Cantab PI
Retail Banking

Empowering retail banks to rapidly generate value from their customer data

Automated Machine Learning for Retail Bank

Enabling data science teams in retail banks to rapidly deliver value by developing, deploying and managing hundreds of predictive models with proven ML architecture
  • Data Lake: Enabling rapid integration of data from multiple data warehouses and product factories


  • Algorithms and Automated Machine Learning: Partially pre-trained algorithms and advanced ML/AI technology developed and validated on over 20 million of retail banking clients


  • Integration: Modular tools enabling streamlined integration of ML/AI models with the existing legacy systems with minimal disruption and investment

Feature generator

Auto ML (model training)

Model interpretation

Data Lake & Reporting

Model comparison and performance analysis

Model deployment

Hyper-personalization of digital customer journeys

Hyper-personalization of digital customer journeys

Deploying ML/AI models to optimise digital actions, channels and psychological triggers for retail bank’s customers by hyper-personalization

Supported use cases

  • The next best action
  • Optimal psychological trigger
  • The next best channel
  • The optimal contact time
  • Message personalization

Client example

AVO Africa – a „super App” by the leading South African bank Nedbank, powered by Cantab PI autoML

Use cases

  • Lead generation and targeting for cash loans

    Increase in banking products sold with the same marketing spend, enabled by a propensity to buy ML algorithm


    Increase of the banking products sold by


  • Preventing customer churn

    Reduction of customer churn, and measurable improvement of customer satisfaction, enabled by four ML algorithms flagging likely churn by churn driver


    Reduced customer churn by


  • Optimisation of omni-channel targeting

    Increase in revenue due to ML-driven optimisation of sales channels and messages

    Increase in revenue


  • Accurate credit scoring on current account (PSD2) data only

    Deployed behavioral credit scoring models using only data from current account transactions of the client, with performance superior to the legacy models relying also on credit history. 

    Gini coefficient of the developed credit scoring model