Case Study: Covéa Insurance saves £350k detecting ghost-broker fraud with Seldon Deploy (Seldon)

How Covéa plan to save £1 million detecting fraudulent insurance policies

Covéa Insurance, the UK underwriting arm of the Covéa group serving two million policyholders, faced widespread "ghost broking" fraud and a deluge of over two million new quotes per day while having limited direct customer data for detection. To scale detection quickly and meet strict regulatory explainability requirements, Covéa evaluated tools and partnered with Seldon, selecting Seldon Deploy for model serving and explainability.

Using Seldon Deploy, Tom Clay’s data science team productionised explainable ML models with a human-on-the-loop workflow; a four-person team deployed models that recommended over 1,000 policies for review and saved £350,000 in the first four months. Covéa is now extending the approach toward real-time quote checks, showing Seldon’s measurable impact on fraud detection and operational efficiency.

Covéa Insurance

Tom Clay

Chief Data Scientist