Case Study: Design of an AI-Driven Solution to Reconcile and Clean Customer Data

The Client

  • The client is one of the largest insurance brokers in Germany, providing policies to private and corporate customers.
  • The client wanted to standardize their partner numbers codes and create a primary key to uniquely identify each customer and its relationship with channel partners.

The Challenges

  • Customer names, contact details, contract history and other critical identifying attributes were consistently incomplete or contradictory, which lead to policy errors and the need for excessive manual intervention.
  • As a result of merging data sheets over time, individual customers had multiple channel partners assigned to them, making it difficult to distinguish individuals across the client’s distinct branches.

Solution Delivered

  • Design and execution of an AI-driven solution to reconcile and clean customer and partner data records using natural language processing (NPL) techniques that produced distance-based similarity scores across all customers, stratified by key variables to not only clean but also enhance the data, enabling accurate customer scoring.
  • Identification of valid partner numbers using waterfall logic constructed to prioritize the client’s branch and active cases for the purpose of re-assigning unique primary identification IDs to each customer.
  • Built the technical framework powered by R, Excel and Equipped’s proprietary NLP engine that simplified future reiterative deliverables.

Results & Benefits

  • Client was able to easily identify each customer and its attributes by newly assigned primary number codes which facilitated both highly accurate cluster building and segmentation.
  • Customer identification and proper cohort grouping aided pricing models for new client policies, renewals and underwriting.
  • Granular customer insights fuelled unpreceded modelling depth, scenario analysis and pricing optimisation.