If you want to protect customers, you need data

 

 

The analysis and use of internal and external data is not necessarily new to successful insurance companies. In digitizing products, services and processes, many insurers have the advantage of being able to draw on data collections that have been built up over many years. However, few succeed in analyzing this interconnected data holistically and visualizing it in a 360-degree view of the customer.

Customer-centric services also means fraud detection

This was the challenge facing Allianz Benelux. The subsidiary of the 130-billion-euro insurance giant Allianz, employs more than 2.000 employees, insures customers across countries in Belgium, the Netherlands and Luxembourg, and has an estimated annual turnover of around four billion euros. After a series of mergers and acquisitions, the company faced the task of consolidating its customer data across multiple data silos. The aim was not only to improve sales efficiency, but also to gain a complete overview of customer risks, make forecasts for the future and thus improve customer service.

A central role is also played by fraud detection. This is where Allianz Benelux takes a zero-tolerance stance. The insurer not only wants to put a stop to criminal activities, comply with legal requirements and protect its own company, but also ensure that the premiums paid in actually benefit the policyholders and that premiums do not have to be increased in the long term. If you identify fraudsters early on and spot false claims, you can offer your customers better offers and prices on balance.

Graph technology and graph analytics

A snapshot of customers, including all their data and data relationships, could not be realized with the existing systems of Allianz Benelux. Relational database systems in particular are simply not capable of linking heterogeneous data across multiple data silos. Suspicious behavior patterns, fraudulent networks, and opaque connections can only be mapped visually, if at all, at great computational expense. The tables in SQL databases with rows and columns do not provide deep, contextual data connections and cannot track relationships across multiple levels. Frequently retrieved, "relevant" data (warm data) cannot be extracted.

Allianz Benelux therefore decided to use graph technology. Your advantage: The data model of graphs can be intuitively understood and thus vividly visualized. Data is represented as circles (nodes) connected by lines (edges). This allows even highly networked data to be analyzed and queried in real time. Relationships between customers, same addresses or phone numbers, and past claims and insurance benefits can be considered in context. The query speed does not depend on the total amount of data in the database and the number of linking operations, but only on the number of concrete relationships that are relevant for the desired query.

Graph algorithms are used for fast and precise analysis. The "PageRank" algorithm developed by Google is probably the best known example here. It measures the importance of each node within a graph based on the number of its relationships and the importance of the nodes connected to it. Or simply put: The more a data set is linked to other data, the more important it is. Google orders its search results according to this principle. Transferred to the insurance world, important customers, insurance companies with high claims coverage or even a suspicious accumulation of claims can be identified. Nodes with a high PageRank score appear larger in the graph and attract the attention of customer advisors, risk analysts and investigators. In combination with data warehouses, graph databases and graph algorithms thus form the foundation of modern data analysis in companies.

Fraud detection and 360-degree view

After an extensive market evaluation, Allianz Benelux chose Neo4j as its central database. In addition to the high scalability and flexibility, the enterprise features as well as the dominance of the enterprise solution for graphs on the market were decisive factors. In the Neo4j graph, data scientists, analysts and consultants can now look at customers from different points of view such as place of residence, address and the other people living there. This gives them an accurate picture of their relationships, lives and potential needs very quickly. Offers can be personalized and existing services optimized, with customer advisors being able to draw on automated recommendations from the system. As the customer view goes beyond the usual parameters such as personal advisor and claims, Allianz experts can see at a glance how many insurance policies a person has taken out and whether a new contract with combined benefits or extended coverage is worthwhile.

This level of detail in customer profiles also takes fraud detection to a new level. For example, a staged car accident is very difficult to detect on paper. Only when the people involved in the accident are put into a data context can correlations and relationships be uncovered. Do the people involved in the accident, the vehicle owners or the witnesses know each other?? Several such claims have recently been filed by the insurer? Does the insurance for the vehicle expire? Or is it due for inspection or maintenance? If, for example, a 15-year-old car is involved in an accident shortly before the next MOT, the closer circumstances of the accident should be checked again. Data analysis provides important starting points here, which the Allianz Benelux team is investigating further.

Operating profit: two million euros

After a successful proof-of-concept, Allianz Benelux implemented the graph-based solution – with immediate success. Over the course of just two years, the insurer was able to determine an operating profit of two million euros. However, the actual value is much higher. Existing systems, combined with graph technology, are much more powerful and simplify the verification of complex, heterogeneous data. The field of application for graphs is far from exhausted. The agenda includes data scouting and contextualizing data for business-relevant questions as well as an analytics engine for customer support. Further plans to use Neo4j together with machine learning techniques to further automate Allianz Benelux's core processes and make them more efficient.

The more data that can be converted into a visual view, the more precisely risks can be identified and customer service improved. Graph technology enables this 360-degree view even with large, complex data sets. The graph is constantly evolving, growing with the customer, matching needs and contracts with the insurer's current portfolio of benefits and services, and can analyze risks not only descriptively but also predictively.