Performance dynamics of warehousing innovations: a productivity and efficiency perspective

Assessing warehouse productivity in a dynamic and interpretable manner remains challenging due to system heterogeneity and the prevalence of qualitative data. This study develops a hybrid analytical framework that combines the Malmquist Productivity Index (MPI) – which decomposes productivity change into efficiency and technological components–with machine learning–based analysis to enhance interpretability. The framework extends conventional MPI applications by incorporating ordinal performance indicators, such as automation level and service flexibility, into a Data Envelopment Analysis (DEA)-based model. Using a two-period panel dataset of 102 warehouses, the empirical analysis quantifies productivity change over time and compares MPI-based results with static DEA and cross-efficiency evaluations. The findings indicate that static assessments alone may not capture longitudinal improvement, whereas the MPI decomposition reveals both frontier performance and catch-up effects. The machine-learning component further identifies operational factors most closely associated with productivity change across warehouse types. Overall, the study contributes to performance-measurement literature by offering a data-driven, nonparametric, and interpretable approach to productivity analysis, delivering actionable insights for warehouse benchmarking, design, and investment decisions.