Signals for review only: Flags reflect statistical and rule-based anomalies worth reviewing — never confirmed fraud. CSV import only; customer emails are salted-hashed.

Signal Triage Ledger

Flagged Anomalies (4)

Every flag is a statistical or rule-based signal worth reviewing — filter, bulk-mark decisions, or open any row for AI explanation and linked transactions.

Showing 4 of 18 signals
Severity / ScoreRuleSignal & ExplanationStatusAction
highScore 94/100
Velocity

5 row(s)

Customer refund velocity outlier: r***@example.com · #d917b629 (5 refunds, $800.00)

Customer r***@example.com · #d917b629 accumulated 5 refunds totaling $800.00, exceeding the cohort 3σ threshold of 2.9 refunds (mean 0.39, σ=0.83). Worth reviewing account history and support tickets.

OpenInspect
highScore 94/100
Velocity

7 row(s)

Customer refund velocity outlier: v***@example.com · #af39114a (7 refunds, $579.95)

Customer v***@example.com · #af39114a accumulated 7 refunds totaling $579.95, exceeding the cohort 3σ threshold of 2.9 refunds (mean 0.39, σ=0.83). Worth reviewing account history and support tickets.

OpenInspect
highScore 92/100
Velocity

16 row(s)

Daily refund velocity spike on 2026-11-12 (16 refunds, $1,547.60)

2026-11-12 recorded 16 refunds totaling $1,547.60 compared to a daily baseline of 2.5 refunds (σ=2.8, 3σ threshold=10.9) and $226.83/day. This statistical outlier is worth reviewing for an incident, batch error, or coordinated return wave.

OpenInspect
mediumScore 86/100
Velocity

4 row(s)

Customer refund velocity outlier: r***@example.com · #f1863b43 (4 refunds, $388.50)

Customer r***@example.com · #f1863b43 accumulated 4 refunds totaling $388.50, exceeding the cohort 3σ threshold of 2.9 refunds (mean 0.39, σ=0.83). Worth reviewing account history and support tickets.

OpenInspect
Flagged Signals · RefundRadar