Supplying the mix: planning the factory, suppliers and subcontractors for ETO and CTO manufacturers

Part 3 of 3: Planning what you haven’t designed yet

Part 2 ended with a forecast of each model and its expected option mix. How does that become a plan for the factory, material suppliers, subcontractors and distribution, for products that aren’t defined yet, and stay executable as orders arrive? The short answer: attribute-based planning replaces thousands of fixed BOMs with generic parts. The specifications on each demand choose the BOM and route options, and the plan is optimized against real capacity and materials.

Why supply planning is harder in ETO and CTO

The combinations that make forecasting hard explode again on the supply side, for both orders and forecasts. Orders must be built exactly as the customer configured them. Traditional systems handle that by creating an order-specific BOM for every order, which multiplies master data and maintenance with every sale.

Forecasts follow a different logic. They have to drive the statistical demand for end-item options, intermediates and materials, and that demand often follows complex rules about which options depend on each other. In the middle horizon, both exist at once, tied together by forecast netting rules. And the traditional supply problems don’t go away: finite capacity, supplier lead times and limits, and load that shifts with the mix.

The simplification: generic parts with BOM and route options

With attribute-based planning, each model has one generic BOM and route, with options in place of thousands of fixed structures. The specifications on each demand choose the options. For a configured order, those are the attributes the customer selected. For a forecast, they are the statistical breakout of the expected mix. No order-specific BOMs are needed, and the same structure serves orders and forecasts alike.

For the forklift from Part 1, choosing the Type2 engine brings its own components and its longer test step. Whether that choice comes from a firm order or from a forecast that expects 40% Type2, the plan handles it the same way.

Capacity and materials from the mix

Because each option carries its own route, the load on every resource follows the mix, not just the unit count. A shift toward Type2 engines shows up as extra load on final test before it becomes a bottleneck. Intermediates and long-lead materials are planned from the same statistical breakouts. The plan is built against real limits, including subcontractor and supplier capacity, at a high level for S&OP and in detail for S&OE.

The middle horizon: orders and forecast together

As orders arrive, they net against the forecast, but not combination by combination. Matching each order to an exact forecast configuration is where traditional netting breaks down: there are too many possible combinations to match accurately. Instead, netting works in two parts. The generic model is netted first, so each forklift ordered consumes one forklift of forecast. Then the options are netted differently, to avoid the downside of combinatorial explosion. For example, a Type2 engine on an order consumes Type2 engine forecast, whatever cabin or chassis it came with. The netting considers option dependencies where they are real, but does not create artificial ones.

The rule is simple: forecast reserves, orders execute. The remaining forecast holds capacity and buffer materials but never releases work orders. Configured orders generate the exact work orders and material requests for the product the customer bought. The company’s own lead times set where one horizon ends and the next begins.

Committing ahead and quoting reliable dates

S&OP commits what can’t wait for an order: long-lead purchases, sub-assembly builds, in-house and subcontract capacity, and the buffer targets set in demand planning. Quoted dates check both material and capacity instead of relying on standard lead times.

Handling the three horizons of ETO and CTO planning means S&OP and S&OE need to run in one system on one model, so there is no hand-off to reconcile between the long-term, mid-term and short-term plans.

A planning model built on attributes also makes agentic AI more effective. Plans are easier to explain and update, because they are described in the terms the business uses, such as engine type or chassis, rather than thousands of part numbers. Instead of sifting through part-level detail, AI agents can spot patterns sooner, such as a shift in the option mix, and propose changes for a planner to approve.

The message of this series is simple: ETO and CTO businesses can’t be planned well in a system built for make-to-stock. Plan the model and the mix, and let the specifications drive supply. That is the approach Eyelit SIOP and Agent EyeQ from Eyelit Technologies are built on. We’d be glad to start with a working session to map your models and attributes.

About the authorWilliam Green