Inventory Record Inaccuracy: Optimal Inspection and Labor Effects

Inventory record inaccuracy (IRI) is a pervasive problem in retailing and causes non-trivial profit loss. We first propose a formulation to capture the dynamics of information decay and use a fairly general cost structure of inspection programs to tackle IRI. Within this formulation we devise two optimization models that represent current practices in industry: daily-fraction inspection and all-or-none inspection. Some qualitative insights about the interaction between inspector fallibility and inspection efforts are derived from steady-state analytics assuming risk-neutrality. We perform an empirical case study in which we derive estimates of all model parameters and identify deficiencies of store operating practices using actual data from a global retailer. We further postulate that labor is key to reducing human-related errors that drive IRI. We derive two hypotheses on the association between IRI and labor. After obtaining longitudinal data from the retailer, we adopt Bayesian shrinkage estimation to derive a robust measure of IRI based on more than 60,000 correction records. The panel data analysis shows that both the level and the mix of store labor have substantial impacts on IRI. Our modeling efforts derive qualitative insights that will help managers design cost-efficient inspection policy and allocate labor capacity with foresight.