Graph Analysis

Modelling the network of shared entities between records to expose coordinated activity that looks unremarkable case by case.

Uses AIestablishedgraph

Graph analysis changes the unit of review from the individual record to the network around it.

It is the technique to reach for when the concern is coordination between actors rather than any single actor’s behaviour.

It depends entirely on entity resolution being correct across identifiers such as people, addresses, and account numbers, so weak identity matching undermines everything built on top of it.

Skip it when records are genuinely independent of one another, since building and maintaining a graph is not worth the cost.

General strengths

• Detects coordinated or organised activity that individual case review cannot see.
• Shared entities give a reviewer a concrete, explainable lead to follow.
• Value grows as more records get connected into the graph.

General limitations

• Entity resolution errors create false links and wasted review effort.
• Computationally heavier than scoring records independently.
• A shared entity is often innocent, such as a large, legitimately busy counterparty.

Applied to business problems

Each of these is written for its own problem. The technique is shared, the solution is not.

• Network layering detectionGraph analysis linking accounts through shared counterparties and rapid pass-through transfers to expose layering chains and money mule networks.Transaction Monitoring
• Organised ring detectionGraph analysis linking claims through shared people, vehicles, addresses, and providers to expose coordinated fraud rings.Claims Fraud Detection