Case Study: Personalizing credit card risk management at one of Canada’s largest banks

The company

A multinational Canadian bank with a broad range of products and services including personal and commercial banking, wealth management, corporate and investment banking. The bank is one of the largest both nationally and globally.

The challenge

The bank wanted to identify and deploy a new model that could track and analyze behavioural data for millions of its credit card customers. Objectives for the new predictive model included:

  • Personalizing customer experience
  • Improved collection recoveries and lowered risk
  • Installation without the need to overhaul existing IT systems

Old statistical models used to predict customers at high-risk for delinquent payments were limited in the following ways:

  • Refreshing models with new customer data took over one year to deploy and was very costly
  • Models could not track real-time changes to customer data
  • Models could not identify complex patterns within data
  • Models could not produce personalized suggestions on how to best reach customers for payment resolution

The solution

The bank approached to design, build and deploy a deep learning model that would offer solutions to the old models’ challenges in the following ways:

  • created a machine learning model that could analyse data holistically
  •’s model integrated and analyzed updates to customer data in real-time
  •’s model drew insights from complex data relationships
  •’s model produced insights on how to best reach customers to resolve late or missed payments


In a cross-comparison between the bank’s old statistical model and’s deep learning software,’s software consistently outperformed the bank by at least 20% each month.

Unlike the bank’s old model,’s model also offers banking executives the power to generate personalized customer insights from data.

Powered by their platform Frontiers, the team was able to quickly identify the best deep learning model to build to reach the bank’s objectives for credit card management. By pairing Frontiers with technological expertise, the team also ensured the model was seamlessly installed, without the need to overhaul the bank’s existing legacy IT infrastructure.’s successful deployment of a model to enhance the bank’s credit card risk management was a great first step in demonstrating the technology’s potential for new solutions in retail banking. With optimized customer communications and a better bottom line as the result of’s model, the bank will continue to partner with our team to move beyond a narrow and specific collections tool. In the coming months, will work with the bank to build holistic strategies for use in branches and across its digital channels.


March 1, 2017