Just-in-time
pull system · lean inventory · Toyota production system
Producing or ordering only what the next step actually needs, when it needs it, rather than building stock in advance. Developed at Toyota, where the point was never cost — it was that holding no buffer makes every problem visible immediately.
In practice
A business with three months of work booked cannot tell whether its sales process is working. The backlog is a buffer, and a buffer hides the signal for exactly as long as it lasts.
The common mistake
Reading it as an inventory cost-cutting measure. Cutting the buffer without first fixing what the buffer was concealing produces the failure, on schedule, with no slack left to absorb it.
Just-in-time is usually explained as holding less inventory, which is what it looks like from outside and is not what it is for.
Taiichi Ohno's argument at Toyota was about information. Inventory between two steps means the second step can keep running when the first one has a problem. That is useful and it is also how the problem stays hidden: the buffer absorbs the fault and nobody has to deal with it. Remove the buffer and the fault stops the line immediately, in front of everyone, which is unpleasant and is the entire point.
Ohno's image was water covering rocks. Lower the water and you find out where the rocks are. You are not lowering it to save water.
Why the imitations failed
Western manufacturers adopted the visible half through the 1980s and 1990s — cut inventory, push stock onto suppliers, run lean — and a good number of them got worse. They had removed the buffers without building the thing the buffers were standing in for.
Toyota's system works because stopping the line is normal, a stoppage triggers an immediate investigation, the cause is fixed rather than worked around, and suppliers are close partners rather than pressured vendors. The low inventory is a result of those, not a substitute for them. Cutting the buffer first produces a fragile operation that discovers its rocks by hitting them.
The business version of a buffer
Most buffers in a small business are not stock, which is why the idea gets filed as irrelevant to services.
A backlog. Three months of booked work is a buffer against a sales process that may not be working. You will find out in three months.
Slack in estimates. Padding every task hides which steps are actually slow, because everything finishes roughly on time by construction.
A cash reserve. This one is different and is the important exception below.
Unfinished work in progress. Fifteen half-done projects buffer against running out of things to do and guarantee that everything takes longer, which is Little's law.
Removing the first, second and fourth surfaces information. Removing the third does not.
Where it stops, and why cash is not a rock
The distinction that keeps this from being dangerous advice: a buffer against variation you can fix is hiding a problem; a buffer against variation you cannot fix is a margin of safety.
A queue in front of a step that keeps breaking is hiding the breakage — fix the step. A cash reserve is not hiding anything; it is protection against a world that will occasionally do something you did not predict, and running without it does not surface information, it just means the next surprise is terminal.
The pandemic settled a version of this argument publicly. Supply chains run to minimum inventory discovered that they had optimized against a distribution of disruptions that turned out to be wrong, and the businesses that survived best were frequently the ones carrying buffers that had looked like inefficiency for a decade. Buffers are not free and neither is their absence.
Using it honestly
The useful question is not how much buffer can be removed. It is: which of our buffers is concealing something we could fix, and what would we learn by lowering it a little?
Lowering it a little is the operative phrase. The version that works is incremental and reversible — take a week out of the lead time, cut work in progress from fifteen to eight, shorten the reporting cycle — and each step exposes one rock at a time, which is the pace at which rocks can actually be dealt with.
Concept web
Open the full webQuestions
What is just-in-time?
Producing or ordering only what the next step needs, when it needs it, rather than building stock in advance. It was developed at Toyota, where the purpose was to make problems visible rather than to cut inventory cost.
Why did just-in-time fail for many companies that copied it?
They removed the buffers without building what the buffers were standing in for — stopping work when something breaks, investigating the cause, and fixing it. Low inventory is a result of that discipline, not a substitute for it.
When should you keep a buffer?
When it protects against variation you cannot fix. A queue in front of a step that keeps breaking is hiding a fixable problem. A cash reserve is not hiding anything, and removing it surfaces no information.