PERFORMANCE MEASUREMENT IN THE WAREHOUSING INDUSTRY Abstract Warehouses are a substantial component of logistic operations, and an important contributor to speed and cost in supply chains. While there are widely accepted benchmarks for individual warehouse functions like order picking, little is known about the overall technical efficiency of warehouses. Lacking a general understanding of warehouse technical efficiency and the associated causal factors limits industry’s ability to identify the best opportunities for improving warehouse performance. The problem is compounded by the significant gap in the education and training of the industry’s professionals. This paper addresses this gap by describing both a new methodology for assessing warehouse technical efficiency based on empirical data integrating several statistical approaches and the new results derived from applying the method to a large sample of warehouses. The self-reported nature of attributes and performance data makes the use of statistical methods for rectifying data, validating models, and identifying key factors affecting efficient performance particularly appropriate. This paper also identifies several opportunities for additional research on warehouse assessment and optimization. Supplementary materials are available for this article. Go to the publisher’s online edition of IIE Transactions for appendices and additional tables. Keyword Warehouse, facility logistics, data envelopment analysis, outlier detection, two-stage DEA 1. Introduction Performance assessments in warehousing are needed to identify the options in design and operations that confer the greatest benefits (i.e. “speeding up” the supply chain, minimizing order picking costs, etc.). There are two related but distinct approaches to performance measurement: economic (i.e. revenue related to cost), and technical (i.e. outputs related to inputs). Economic performance assessment is somewhat difficult because warehouses typically do not generate revenues; rather, their function is tosupport the supply-chain including bricks-and-mortar and Web-based outlets. Moreover, since a firm’s warehouses can be sited in urban, rural, or international locales, the differences in the settings will have a major impact on the costs of the resources used by each warehouse, such as labor and building space. Further, the acquisition costs of capital equipment specific to warehouses vary depending on general economic conditions and the buying power of the specific warehouse owner (e.g., large 3PL versus start up company). For these and other reasons technical measures based on output generated and resources consumed tend to give a clearer picture of operational performance when assessing warehouses across a group of warehouses because the measures avoid the uncertainty or variation introduced when using financial measures directly. Technical performance measurement in the warehouse industry traditionally employs a set of single factor productivity measures that compare one output to one resource (or input). This is sometimes called the ratio method (see Tompkins et al. 2003, Chen and McGinnis 2007). However, using a set of ratio measures can lead to confusion—if some measures are good and some are poor, is the warehouse performing well? Thus it is more useful to employ a measure that considers simultaneously all of the significant inputs and outputs. The field of production economics (Fried et al. 2007; Coelli et al. 2005) provides a variety of approaches to assessing technical efficiency when there are multiple inputs and outputs. This paper uses the approach of Data Envelopment Analysis (DEA) (Charnes et al. 1978), presents several adaptations which make the approach more applicable to self-reported warehouse data, and summarizes the results of applying the adaptations to a large and diverse sample of warehouses. Data on warehouse performance collected over a 5-year period (2001-2005) is used to benchmark the performance of each observation against all other observations in the data set. The purpose of the reported study is two-fold: (1) to develop useful methods by which both individual warehouses and groups of warehouses can be evaluated with regard to technical efficiency, and (2) to identify the operational policies, design characteristics and attributes of warehouses that are correlated with greater technical efficiency. The results reported are reflective of the data set and methods used. To the extent that the data is representative of generalwarehouse operations and the practitioner accepts the assumptions related to the models and methods used, the conclusion of this paper are reflective of warehousing best practices for general warehousing operations. 2. Literature Review Surprisingly, the technical literature on warehouse performance assessment is meager. Two streams can be identified: papers which propose a framework for designing or analyzing warehouses and those which directly address performance assessment. Rouwenhorst et al. (2000) and Gu et al. (2007) typify the first category. Both address the coordination problems that arise from the investigation of warehousing subproblems. Rouwenhorst et al. suggest a framework in which to place these problems, but it is largely descriptive and does not provide an operational technique to coordinate the design decisions. Gu et al. categorize the decision problems associated with design and operation rather than overall warehouse performance assessment. The second category, logistics benchmarking, is used by more than half of the Fortune 1000 companies to improve productivity and quality (Foster 1992). However, prior to the start of the benchmarking project discussed in this paper (2001) relatively few warehouse benchmarking results are described in the literature. Three notable exceptions are Stank et al. (1994), Cohen et al. (1997) and Hackman et al. (2001). Stank et al. gathered survey data from 154 warehousing companies to determine if they employed benchmarking and in what