#353 1969 · Toyota · Manufacturing
Toyota shrank every batch until defects had nowhere to hide
the problem
Flaws surfaced only after thousands were made
background
By the 1950s the world's dominant manufacturing doctrine was Ford-style mass production: huge batches, dedicated tooling, and changeovers so slow that swapping a die on a stamping press could tie up a line for hours. Toyota, rebuilding after the war with a fraction of Detroit's capital and demand for only a handful of units per model per day, could not afford the giant production runs the American model assumed.
Large batches were the obvious answer because every textbook and every competitor used them: spreading a slow, expensive die change across thousands of identical parts minimized the changeover cost per unit. Idle-machine time during setup was treated as an unavoidable tax on production, not a variable anyone expected to shrink — so nobody optimized for it.
what everyone would do
The obvious response to catching defects late was to inspect harder or more often within the large batch — more quality checkpoints on the mass-production line already running. That doesn't fix the underlying problem: with changeovers taking hours, no plant could afford to run small batches to catch defects sooner even if it wanted to, so the entire industry inspected around a batch size everyone treated as fixed rather than questioning the batch size itself.
what they saw
Shingo saw that setup time wasn't a fixed physical cost of changing a die — it was mostly time spent doing steps that only needed the machine to be running, once those steps were separated from the ones that genuinely required it to be stopped. Changeover speed, not batch size, was the real constraint the whole industry had been optimizing around instead of attacking directly.
the move
Starting around 1969, Shigeo Shingo worked with Toyota's press shops to separate 'internal' setup steps, which require the machine to be stopped, from 'external' ones that can be done while it keeps running, and to convert as many internal steps to external as possible. This turned die changes that had taken roughly one to several hours into changes measured in single-digit minutes — the method he later named Single-Minute Exchange of Die (SMED).
why it works
Splitting setup into 'internal' steps (machine must be stopped) and 'external' steps (can be prepared while the machine still runs) and moving as much as possible into the external category means the machine's actual downtime shrinks to only the steps that truly need it idle, cutting changeovers from hours to minutes. Once changeovers are cheap, running small batches stops being economically irrational — a defect introduced by a bad die or a drifting process shows up within the same shift's small batch instead of being buried inside a week's output, so the feedback loop between a defect occurring and someone noticing it collapses from days to hours. That faster loop is what let Toyota catch and fix problems at their source instead of discovering them downstream in a mountain of already-defective parts.
the payoff
By later Toyota Production System accounts, press die-changeover times fell from hours to an average of around three minutes by the early 1970s, which let Toyota run batches small enough that a defect surfaced within the same shift instead of being buried in a week's worth of parts. The precise plant-by-plant sequence is disputed — historians differ on how much credit belongs to Shingo personally versus Toyota's own engineers — but the order-of-magnitude time reduction and its role in enabling just-in-time, small-lot production are well corroborated across Toyota Production System literature.
where it breaks
The method only pays off where changeover cost is genuinely the binding constraint on batch size and where a real internal/external split exists to exploit — a process with almost no external-izable setup steps, or one where changeovers are already fast relative to run time, has little slack for SMED to capture. It also requires the organizational discipline to actually run the resulting small batches once changeovers are fast; a plant that halves its changeover time but keeps batch sizes large out of habit captures none of the quality-feedback benefit, only faster changeovers. And the approach assumes defects are worth catching early — for a process where per-unit inspection is cheap and batch size doesn't affect time-to-detection, the whole mechanism has nothing to bite on.
what came after
Toyota folded fast changeovers into the broader Toyota Production System formalized by Taiichi Ohno, and Shingo's own books carried the method worldwide after he began consulting internationally in the late 1970s and 1980s. SMED became a named, teachable discipline independent of Toyota, applied across manufacturing, healthcare, and software as a general method for attacking any setup time assumed to be fixed.
references
- [1]A Revolution in Manufacturing: The SMED SystemProductivity Press (via Internet Archive), 1985archive.org
- [2]Shigeo Shingo – The People Behind The Big Ideas of Operations ManagementUtah Education Network (Pressbooks OER textbook), 2021uen.pressbooks.pub
- [3]The History of Quick Changeover (SMED)AllAboutLean.com, 2018allaboutlean.com