Make-to-order: turning a confirmed order into a production slot

Which products earn the right to be built on demand

The model suits products where variety is genuinely wide, where a unit has value only to the buyer who specified it, or where the item is costly enough that holding one speculatively would be reckless. Industrial equipment, specialised assemblies, converted and printed goods and most bespoke fabrication sit here. Accepting the model means the plant holds open capacity rather than open stock. Every quote is a claim about a future slot on a machine, and every accepted order consumes one. That makes the order book, not the warehouse, the thing management reads daily, and it drags the sales team into the scheduling conversation whether operations wants them there or not.

Machines that tolerate a mix, and a schedule anyone can believe

Assets have to cope with a changing mix rather than a fixed one, so general-purpose machines, quick-change tooling and operators who can run more than one process are worth more than raw throughput. The capital bill is usually lighter than a dedicated line while utilisation is worse, and that exchange is the whole point. Software carries more weight here than in any high-volume model. The plant needs routings and standard times good enough to quote from, finite scheduling that respects real constraints instead of assuming infinite capacity, and job-level visibility against the promised date. A planner fed optimistic times will publish a schedule that has never once been achieved.

Cash sitting inside a job rather than on a shelf

Finished goods barely exist, because an item is shipped rather than stored. The weight shifts into work in progress, where cash sits while a job crawls through the shop, and into whatever raw material must be held because its replenishment time exceeds the delivery the customer accepted. That last point is the procurement question the model turns on. Buying strictly to the order is clean but slow. Stocking common material shortens the quoted lead time and quietly reintroduces some of the risk this model was meant to remove. The workable split is by commonality: hold what many jobs consume, buy specifically what only one job will.

Getting it right first time, with a supply base that keeps pace

Rework is disproportionately painful because no substitute unit is waiting in the warehouse. A scrapped part is a missed delivery date, full stop. That pushes quality effort upstream into first-article verification, drawing and specification review before release, and in-process checks rather than a final inspection that finds the problem too late to matter. Suppliers inherit the same logic. A vendor whose parts are broadly acceptable but occasionally late or marginally out of tolerance does far more damage here than in a stocked plant, because their variability lands straight on a customer commitment. A short list of capable suppliers with agreed response times usually beats a wide panel bought on price.

Growth by added capacity, and the day promised dates start slipping

Scaling means adding capacity and the people who can run it, and that works until the shop is congested enough that queue time rather than machine time sets the lead time. Past that point every additional order stretches the quoted date for all the others, so revenue grows while delivery performance decays. The recognisable failure sequence starts with over-promising to win work, moves to expediting individual jobs past everything else, and ends with a schedule nobody trusts and a shop run by whoever shouts loudest. The defensible remedy is unpopular: quote from measured capacity, and be willing to lose an order rather than accept a date you cannot hold.

Frequently asked questions

How do you quote a delivery date the shop can actually hold?

Start from measured capacity rather than theoretical capacity, and from routing times that reflect what the shop really achieves including setup, inspection and the waiting between operations. Then check the material path, since a purchased item with a long replenishment time will set the date regardless of how free the machines look. Finally, quote against the committed order book, not against an empty calendar. Most broken promises come from quoting each job as though it were the only one in the building.

Does make-to-order remove inventory risk?

It removes most finished-goods risk and replaces it with two others. Work in progress rises, because value is tied up in partly built jobs that cannot be sold to anyone else if the customer cancels or changes the specification. And any material you stock to shorten lead times carries the same obsolescence exposure a stocked plant has, only in smaller quantities. Progress payments and clear change-order terms matter more in this model than in almost any other, because they decide who funds a job in flight.

When should a make-to-order plant start stocking components?

When the customer will not wait as long as your supplier takes, and only for items shared across many jobs. Rank purchased parts by two attributes: how many different orders consume them, and how long they take to arrive. Anything high on both is a candidate for stock, because the holding risk is spread across future demand you are confident of. Single-job specials never qualify, however tempting the price break for buying a larger quantity looks at the time.