ANOMALY AGENT — Q&A
Q: How does the Revenue Anomaly Agent detect anomalies? Uses a statistical baseline approach (Z-score). For each metric (MRR, invoice count, payment amount, credit note volume), it maintains a rolling average and standard deviation over the last 12 periods. Any observation more than 2 standard deviations from the mean is flagged. Examples of what it catches:
- MRR spike from a duplicate subscription being activated
- Unusual credit note volume suggesting billing errors
- Payment receipts far higher than invoiced (over-payment from customer)
- Journal entries that are outliers vs historical amounts
Q: What is the Z-score rolling baseline? Level 1 enhancement. The Anomaly Agent now uses a rolling 12-period Z-score rather than a fixed threshold. Threshold automatically adjusts as the business grows. A company at ₹5Cr ARR and a company at ₹50Cr ARR have very different normal ranges — the Z-score baseline adapts to each.