BA010 · Correspondent Banking

Peer Group Anomalies

Most correspondent-banking analytics compare an account against a fixed, absolute threshold — a dollar cut-line, a transaction count, a ratio. Peer group anomaly detection compares an account against the STATISTICAL PROFILE OF ITS OWN PEER SEGMENT instead: accounts of a similar type, size and declared business purpose. A transaction volume that is unremarkable for a high-turnover corporate current account can be a strong deviation signal for a retail deposit account, and vice versa — a fixed threshold misses this, a peer-relative model does not.

The finding lives in how far the account's own transaction volume, turnover velocity, product/channel usage and counterparty breadth sit from its peer segment's mean, measured in standard deviations (a z-score), not in the raw number itself.

Risk indicators

Volume Z-Score

• Transaction volume running multiple standard deviations above the peer segment mean
• Volume growth with no plausible peer-consistent business explanation

Velocity Z-Score

• Turnover velocity (transactions per period) inconsistent with the peer segment's norm
• A rate of fund movement that outpaces the segment's typical pattern

Product Usage Deviation

• A product/channel footprint markedly narrower or wider than the peer segment's norm
• Usage inconsistent with the account's declared segment and business purpose

Counterparty Diversity Z-Score

• An unusually wide or unusually concentrated counterparty footprint against the peer norm
• A single anomalous counterparty accounting for a peer-inconsistent share of activity

Investigation actions

1. Resolve the subject account and its declared peer segment.
2. Compute the volume z-score (own transaction volume vs the peer segment's mean
3. Compute the velocity z-score (own turnover velocity vs the peer segment's mean
4. Compute the product usage deviation z-score (own product/channel count vs the
5. Compute the counterparty diversity z-score (own distinct-counterparty count vs
6. Score each of the four components against its |z| band cut-lines.
7. Aggregate the weighted exposure score; escalate at or above the threshold.

Detection logic

1. Resolve the subject account and its declared peer segment. 2. Compute the volume z-score (own transaction volume vs the peer segment's mean and standard deviation). 3. Compute the velocity z-score (own turnover velocity vs the peer segment's mean and standard deviation). 4. Compute the product usage deviation z-score (own product/channel count vs the peer segment's mean and standard deviation). 5. Compute the counterparty diversity z-score (own distinct-counterparty count vs the peer segment's mean and standard deviation). 6. Score each of the four components against its |z| band cut-lines. 7. Aggregate the weighted exposure score; escalate at or above the threshold.

Related typologies

Curated from fc-06-correspondent-banking-analytics-toolkit/typologies/BA010_Peer_Group_Anomalies.md