Employee attrition modeling is one of the most widely touted applications of predictive modeling. The process of building and deploying an attrition model can seem deceptively simple: prepare the data, apply a predictive algorithm, and voila, you have a reliable indication of which high-value employees are most likely to quit. While building a basic employee attrition model might seem easy, building one that works in action – that truly and effectively serves your business need – is another matter entirely.
Perhaps the most common source of model dysfunction is a failure to align the model’s design with the business need at hand. Many beginners learn to build models optimized for Risk Segmentation. These models work well for looking at a large group of employees and identifying who among them is most likely to quit. However, while risk segmentation is valuable, many organizations need predictive models to serve other needs. Some organizations need models that can reliably Forecast attrition rates, while others use predictive modeling to conduct Policy Evaluations that can help shape interventions to reduce attrition. Models optimized for risk segmentation will not necessarily serve these other needs.
In this presentation, we’ll review the basics of building an employee attrition model, and then show how it can be optimized to suit a wide range of business needs.