From forecast to a feasible production plan
Follow demand into a master production plan, check uncertain supply and capacity, and communicate a feasible response.
A forecast describes expected demand. A production plan decides what the operation intends to make and when. Connecting them requires explicit assumptions about orders, inventory, materials, capacity and uncertainty.
1. Keep the planning layers distinct
Start with a demand picture at the product and time bucket used for the decision. Reconcile known customer commitments and business events without counting the same demand twice. Convert the resulting requirement into a time-phased supply response: use stock, receive supply, make product, or change the date or allocation.
A master production schedule (MPS) is a statement of planned end-item production over time. Material requirements planning (MRP) translates that schedule into component needs using bills of material, inventory records and lead times. Neither calculation proves that every supplier and work center has room; those limits need to be checked.
2. Worked example: turn a net requirement into a plan
Suppose a month has a demand forecast of 120 units and 8 units of backlog. There are 28 usable units on hand and 20 confirmed receipts due in time. A simplified requirement is 120 + 8 − 28 − 20 = 80 units to replenish.
The main supplier confirms components for 72 units. The line can build 76 units in normal time, and dispatch can move 82. The currently feasible output is therefore limited to 72 units by supply. The plan needs an explicit choice for the remaining 8: secure approved alternate material, pay for a recovery action, allocate the available units, or move the promise and track backlog.
3. Carry uncertainty into the decision
A single point forecast can hide the range of plausible demand. For a constrained product, compare a few demand scenarios or use a probability distribution if the data and planning process support it. Then evaluate which feasible plan meets the chosen service and cost priorities under those scenarios.
A 2025 peer-reviewed study examined probabilistic demand forecasts inside capacity-constrained master production scheduling across 17 datasets and 303 products. Its results support testing forecast uncertainty within the production decision; they do not prove that the same model or gains will transfer to every plant.
4. Make the plan executable
- Choose the bucket: weeks, days or shifts, aligned to planning lead times.
- Check inputs: inventory status, receipts, bills of material, yield and supplier confirmations.
- Check resources: labor, equipment, shifts, setup and outbound capacity.
- Show exceptions: name the short quantity, affected date and owner.
- Replan deliberately: change the schedule when new evidence arrives, and communicate what the change affects.
What the cited sources report
Paull, Bubak and Stuckenschmidt study probabilistic machine-learning forecasts within capacity-constrained master production scheduling. Their published article reports experiments on 17 datasets containing 303 products. MIT OpenCourseWare materials cover MRP, capacity planning and coordinated supply-chain planning as related planning layers.
How to use this in a planning conversation
The worked numbers on this page are invented for instruction. The practical lesson is to separate the demand estimate from the executable supply commitment, then expose each binding limit. Model outputs should be judged against simple baselines, operational constraints and the organization’s actual shortage and overage costs.
Sources and publication details
Source findings are summarized in our own words. Follow the links for the original work and context.
- Sebastian Paull, Alexander Bubak & Heiner Stuckenschmidt, “Machine Learning for Master Production Scheduling: Combining probabilistic forecasting with stochastic optimisation,” Expert Systems with Applications, 271, Article 126586.Peer-reviewed research article · 1 May 2025
- Introduction to Manufacturing Systems: Material Requirements Planning, MIT OpenCourseWare, 2.854.Lecture notes · Fall 2016
- Supply Chain Planning, MIT OpenCourseWare, 15.762J.Course overview · Spring 2011