Buffers are not waste

Why some inventory is a design choice rather than a failure.

Buffers are not waste because a buffer's entire job is to sit unused most of the time, absorbing the ordinary unevenness between how fast one stage of a process produces and how fast the next one consumes, and judging it by how often it is drawn down misses the one day it is needed.

An umbrella in the boot of the car

An umbrella kept in the boot of a car sits there unused on most days, taking up a little space for no visible return, and judged purely by how often it is reached for, it would be easy to call it dead weight worth clearing out. On the rare day that rain arrives unexpectedly its purpose becomes obvious, and removing it to save space guarantees that the one day it was needed is the day it will not be there. Its value lies entirely in being present for the infrequent event it exists to cover.

A production buffer sits in the same position. Judged by a snapshot on an ordinary day, it looks like idle stock earning nothing. On the day one stage falls behind or breaks down, the buffer keeps the next stage running on work it has already banked, instead of stopping the instant the problem appears.

A driver who has never needed the umbrella might conclude, after a long run of dry months, that carrying it was pointless all along. A manager who clears out a buffer after a long run of trouble-free weeks makes the same mistake. A long stretch without the covered event says the weather, or the process, has been kind lately, and says nothing about whether the protection will be needed next week.

Two stages that never run at the same pace

Two stages of a process rarely run at exactly the same pace at the same moments. One machine runs slightly faster this hour and slightly slower the next, an operator takes marginally longer on one part than the one before, a delivery arrives an hour early or an hour late, and none of that ordinary variation is a failure of either stage.

A buffer, a small stock of finished or part-finished work sitting between the two stages, exists to absorb that mismatch. The faster stage can keep producing into it when the slower stage cannot immediately take what it made, and the slower stage can draw from it when the faster stage briefly falls behind. Without it, the two stages are locked together so tightly that any small variation in either immediately stops the other, since the mismatch has nowhere to go except into an idle machine or an idle operator waiting on work that has not arrived.

The theory of constraints treats a buffer placed deliberately in front of a system's constraining resource as one of its central tools for protecting output. An idle minute at the stage that sets the pace of the whole system can never be recovered, while an idle minute at a stage with spare capacity costs comparatively little, so the stock in front of the constraint earns its keep on the days it is used.

Sizing a buffer from the mismatch it covers

Treating every piece of inventory between two stages as waste to be eliminated on principle, without first asking what mismatch it absorbs, removes the buffer's protection along with whatever real excess was sitting alongside it. The two stages it used to decouple become locked together, and a small hiccup at either one now stops the other immediately.

The useful question about any stock is what specific variability it is there to cover. A buffer sized against a real, understood mismatch is doing its job even while it sits still, in the same way the umbrella in the boot is doing its job through every dry month.

The same question sets the size. Suppose the downstream stage uses one part every two minutes, and on a bad hour the upstream stage can fall up to twenty minutes behind before recovering. Ten parts in between cover that gap exactly, so a buffer of ten, perhaps with a little margin, lets the downstream stage run straight through the upstream stage's worst ordinary hour. Five parts would run out halfway through the bad stretch and stop the line on exactly the day the buffer was meant to help. Fifty would cover five bad hours back to back while tying up money and floor space, quietly becoming the unexamined waste the exercise was meant to avoid. The right size follows from measuring how far behind the slower stage typically falls and for how long, and an arbitrary round number chosen because it looked reasonable is a guess in either direction.

Stock that nobody chose

Some stock sitting between two stages has simply accumulated. Parts sized by habit, or by whatever a supplier's minimum order quantity happened to be, with no measured variability behind them, are waste, and removing them costs nothing worth protecting.

Telling the two apart requires knowing why a given buffer's size was chosen in the first place. A plant that cannot answer that question for a stock of parts between two stations has no way of knowing whether it is looking at a deliberate umbrella or an accident nobody has got around to clearing out. Writing down, for each buffer, which mismatch it covers and how its size was worked out turns that question from an argument into a quick check, and makes it obvious which stocks can be trimmed safely and which are holding the line together on its worst days.