Our client is a major provider of health insurance plans in the US. For its members, reaching out to the call center is the most common way to resolve any issue regarding healthcare plans. The Head of Customer Service wanted to identify analytical opportunities to reduce occurrences of repeat calls by predicting members that are at high risk of calling repeatedly.
More specifically, the client wanted to:
- Develop a predictive model to flag members who are at high risk of making a repeat call
- Identify top drivers that cause a member to call repeatedly
- Alert and enable advocates to realistically engage with customers to avoid repeat calls
- Defining a repeat call considering a range of call topics and different intervals between the calls
- Deciding upon the right intervention strategy and tool to display repeat caller’s information
- Performed a detailed analysis to define a repeat caller through EDA and First Contact Resolution metric.
- Tested specific hypotheses with the available data, through features related to call & inquiry history, prior authorization requests, claim denials, provider details, active usage of web apps, and member profile
- Developed a predictive model using XGBoost algorithm in Data Science Studio
- The model is currently used to display an alert in a contact center application and guide the agent with tip-offs to resolve the call
- There is a projected estimate of over USD 2 M in annual operational savings
- Improved NPS scores that would save millions of dollars if performance guarantees are not met as per contract