How Do You Stop Making Things Nobody Wants Yet? Wait to Be Asked

A conventional factory works from a forecast, producing according to a schedule prepared in advance and pushing the output onward whether or not the next stage needs it, which builds inventory wherever stages run at different rates. The alternative reverses the direction of the instruction. Each stage makes something only when the stage after it signals that it has consumed one, so work is drawn through the system by actual demand rather than pushed through it by a plan. The final signal comes from a real customer order, and every upstream instruction ultimately traces back to that.

The mechanism in its original form is deliberately physical:

• •A card travels with each container of parts through the process
• •Taking a container into use releases its card
• •The released card goes back to the supplying stage as authority to make more
• •No card means no authority, so nothing is produced
• •The number of cards in circulation caps the inventory absolutely
• •Removing cards tightens the system and exposes where it struggles

The system was developed at Toyota from the late 1940s under conditions that make its shape comprehensible. The company was short of capital and could not afford to hold stock, the Japanese market was small and demanded many models in low volumes, and space was limited. Taiichi Ohno, who led the work, credited American supermarkets as the inspiration, since a supermarket shelf is refilled according to what customers actually take rather than according to a forecast, and the shelf itself is the signal. The approach developed over roughly two decades and was not described in English until the 1970s, after which Western manufacturers adopted it widely and frequently misunderstood it.

The idea moved into software development and office work in the 2000s, where it looks different and follows the same rule. A board divides work into columns representing stages, cards represent individual items, and each column carries an explicit limit on how many cards it may hold at once. Nothing may be pulled into a column that is already full, so a bottleneck stops upstream work rather than allowing items to pile up in front of it, and the team must clear the blockage instead of starting something new. That limit is the whole mechanism, and boards without limits, which are common, are visual displays rather than the system they are named after.

The method is frequently sold as a way to reduce inventory, which it does, and that framing misses the more important effect. Limiting inventory removes the cushion that hides problems, so a machine that breaks down, a supplier that delivers late or a process that produces defects stops the line immediately instead of being absorbed by a pile of spare parts. That is uncomfortable and is the point, since a problem that stops production gets fixed while a problem that is absorbed continues forever. Adopting the card system without also building the capability to fix what it exposes produces a fragile factory rather than an efficient one, which is the commonest failure of the imported version.

A card returned from the next stage is the only authority to produce, so work is drawn through a factory by real consumption rather than pushed through by a forecast, and the number of cards caps inventory absolutely. Toyota developed it from the late 1940s while short of capital and space. Removing the inventory cushion exposes breakdowns and defects immediately, which is the actual benefit and the part most often skipped.