Engineer-to-order: when design hours are part of the cost of goods

Signing up to deliver something nobody has drawn yet

An engineer-to-order commitment binds the factory to deliver a result rather than a part. Design, verification and manufacture all sit inside one order, and the buyer is purchasing an outcome that has not been proven anywhere else. Quality assurance therefore begins in engineering rather than on the shop floor: requirement capture, design review, and a documented trail showing that what was built matches what was specified and approved. Acceptance is usually witnessed, often at the customer site, and the definition of conforming must be agreed in writing before fabrication starts. Arguments about what good looks like, held after the unit is welded together, are almost never won.

The buyers this fits, and why the model resists volume

Customers here are usually buying capability for a site they own: process plant, handling equipment, tooling, test rigs, specialised vehicles, building-integrated systems. Demand is lumpy by nature because it follows their capital cycles rather than their consumption, so an order book can look full and then empty within a season. That lumpiness is exactly why the model resists conventional scaling. There is no same thing to run faster. Growth means more engineers and more project managers, both slow to recruit and slower to become productive, so a firm that chases volume without adding design capacity simply queues work in front of the drawing office and calls the delay a production problem.

Purchasing against a bill of materials still being written

Buying starts before the design is finished, which is the uncomfortable reality of this model. Castings, large fabrications, drives and specialised instruments have to be committed while the detail around them is still moving, so buyers work from preliminary information and accept some rework risk to protect the delivery date. That argues for pulling suppliers into the design conversation on the few items that set the critical path, and for terms that allow secondary details to be defined late. The supply base tends to be narrow and technically capable rather than broad and inexpensive, because a vendor who reads a specification and queries it is worth more than one who quotes quickly.

One job, one record, and the cash trapped inside it

Everything is tracked against the job rather than the item. Hours, drawings, revisions, purchased material, inspection records and change requests all belong to a single order, and the systems must show committed cost against estimate while work is still running, not after the final invoice. Inventory is almost entirely work in progress, and it can be very large relative to turnover, because one unit may absorb months of material and labour before anything leaves the building. Payment milestones tied to defined events, and disciplined recording of variations as they happen, therefore do more for the cash position than any negotiated purchasing saving.

Skilled hours as the binding constraint, and how a job unravels

Capacity is counted in skilled hours, not machine hours, and the constraint is usually engineering rather than fabrication. Equipment tends to be general purpose and long-lived: cutting, welding, machining and lifting capability sized for large items, plus assembly and test space that is costly to provide and hard to fill densely. Jobs come apart in a familiar sequence. The estimate rests on optimistic design-hour assumptions, scope creeps through informal customer conversations nobody records, late changes collide with material already bought, and recovery is attempted through overtime that raises cost without moving the date. Every one of those steps is visible weeks before the overrun is admitted.

Frequently asked questions

Why do engineer-to-order jobs so regularly overrun their estimate?

Because the estimate is written when least is known and rarely revisited when more becomes known. Design hours are the usual culprit, since novel work refuses to behave like the previous job it was benchmarked against. The second cause is uncontrolled scope: small accommodations agreed verbally with a customer engineer, none of them individually significant, none of them priced. Firms that control this compare committed cost to estimate weekly during the job, not at handover, and force every variation through a written change note.

How should a change request be handled once fabrication has started?

Treat it as a commercial event before treating it as a technical one. Establish what is already cut, welded or purchased, what has to be scrapped, and what the delivery consequence is, then put price and date in front of the customer before any drawing is revised. The failure pattern is engineering absorbing the change quietly to keep the relationship comfortable, which converts a chargeable variation into an unrecoverable cost and, worse, sets an expectation that the next change is free too.

Can an engineer-to-order business ever become more repeatable?

Partly, and the gains come from structure rather than from standardising the product. Build a library of proven subassemblies, calculation methods and supplier packages that can be reused across dissimilar jobs, so novelty is confined to what genuinely differs. Standardise the process even where the product cannot be standardised: consistent design review gates, a fixed estimating method, one way of releasing work to the shop. Most firms that move toward configure-to-order started by discovering how much of their bespoke work was quietly identical.