Abstract
Employees are vital assets to any organization, and their departure can result in reduced human capital and operational disruptions. To mitigate this, companies employ predictive analysis to forecast potential employee churn. Probability-based modeling for projecting employee churn is an underexplored area in HR analytics. This paper tests the applicability of the shifted-beta-geometric (sBG) and beta-discrete-Weibull (BdW) models within the context of employee survival projection. Using data from three cohorts of employees, we compare the results of these models with each other as well as with linear and logarithmic regressions. Our key finding is the superior performance of the BdW model, which can capture differences in churn rates between employees and within employees over time. The beta distribution captures the heterogeneous employee loyalty, while the Weibull distribution effectively captures retention rate changes over time. Our research demonstrates that parsimonious probabilistic models, which require minimal data and have so far been used only in customer analytics, can be applied in HR analytics for projecting employee retention curves.