Customer Marketing Insights

Key Highlights

  • The enhanced customer retention model enabled by machine learning predictions improved the client’s ability to successfully predict members who were most likely to discontinue their memberships. This has the potential of around half a million dollars in savings annually as well as the benefit of enhancing the brand for attracting new members and organizations looking to purchase physician data

Challenges

The client ask was to segment customers based on their ethnicity, disposition type, gender, net sales and household income

Approach

  • Completed an in-depth study of existing approach, data characteristics, features used to understand the pitfalls in the solution
  • Enriched the data, did feature engineering with domain and stats driven approach
  • Model Selection - The Random Forest model with SMOTE was the best performing model with a 2% increase, including marketing insights data. The ROC(Receiver Opera/ng Characteristics) curve was used to measure model accuracy. Under sampling, oversampling, synthetic data genera/on and cost-sensitive learning, methods used to address the imbalanced dataset.
  • Operationalizing the Model - KNIME Server, a Data Science & Machine Learning Platform, was implemented and used for versioning and deployment

Tech Stack

R, DB2 (IBM), KNIME Server and SAS

Download 14+ Case Studies

Related Case Studies