Leveraging Curve’s approach to data science

Firstly, we worked with the team to understand their challenges, objectives and requirements.

From there, we followed Curve’s structured approach to data science:

1. Use case definition – including hypotheses to be tested
2. Data acquisition – gathering relevant data such as demographics, transaction history, engagement and usage patterns
3. Data preprocessing and cleansing – to handle missing values, outliers, and inconsistencies
4. Feature selection and engineering – identifying relevant features that could be used for the model
5. Model selection – evaluating different ML algorithms considering factors such as accuracy, interpretability, and scalability
6. Model training, evaluation and optimisation – training the selected churn model using historical data and fine-tuned parameters to optimise performance
7. Validation and testing – testing the model’s performance on unseen data to validate effectiveness in predicting outcomes
8. Deployment – integrating the final model within client systems and setting up automation based on their needs
9. Handover or support – ensuring maintenance procedures are in place to support ongoing usage and improvement

Award-winning churn mitigation programme

Our client informed us that “the business has seen excellent results from the personalisation campaigns enabled by the models”.

The churn model enabled a new Customer Programme which won Gold at the 2023 Digital Marketing Awards:

“Some 62% of consumers said they would recommend BAT to friends and family. In addition, the campaign drove an increase in sales conversion and total revenue that would not have been achieved without the churn mitigation programme.

The personalised omni-channel initiative reduced the net-new subscriber attrition rate by an average of 25 percentage points year on year. BAT’s digital and call-centre activation has helped to reduce churn by 32%.”