Benchmarking a warehouse

This is material to supplement the book Warehouse & Distribution Science, a graduate course taught at the Georgia Institute of Technology by John J. BARTHOLDI, III and Steven T. HACKMAN. Everyone is welcome to use the book and materials here for educational purposes, so long as the copyrights remain intact. You will find more information and technical details on these topics within the book.

Ratio-based benchmarks

Standard statistics that are gathered for most warehouses include pick-lines per person-hour, orders shipped per year, warehousing costs as a percentage of sales, and so on. Each of these is of the form "(number of units of output generated) divided by (the number of units of input required)". Such statistics are fairly easy to measure and can give some insight into performance. See, for example, the annual Warehouse Benchmarking Survey of The Supply Chain & Logistics Institute.

Such simple ratios are attempts to gauge warehouse productivity and many warehouses are managed by examining a collection of such measures (or key performance indicators, KPI's). But each one represents a narrow point of view and so can be misleading. For example, warehouse costs as a percentage of sales depends on sales, and so the marketing department can affect how the warehouse is scored.

The problem with this is that the collection of productivity measures is no more than that: just a collection. There is no structure relating them and so each one is just another "hand on the elephant", another point of view but without any hints about how to integrate them.

System-based benchmarks

Data-envelope analysis (DEA) is a way to measure how well a facility generates a portfolio of outputs for a given portfolio of inputs. But to use DEA, we must accept a few simple economic assumptions, the most important of which is that warehouse operations are scalable, so that, for example, a warehouse with twice the inputs will produce twice the outputs. In the terminology of economics, we must assume "constant returns to scale".

The key feature of DEA is that, in effect, it evaluates your warehouse against an entire community of other warehouses. It does this by using linear programming to blend the best warehouses in the community to create a "synthetic" warehouse that can be directly compared to your. The synthetic warehouse does not actually exist, but rather is a scaled combination of real warehouses.

See the textbook for details.

This type of benchmarking has several advantages over simple ratios. First, it accounts for entire portfolios of inputs and outputs. Second, the warehouses of the community can remain anonymous, which reduces worry about divulging confidential information.

One must be aware of possible problems with this approach as well. For example, the evaluation of a warehouse can be quite sensitive to outliers, so that inaccurate data can skew everyone's ranking.

Professor Hackman was the first to apply DEA to warehouse benchmarking and an on-line implementation of his DEA model may be found here.