The Acme Stamping benchmark
A value stream tool is a pile of arithmetic wearing a diagram. The only honest way to show the arithmetic is right is to run it against numbers somebody else published first.
What is checked
The current-state example in Rother & Shook’s Learning to See is the most widely taught value stream map there is, and the book states its totals. The engine is given the same plant — the same demand, calendar, cycle times, changeovers and inventory counts — and asked for the same three numbers.
| Figure | Published | Computed |
|---|---|---|
| Takt time | 60 s | 60 s |
| Processing time | 188 s | 188 s |
| Production lead time | 23.6 days | 23.62 days |
The finding it has to reach
Matching totals are necessary and not sufficient: three numbers can agree by coincidence in a model that is wrong everywhere else. So the engine is also held to the example’s central conclusion — assembly 1, at a 62 s cycle time, cannot meet a 60 s takt. A tool that reports the right lead time and misses the constraint the whole exercise exists to find has not reproduced anything.
The figure it refuses to report
The book’s data box gives assembly 1 an uptime of 85%. That is one factor. OEE is availability × performance × quality, and a product that multiplies the one number it has by two it assumed would produce a confident figure with no measurement behind it. This engine reports an A×Q score instead and names the component it is missing. Declining to answer is part of what is being verified here, and it is why the benchmark includes a case with no answer in it.
What this does not prove
The fixture is reconstructed from the published map, not transcribed from a dataset the authors released — the provenance argument and its limits are set out in the methodology.
It is also a partial exercise, and deliberately reported as one. The current state was published in 1998 without quality data and without the time accounting OEE requires, so it can only test the flow half of the analysis: every yield reads 100%, no OEE is computed, no connection is pulled and no process carries a rework loop. That makes it the right thing to verify against and the wrong thing to evaluate the product on — which is why the tool opens on a worked reference study of a compressor-bracket line, not on this.
Reference: Mike Rother & John Shook, Learning to See: Value Stream Mapping to Create Value and Eliminate Muda, Lean Enterprise Institute.