Italy holds an undisputed and peculiar record in the world of car insurance: with more than 5 million of black boxes, it is the country with the major number of telematics clients in the world.
The introduction of black boxes has completely changed the world of insurance by introducing new services such as: discounts on insurance premium based on driving style, real-time assistance in case of crash, positioning of cars in case of theft, danger detection on a specific street, and control over speed limit.
A whole new architectural design was introduced in latest few years to provide real-time assistance in case of crash and geo-locationing services. Data from black boxes are collected in real time with Kafka, processed, enriched, and, according to necessity, used to provide assistance and stored in order to be usable for analytic analysis.
The enrichment process consists of the addition of information regarding personal data of the customers. Information about accidents is processed in real time to detect false-positive events, using a machine learning algorithm that has proved to have high accuracy allowing to prevent fraud against the insurance company. With the information acquired about trips, we can determine a coefficient related to the driver’s ability in order to offer rates tailored for each customer and even report potential danger on a specific street.
In this presentation, we will focus specifically on the data management challenges that occur in a streaming context where the amount of data to process is really huge and the architecture is based on Kafka. We provide criteria for selecting a data management model capable of providing the scalability and performance needed to support massive growth.