Predictive analytics and HR data
How to predict the retirement age of your employees using HR Analytics?
The client’s starting point
- The HR division of the shared services centre of a major international energy group
- A division that partners with the other entities of the group, seeking innovative solutions to offer to its internal clients, in particular by making the most of the HR data at its disposal
- The desire to accurately forecast retirements to better manage the payroll, anticipate knowledge transfers and succession plans
- The development of five retirement hypotheses that will be confirmed or denied by data
- The implementation of a methodology showing a calculation variable established as the difference between the actual departure date and the theoretical departure date
- The construction of two statistical models (one linear and one random forest) to forecast retirements
The whole story
Our client, the HR division of the shared services centre of a major energy group, is looking for a new value-added service proposition that would leverage the large volume of HR data at its disposal and help identify HR Analytics as a powerful driver of productivity and efficiency.
It has been determined that predicting the retirement age of employees would allow for more efficient management of payroll, knowledge transfer and succession planning. Akoya Consulting was mandated within the framework of a Proof of Concept (PoC) to develop a predictive model of retirement on a perimeter cumulating several thousands of retirements over the last 10 years.
The firm took stock of the available HR data and aggregated the potential causes of retirement – demographic, social, geographical, etc. – and drew up five initial hypotheses which the analysis had to confirm or refute during the process.
Akoya Consulting then set up a methodology allowing the comparison of retirement ages between individuals (thanks to an actual retirement date expressed as a delta with a theoretical retirement date) and the realization of two predictive models, linear and random forest, for the client teams.
The models delivered to the client reduced the uncertainty about the actual retirement date of an employee to a few months, compared to more than 12 months previously. Akoya Consulting also established a list of recommendations to improve the accuracy of the models developed in the PoC phase.