Labor Planning and Shift Scheduling in Retail Stores Using Customer Traffic Data

Abstract

Labor costs are one of the largest expenses of retail companies. This work develops a decision support tool to manage staffing levels and schedule working shifts, balancing the sales contribution of employees with salary costs. We extend recent empirical studies that estimate the impact of staffing levels on sales, combining detailed sales transaction and hourly customer traffic with employee staffing data. The empirical strategy uses exogenous deviations between the planned and actual schedule to identify the causal effect of staffing on conversion rates and basket value. This empirical model combined with traffic forecasting models are used as an input to a scheduling algorithm that optimizes the workforce schedule by balancing revenues with employee salaries, providing a feasible and detailed specification of working shifts that complies with labor regulations and other practical constraints. We also propose a robust optimization model to face uncertainties coming from the parameter estimation. The methodology is tested using real data from a children apparel retail chain in Latin America. The results suggest that most stores are understaffed during the weekends and at the same time overstaffed during lunch hours in weekdays, which is in part explained by the reduced flexibility imposed by regulatory restrictions for full-time workers. We show that our approach to optimize the workforce schedule can effectively improve the efficiency of the current labor plan of the company. Finally by using a robust approach we find staffing plans that produce less variance in the outcomes without losing too much in terms of expected revenue. The methods and models developed in this work have been integrated into a packaged solution for retail chains.

Keywords: Retail Operations, Empirical Research, Forecasting, Staff Scheduling, Robust Optimization