The Variance Learning Curve in Retail Order Picking

Abstract

Listen

Translate article

The notion that prior experience can improve future performance has reached general acceptance in the operations and supply chain management literature. However, prior experience may not only have a positive impact on average performance but in addition also on the variance of performance. We empirically examine this issue by considering the performance of 19 temporary workers in retail order picking new to this task, performing 22,603 storage location visits and use a two-stage estimation method of a heteroskedastic learning curve model. In the first stage, we estimate the impact of prior experience on task performance time. For the second stage, we extract the error terms from the first stage model and use it as the dependent variable in the second stage. Through this, we can estimate the impact of prior experience on performance variability. Our first stage model suggests that prior experience can improve average task performance on a system level for all new temporary order pickers observed. However, on an individual level, these learning curves are heterogeneous, with varying intercepts, slopes, and functional forms. Going beyond the impact of experience on average performance, our second stage model finds that one week of experience working in the examined warehouse can reduce performance variability by 26%, but at diminishing increments. These findings suggest that prior experience can mitigate variability in task performance time, which indicates that the recently discussed variance learning curve exists for order picking tasks. Our results allow managers to make informed decisions on how to benefit from the positive effects of prior experience on average performance and performance variability when employing a temporary workforce for the example case of retail warehouse order picking.