Putting the brakes on dubious claims for Schadegarant

To reduce costs and improve processing speed for Schadegarant, we automated the detection of suspicious insurance claims by building a cloud-based end-to-end anomaly detection system.

The Xomnia X-Force team has shown its broad and deep knowledge and skills; designing and implementing a data-driven application. With a lot at stake, Xomnia showed to be flexible and trustworthy delivering within time and budget.
Frank van Donk, director Schadegarant


Schadegarant is a cooperation between 15 labels of 8 car insurance companies in the Netherlands. They negotiate on behalf of those companies, with car repair shops to get better pricing. It’s important for them to be able to spot anomalies in the pricing, or mistakes so that they can make sure that a claim is genuine.

The challenge of finding anomalies in car damage claims, and sending experts out to ambiguous cases is time-consuming and expensive for Schadegarant. Xomnia’s job was to help them automate the process of finding the anomalies in these claims, instead of having to go through them manually.


Xomnia designed a set of algorithms and filters which can assess dossiers of car damages and highlight questionable ones. Historical dossiers are analyzed by advanced algorithms after training them on a subset of the historical data. The accuracy of the estimation can be assessed by comparing it with the actual repair costs of several historical dossiers that have been approved by the insurance companies’ experts.

On average, the estimation deviates from the actual costs, both in high- and low-cost ranges. Now, new/unseen dossiers can be passed to the algorithm and an estimation of the repair costs can be retrieved. If the estimation of a new dossier deviates significantly from the average deviation, the dossier could qualify for being checked by an expert.

The algorithms run real-time in a highly customized infrastructure that is built using i.a. Google Cloud Platform and Kubernetes technologies. The main application is built with Python and Java. For the algorithms, the python machine learning library Scikit-learn is used.


Schadegarant is now able to act in a more data-driven way, relying on algorithms’ assessments regarding deployment of expensive expertise. This works in real-time, which has meant that there’s no delay in anomaly detection and they can save time and money that used to be spent searching for these anomalies manually. Repair shops can start repairing a car straight away without having to wait for approval from Schadegarant so the whole process is quicker.