Data envelopment analysis (DEA) is a method that computes a relative efficiency score for an entity by comparing it to a population of "peer" entities. The method requires defining a set of inputs (resources) and outputs (products or services) that is common to all entities. In the context of warehousing, for example, the resources might be space, equipment, labor, and inventory, and the services might be customer orders shipped. To compute an "input efficiency" for a particular warehouse (the "target" warehouse), DEA considers the output of the target, and constructs a hypothetical warehouse (the "reference" warehouse) that produces the same output, but uses the least possible input. The reference warehouse is a composite of other warehouses in the peer group, so it represents what is hypothetically possible. Given a reasonably large peer group, DEA can do a reasonably good job of assessing the relative efficiency of the target. A major barrier to widespread use of DEA has been the difficulty of data collection. The iDEAs (www.isye.gatech.edu/ideas) project in the Keck Virtual Factory Lab has developed a web-based application using DEA to perform warehouse efficiency analysis. To date, users have provided over 600 warehouse descriptions by logging on to the website. Conservative screening of the data reduces the peer group to just under 200 distinct warehouse records. Preliminary analysis of the peer group indicates that a large fraction of the warehouses operate with relative resource efficiencies of 50% or less. Based on these results, it appears there are very large opportunities for improvement. The problem is to diagnose, for a particular warehouse, exactly what that warehouse needs to do to improve efficiency. The traditional approaches have been to perform a "benchmarking study" comparing to other warehouses to identify best practices, or the implicit benchmarking achieved by hiring consultants. Recently, the iDEAs team began exploring the possibility of using our DEA-based performance assessment tool as a platform for benchmarking. The fundamental idea is fairly straightforward - in addition to input-output data, collect data on warehouse attributes (seasonality, order variability, etc), and warehouse technology and practices (pick-to-light, velocity-based slotting, etc); periodically analyze the database with respect to possible correlations between efficiency score (the dependent variable) and warehouse attributes, technologies, and practices (the independent variables). Suppose there is a strong correlation between some attribute, technology, or practice and system efficiency. A target warehouse can evaluate the correlation and where they currently stand relative to that attribute, technology, or practice in making decisions about initiatives to improve system efficiency. Clearly, the success of this approach to benchmarking depends upon the participation of large numbers of warehouses. However, the cost per warehouse is very small (the cost of collecting the data), and the cost of maintaining the website is relatively small. The payoff is potentially quite large, if individual warehouses can identify opportunities for improvement, either by modifying their attributes (pricing policies to reduce variability), policies (reducing employee turnover), and adopting technologies for which strong efficiency results are indicated. The presentation will describe the tentative benchmarking results obtained to date, and discuss technical and practical issues for this approach to benchmarking.
|Number of pages||1|
|State||Published - 1 Dec 2004|
|Event||IIE Annual Conference and Exhibition 2004 - Houston, TX, United States|
Duration: 15 May 2004 → 19 May 2004
|Conference||IIE Annual Conference and Exhibition 2004|
|Period||15/05/04 → 19/05/04|