Work in progress control: keeping the floor from filling up with unfinished work

Capping what enters rather than chasing what is already there

The amount of unfinished work on a floor is governed almost entirely by the rate at which jobs are released compared with the rate the plant completes them. Expediting inside an overloaded floor moves the problem around without reducing it. The effective control is a cap on released work, whether expressed as a number of open jobs per area, a fixed quantity in each queue, or release tied to completion. Plants find this counter-intuitive because holding material back looks like idleness in the first operations, so the cap needs explicit management support before supervisors will hold the line.

Finding the places unfinished work hides

Unfinished stock rarely accumulates where the system says it is. Look behind machines, in the rework corner, on the mezzanine, in the area holding parts awaiting an engineering decision, on pallets marked with a name and no job number, and in whatever building the plant uses for things nobody wants to scrap. A physical walk with a supervisor and a printed list of open jobs finds more in a morning than any report. Do it after a period of high demand, when the accumulation is worst, and photograph the locations, because the same corners refill within months unless someone owns them.

Ageing analysis turns a pile into a decision list

Sort open jobs by how long they have been open, and the tail is immediately informative. Old jobs are usually blocked rather than slow: waiting for a component that was never ordered, for a decision on a deviation, for a customer to confirm a change, or for a tool that is broken. Each needs a decision, and most of the decisions are unattractive, which is why they were avoided. Run the list at a regular meeting with someone who can authorise scrapping, reworking or closing, and give each aged job a named owner and a date. Without that authority the review becomes a monthly recital of the same jobs.

Keeping the recorded quantity close to the physical one

Control depends on the numbers being roughly true, and shop floor records drift for ordinary reasons: partial completions unrecorded, scrap not booked, jobs split without a transaction, material borrowed from one job for another. The correction is not more paperwork but fewer, clearer transaction points, plus regular counting of the areas that matter most. Where a discrepancy is found, investigate the cause rather than simply adjusting the record, because the adjustment removes the evidence. Accountants should be part of this: unfinished stock carried at a value the plant cannot physically locate becomes a problem at the year end.

The cost that is not on the stock report

The obvious cost of unfinished work is cash tied up. The costs that hurt more are indirect: longer elapsed time through the plant because everything queues behind everything else, damage and deterioration in storage, obsolescence when a design changes and a large amount of part-built product has to be reworked, and the sheer difficulty of finding anything. A useful argument for the board is not the stock value but the delivery lead time, since customers experience the queue directly and competitors with a leaner floor quote shorter dates from the same equipment.

Frequently asked questions

What is a reasonable level of work in progress for our plant?

There is no universal figure, because it depends on how many operations a product passes through and how variable each one is. A workable approach is to compare elapsed time through the plant with the total processing time on the routing. When elapsed time is many times processing time, most of the material on the floor is waiting rather than being worked, and the level can be reduced without losing output. Track the ratio over time instead of chasing an absolute target.

Will capping released work reduce our output?

Not if the cap is set above what the limiting resource can consume. Output is governed by that resource, not by how much material is on the floor, so releasing more simply lengthens queues. Set the cap by observing current queue levels and reducing gradually, watching output at the constraint as you go. If output starts to fall, the cap has gone below the level needed to keep that resource fed and should be raised slightly.

Who should authorise scrapping aged work in progress?

Someone senior enough to accept the charge without needing to defend it as a personal failure, typically the operations or plant manager with finance informed. Delegating it too low means nothing gets scrapped, because the individual carries the blame for a loss that usually originated elsewhere. Set a routine review with a standing agenda item, so the decision is normal rather than exceptional, and record the cause so the same category of stranded work can be prevented.