How to use it
- 1. Edit the fleetEight warehouses: monthly cost, staff, and orders shipped. All cells editable.
- 2. Read the frontierEach site is plotted by resources per order. The red hull is best practice - nobody does better.
- 3. Click a rowThe ray shows how far that site must shrink its inputs (same output) to reach the frontier - that’s its score.
- 4. Find the peersInefficient sites are scored against a blend of real efficient ones - their benchmark, by name.
Warehouse efficiency frontier
Data (per month)
| Site | Cost $k | Staff | Orders k | Score |
Reading a score
Click any row.
–
Efficient sites
–
Weakest score
–
Fleet average
–
Recoverable input
Axes are inputs normalised per 1k orders (constant returns to scale). Lower-left = better. The red frontier is the convex hull of best practice; the dotted ray is the selected site’s radial path to it.
Try these experiments
- Give one site absurdly low cost but huge staff - it can still be 100% efficient. DEA lets every DMU pick the weighting that flatters it most.
- Delete-the-star test: make the best site terrible and watch every score change. Efficiency here is relative, never absolute.
- Find a site at 85%: the tile shows exactly how much cost + staff it should shed with zero loss of throughput - that’s the audit conversation.
About this tool
The concept
Data Envelopment Analysis compares units that turn multiple inputs into outputs. No prices, no weights imposed - the data itself envelops a best-practice frontier.
How it works
CCR input orientation, two inputs one output: normalise inputs per unit of output, build the convex hull, and score each site by how far its ray can contract before hitting the frontier.
Winning strategy
Use DEA to find who to copy, not just who to blame: an inefficient site’s peers are real, comparable operations whose blend defines its target.
Background: Data envelopment analysis · Charnes, Cooper & Rhodes (1978)