Production levelling: deciding where demand variation is going to be absorbed

Variation does not disappear, it is relocated

A plant that builds exactly what was ordered each day looks responsive and pays for it in overtime, changeovers, expediting, quality problems during the peaks and idle capacity in the troughs. Levelling relocates that variation into an explicit buffer: finished stock for repeat items, a short order backlog for made-to-order items, or a quoted lead time long enough to smooth the arrivals. Each option has an owner and a cost, and somebody has to agree to carry it. Arguments about levelling that stay inside operations never resolve, because the trade being made sits with the commercial side of the business.

Mix levelling is harder and matters more

Holding total volume steady while running a month of one product and then a month of the next still forces suppliers and downstream processes into a lumpy pattern, and still leaves the plant unable to respond to a mix change without a large stock position. Levelling the mix, so a repeating pattern contains each significant product regularly, stabilises component consumption, shortens the response to a mix change and lets upstream operations settle into a rhythm. It also demands many more changeovers, which is why most plants level volume and stop at that point.

The changeover bill arrives immediately

Running a repeating mix means changing over often, and the arithmetic is brutal where setups are long, since capacity disappears into setup time and the plant misses the schedule it just committed to. Levelling and setup reduction are therefore the same project in practice, and attempting one without the other produces a quick, well-documented failure. Sequence it honestly: establish what the changeover time would have to be to support the pattern you want, decide whether that is reachable, and level to the pattern the shop can actually run today while the setup work proceeds.

What levelling asks of planning and sales

The pattern only holds if orders stop being injected into the current period. That means a frozen window inside which the sequence does not change, a rule covering genuine emergencies and who may authorise one, and acceptance that a customer telephoning at noon does not get their order moved to the front. Sales has to be able to quote from that reality and be rewarded for doing so. Plants that install a levelling board while keeping an unlimited expediting channel end up with a board describing yesterday's intentions and a floor that ignores it.

Where levelling is the wrong answer

Genuinely bespoke work, project work, products with very short life cycles and businesses whose demand is dominated by a few large infrequent orders do not level well, because there is no repeating pattern to level to and the stock buffer would be unsellable. Sharp but predictable seasonality is a different case, where the real choice lies between building ahead into stock and flexing capacity, and should be made explicitly with the cost of each on the table. Applying levelling mechanically to an engineering-to-order shop produces a plan the plant quietly ignores.

Frequently asked questions

How can we level when our demand is genuinely lumpy?

Level the production pattern rather than the demand. That needs a buffer between the customer and the plant, sized according to how lumpy the arrivals are and how long the plant takes to react. For repeat products the buffer is finished stock; for made-to-order products it is a controlled backlog and a quoted lead time. If neither is acceptable to the business, the plant must carry the variation in capacity, and the honest conversation is about what that costs.

Does levelling mean carrying more finished stock?

Often more of the levelled items and less inventory overall, because the work in progress, safety stock and component buffers that existed to cope with an erratic schedule can come down. Whether the trade is favourable depends on the value and shelf life of the finished item against the value of its components and the cost of disruption. High-value or perishable finished goods argue for absorbing the variation in lead time or backlog instead.

Can a process plant with long campaigns be levelled?

To a limited degree, and the binding constraint is usually cleaning or sequence-dependent transition between grades. The practical work is shortening the transition, finding a grade sequence that minimises the wash-out penalty, and then running the shortest campaign the economics allow rather than the longest the vessel will hold. The direction of travel matches discrete manufacturing; the step size is set by process chemistry rather than by tooling.