Configure-to-order: selling from a rule set the factory can honour

The rule set is the thing the factory has to honour

A configurator is not a sales convenience. It is the contract between what may be sold and what can be made. Behind the option menu sit rules that generate a parts list and a routing for each valid selection, plus constraints that block combinations which are physically impossible, uncertifiable or unpriceable. Committing to this model therefore means committing to maintain that logic as products evolve, which is continuing engineering work rather than a project with a completion date. When the rules drift out of step with the shop, the symptom is unmistakable: orders arrive that assembly cannot build, and engineers quietly repair them one at a time.

Buyers who want choice without commissioning a design

This fits customers who need something specific but not novel, and who will trade unlimited freedom for a shorter and more certain delivery: machinery with duty options, commercial vehicles, laboratory and medical equipment, windows and doors, cabinetry, industrial enclosures. For purchasing the consequence is that demand for any single option is far less predictable than demand for the family it belongs to, so buyers plan from option take rates rather than unit forecasts. Those rates drift as the mix moves and need re-examining. The buying approach usually separates options appearing on most orders, which can be committed confidently, from rare ones bought against the order.

Stock held against take rates, tooling held against every option

Inventory concentrates in option-specific components, so the cost of the model tracks the width of the option list rather than the volume behind it. Each rarely chosen option still needs a part held, a supplier maintained and space allocated. Capital behaves the same way: fixtures, tooling and test equipment must exist for every option the catalogue offers whether it sells or not, and maintaining them is a running cost long after the tooling itself is paid for. Plants that never audit slow-moving options accumulate a tail that consumes real capital, occupies engineering attention and returns very little.

Units built wrong on purpose, and the inspection that cannot see it

The most common defect here is not a manufacturing defect. It is a unit built exactly as instructed to a configuration that was wrong at order entry, and no dimensional check will find it because every feature conforms. Detection falls to configuration verification: confirming that what was built matches what was ordered, and that the ordered combination was valid to begin with. Failures cluster in three places, namely where sales can override the configurator, where options are captured as free text, and where a special is accepted once and then quietly repeats. Each shortcut looks reasonable alone and is corrosive in aggregate.

Suppliers serving a catalogue rather than a part number

Upstream, vendors are asked to support a range with uneven and unpredictable draw, which is harder for them than a steady part. Some will hold stock against expected take rates; others quote a lead time that only holds for the popular selections. The growth property is attractive while the catalogue stays stable, because rising volume improves purchasing strength on common items without adding complexity. It reverses once the option list expands faster than sales. Past that point each addition dilutes the volume behind every other option, weakens the position with suppliers, and drags the plant toward the economics it adopted this model to avoid.

Frequently asked questions

What separates configure-to-order from assemble-to-order in practice?

They answer different questions. Assemble-to-order is about where inventory sits, namely in modules held ahead of a short final build. Configure-to-order is about how choice is defined and validated before work is released. Plenty of plants run both, holding modules and selling from a rule set, but each can exist alone: you can configure a product that is then fabricated from raw material, and you can assemble modules to a fixed catalogue with no configurator anywhere in the process.

Who should own the configuration rules?

Engineering should own validity and manufacturability, commercial should own price and availability, and one named person should own the whole rule set as a maintained asset. Ownership split between departments with no single custodian is the usual reason rules rot. Practical governance means a change to any option follows the same release discipline as a drawing change, with a record of what altered and when, so a build problem months later can be traced to the rule that permitted it.

What does a configurator need before the shop can trust its output?

Three things. Generated parts lists that match how the plant actually builds, not an idealised structure inherited from the catalogue. Routings that include the operations specific to each option, so scheduling reflects real work content. And a validation layer that rejects impossible combinations at entry rather than flagging them later. Without the last of these, the configurator merely moves the problem downstream, where it lands on whoever is holding the job when the shortage appears.