PurposeForthcoming requirements in MiFID and RegNMS mean that buy‐side and sell‐side firms need to find ways of showing regulators that they are sifting through their trading volumes in a justifiable, methodical manner looking for anomalous trades and investigating them, in order to prove “best execution”. The objective was to see if a SVM/DAPR approach could help identify equity trade anomalies for compliance investigation.Design/methodology/approachA major stock exchange, a computer systems supplier, four brokers and a statistical firm undertook a cooperative research project to determine whether automated statistical processing of trade and order information could provide a tighter focus on the most likely trades for best execution compliance investigation.FindingsThe support vector machine approach worked on UK equities and has significant potential for other markets such as foreign exchange, fixed income and commodities.Research limitations/implicationsThe research has implications for risk professionals as a generic approach to trading anomaly detection. The prototype compliance workstation can be trialed.Originality/valueAutomated anomaly detection could transform the role of compliance and risk in financial institutions.