baseline (seasonality/trend decomposition)
Supporting component
How Oumi connects to your stack →
Correlated-metric retrieval
Supporting component
Correlated-metric retrieval for root-cause context
Deploy/config-change event feed integration
Supporting component
How Oumi connects to your stack →
Human reviewer confirmation queue before an alert is finalized
Supporting component
How Oumi connects to your stack →
Suppression rules
Supporting component
Suppression rules for confirmed recurring causes
How Oumi connects to your stack →
Specialist classifier
Supporting component
Specialist classifier for anomaly vs. noise
How to evaluate and improve it
Evaluation
- False-positive rate (alert fatigue proxy)
- Time-to-detection vs. manual discovery
- Severity ranking agreement with analysts
Feedback signals
- Analyst marks anomaly as real or false positive
- Root cause confirmed and logged
- Alert dismissed without investigation
- Repeat occurrence of same root cause
- Time-to-acknowledge and time-to-resolve
- Downstream metric correction after fix


