One-piece flow: removing the queue between operations and living with what that exposes

The queue is doing work you may not want done

Material waiting between operations decouples them. It lets an unreliable machine stop without starving the next station, it lets operations of different lengths coexist, and it lets a defective batch continue quietly until somebody downstream notices. Removing it converts each of those into a visible interruption. The gain is speed and feedback, since a fault is detected at the next operation rather than a shift later, and a unit reaches the end in a fraction of the elapsed time. The price is that everything has to work, which is why the first weeks of a genuine flow line are unpleasant.

Stability upstream is a precondition, not an outcome

Teams are told that flow exposes problems, and conclude they should impose flow and let the problems surface. That is sound where the problems are small and somebody is standing by to fix them daily. It is destructive where equipment stops constantly, material arrives short, or changeovers consume a large part of a shift, because the line then fails so often that the organisation decides the concept does not suit its products. Sequence the work: make the equipment reliable, get the material supply right, shorten the changeovers, and only then take the buffer out.

You are choosing to lose apparent utilisation

In a flow arrangement the slowest operation sets the pace and everybody else waits. Machines that used to run continuously now stand idle, sometimes conspicuously, and any measurement system built on equipment or labour utilisation will report a decline. Unless leadership has decided in advance that this is acceptable and has changed what supervisors are judged on, pressure builds to give the idle machines something to do, and the buffers reappear within a month. This is a management-system decision wearing the costume of a layout question.

Where a single unit is genuinely impossible

Ovens, autoclaves, plating baths, curing, heat treatment, presses with expensive tool changes and shared equipment serving several families all impose a minimum practical quantity. The honest response is small-lot flow rather than pretence: run the smallest batch the physics and economics permit, keep it moving as a unit, and put the constraint somewhere it can be seen instead of smoothing it out with stock everywhere. Sometimes the right answer is to change the process, since a smaller oven used more often can beat a large one filled once a shift even at a worse energy figure per unit.

Checking that flow improved rather than relocated

The common outcome of a cell project is that the queue moved to the cell entrance. Inside, work moves beautifully; outside, a pallet of parts waits for the cell to be free, and elapsed time from order to despatch is unchanged. Measure the whole path rather than the cell: units of work in progress across the route, elapsed time from release to completion, and the spread of that time rather than its average. Where none of those has moved, the improvement is local and the plant has bought tidiness at the price of a rearrangement.

Frequently asked questions

Does this mean holding no work in progress at all?

No. There is always the unit being worked on at each station, plus whatever the process physically requires, plus deliberate buffers where the design calls for them, typically ahead of a constraint or before despatch. The distinction is between inventory placed on purpose, with a defined quantity and somebody watching it, and inventory that accumulated because operations were left to run at their own convenience. Only the second kind is what this approach removes.

Can we run flow where machines are shared between product families?

It gets harder and is often still worthwhile. The shared resource becomes a scheduling problem for the whole set of cells, and the practical answers are dedicating equipment where volumes justify it, shortening the setup so the machine can serve several families within a day, or accepting a controlled buffer at that step. What does not work is pretending the sharing is invisible, because the cells depending on it will be starved unpredictably and their operators will build hidden stock.

Is this relevant at low volume and high variety?

The principle transfers and the mechanics change. Where every job is different, flow means moving a job through consecutive operations without letting it sit in a queue between them, achieved by controlling how much work is released rather than by physically linking stations. The gains are similar and frequently larger, because in job shops the elapsed time is dominated by waiting rather than by processing.