Quantitative Models for Workforce Management in a Large Service Operation

This paper develops quantitative models for optimizing employee scheduling in large labor-intense service operations that face complexities such high employee turnover, contractual service-level agreements (SLA), non-billable overtime cost, and rising minimum wages. The workforce analytics part of the model uses payroll data to characterize employee features such as consistency, overtime affinity, and retention. The business analytics part of the model ensures that adequate employees are assigned to a shift to comply with SLA while minimizing anticipated overtime cost based on the prevailing overtime salary. A Mixed Integer-Linear Program is formulated with the objective of balancing total overtime cost reduction, employee preferences, and employee features. Additionally, employees can be clustered to identify distinct patterns based on their features of consistency, overtime, and retention. The proposed approach has been applied at a North American service provider with over 4,000 employees across more than 100 sites. K-means clustering based on employee features identified four distinct clusters. Deeper analytics using mixed-effects analysis can show the contribution to profitability from reliable employees, which are limited in availability. Boosted tree importance scoring can establish the influence of moderately reliable employees on overtime cost. Furthermore, decision tree model highlighted that tactical hiring and scheduling must account for collective workforce variability rather than individual attributes in isolation. A partial dependence plot helps visualize the relationship between employee reliability and overtime, thereby helping to characterize the impact of workforce mix on cost. Key impact from this work is that the company management is working to improve its recruitment, retention, and scheduling policies to better align business needs and human capital.