The Master Production Schedule, Explained

What is a master production schedule?

What is the difference between an MPS and a production plan?

How is projected available balance calculated?

What are time fences in master scheduling?

Where does the MPS sit in the planning hierarchy?

How do you read an MPS grid?

What are time fences and why do they matter?

How do you build and maintain an MPS?

What goes wrong with master schedules?

What do the standards say?

Where Harmony AI fits

Q: What is available-to-promise in an MPS?

Q: What does an MPS look like in make-to-order plants?

A master production schedule is the statement of which specific end items a plant will produce, in what quantities, in which time buckets, usually weekly, over a rolling horizon. It disaggregates the sales and operations plan into buildable numbers, drives MRP's material calculations, and provides the available-to-promise figures that order promising depends on.

The production plan is aggregate: product families, monthly volumes, workforce levels, set through S&OP. The MPS is specific: individual end items in weekly buckets, reconciled so its total matches the approved plan. The production plan decides overall rates; the MPS decides exactly what gets built when.

PAB for each week equals the prior balance plus any MPS receipt minus that week's demand, booked orders in the near term, and the greater of forecast or orders further out. When the arithmetic would go negative, the master scheduler places a new MPS lot in that bucket, sized by the item's lot-sizing rule.

Time fences divide the MPS horizon into zones of firmness. Inside the demand time fence the schedule is frozen and changes require management sign-off. Between the demand and planning fences it is slushy: items can be traded within available capacity and material. Beyond the planning fence it is liquid and replans freely. Fences prevent schedule churn from cascading through MRP.

ATP is the portion of on-hand inventory and scheduled MPS lots not yet consumed by booked customer orders. In the first bucket it is inventory minus orders due before the first receipt; at each MPS receipt it is that lot minus orders booked until the next one. Sales promises new orders against ATP so the plant never promises the same units twice.

Make-to-order and assemble-to-order plants do not master-schedule finished items, because variety is too high to forecast at that level. Assemble-to-order plants master modules and options using planning bills and finish products to order; make-to-order shops master raw or semi-finished materials. The rule of thumb is to master-schedule at the level just before variety explodes.

How the MPS grid actually works: projected available balance, available-to-promise, time fences, and the weekly cycle that keeps it honest.

A master production schedule (MPS) states exactly what end items a plant will produce, in what quantities, in which weekly time buckets, over a rolling horizon. It converts the aggregate production plan into buildable numbers, drives MRP, and is the anchor for every delivery promise the business makes.

We introduced the concept in our master production schedule overview. This post goes deeper into the working mechanics: the MPS grid and how projected available balance and available-to-promise are computed, how time fences keep the schedule stable, and how the MPS connects planning above it to execution below it. If you can read an MPS grid, you can tell in thirty seconds whether a plant's promises are real.

The MPS is the middle layer: above it, sales and operations planning sets aggregate volumes by product family; below it, MRP explodes the MPS into component and material requirements. Production planning answers "how much of each family, per month, with what workforce." The MPS answers "which specific end items, how many, in which week." MRP then answers "what parts and materials must be ordered when." Detailed shop scheduling, sequencing actual work centers hour by hour, hangs off the bottom of the stack.

Each layer has a matching capacity check. The production plan gets resource planning, the MPS gets rough-cut capacity planning against critical resources, and MRP output gets capacity requirements planning. An MPS that never gets a rough-cut check is a wish list, and the ways that goes wrong are covered in our companion piece on capacity-based production scheduling.

The MPS translates family-level plans into end-item weekly numbers and drives MRP. Every layer pairs with its own capacity check.

The MPS lives in a grid: one row set per end item, one column per weekly bucket, with five core rows. Forecast and customer orders capture demand. Projected available balance (PAB) is the running inventory arithmetic. The MPS row is the schedule itself, the planned production receipts. Available-to-promise (ATP) is the portion of inventory and scheduled production not yet consumed by booked orders, the number sales can promise against. Here is a worked example for one item with a starting inventory of 40 and a lot size of 50:

PAB subtracts the greater of forecast or orders each week and adds MPS receipts; a new lot is scheduled whenever PAB would go negative. ATP nets booked orders against each MPS lot until the next receipt.

Two calculations carry the weight. PAB each week is the prior balance plus any MPS receipt minus that week's demand (near-term, booked orders; further out, the greater of forecast or orders). When PAB would go negative, the master scheduler plans a new lot, sized by the item's lot-sizing rule, the same logic families described in lot sizing in MRP. ATP in week one is inventory minus orders due before the first receipt; at each MPS receipt it is the lot minus orders booked until the next receipt. Order promising against ATP is its own discipline, covered in available-to-promise.

What gets mastered also depends on the environment. Make-to-stock plants schedule finished items. Assemble-to-order plants master modules and options using planning bills, and finish to order. Make-to-order shops master raw or semi-finished levels. The choice follows where variety explodes, a topic that connects directly to make-to-stock vs make-to-order.

Time fences divide the MPS horizon into zones of decreasing firmness so the schedule stays stable where stability is expensive and flexible where flexibility is cheap. Inside the demand time fence, typically the near days or weeks within final assembly lead time, the schedule is frozen: materials are committed, capacity is committed, and changes require management sign-off because they cost real money. Between the demand fence and the planning time fence, the zone is often called slushy: the master scheduler can trade one item for another within available capacity and material. Beyond the planning fence, the schedule is liquid and the planning system can replan freely.

Fences make schedule stability explicit. Changes inside the frozen zone cost the most, so they require a human decision, not a system replan.

Fences are what keep the MPS from thrashing MRP. Every MPS change explodes into dozens of component-level changes, so an unstable master schedule multiplies into hundreds of expedites and de-expedites downstream, the nervousness problem. The fences damp it at the source. The same stability logic, applied at shop-floor timescale, shows up in dynamic production scheduling as the frozen window.

The working cycle, run weekly by most plants:

Disaggregate the production plan. Break family volumes into end items using demand mix, so the MPS total reconciles with what S&OP approved.
Load demand. Bring in forecast and booked orders per bucket, netting per your consumption rules so demand is not double-counted.
Compute PAB and place lots. Run the balance arithmetic; schedule receipts where PAB goes negative, sized by lot rules.
Rough-cut the critical resources. Convert the draft MPS into hours on the few resources that break first; move lots between weeks until the load fits.
Publish ATP. Recompute available-to-promise so sales promises track the new schedule.
Manage exceptions inside the fences. Handle change requests per zone: liquid changes flow through, slushy changes get traded, frozen changes get escalated with their cost attached.
Measure and close the loop. Track MPS performance, planned versus produced per bucket, and feed the misses back into lot sizes, fences, and capacity data.

Step seven is where the MPS earns or loses its authority. If the plant ships 80% of what the schedule said, every downstream system is planning from fiction, and the fix starts with measuring schedule attainment honestly. A simple structured pass with our free production schedule builder is a reasonable way to prototype the grid before formalizing it in planning software.

The classic failure is the overstated MPS: the schedule carries more volume than the plant has ever demonstrated, usually because someone treats it as a target rather than a plan. The damage is systemic. MRP believes the schedule, so it orders material for output that will not happen; inventory climbs while service falls, because the material arrives but the hours do not exist to convert it. Meanwhile expediting decides what actually ships, which means the real schedule is the hot list and the official one is theater. The fix is unglamorous: load the MPS to demonstrated capacity, and let the rough-cut check kill overloads before they are published.

Three other patterns show up constantly. Double-counted demand, forecast plus the orders that consume it, inflates the schedule quietly; consumption rules exist precisely to prevent it. Fence-jumping, where every customer request breaks into the frozen zone, destroys the stability the fences were built to buy, and it usually traces to nobody attaching a cost to the change. And stale planning data, lead times, lot sizes, and rates that have not been re-verified in years, makes every calculation precise and wrong. Each of these is visible in the numbers if the plant compares planned to actual per bucket and asks why. A master schedule that is measured weekly gets better; one that is never scored just gets longer.

Source material worth having on the shelf:

The ASCM/APICS body of knowledge defines the MPS, time fences, projected available balance, and available-to-promise, and treats master scheduling as its own discipline with certification tracks (CPIM).
The MPS-drives-MRP architecture comes from the MRP and MRP II tradition of Orlicky and Wight; the MRP overview summarizes the history and the dependent-demand logic underneath it.
Practice ranges, not rules: frozen zones commonly span final assembly lead time (days to a few weeks), planning fences commonly sit at or beyond cumulative lead time, and most plants re-run the full MPS cycle weekly with daily exception handling in between.

Customer orders

MPS receipt

Projected available (start 40)

Available-to-promise

The MPS runs on numbers the floor generates: what was actually produced per bucket, real rates, real changeover losses, real scrap. In many plants those actuals arrive late through end-of-shift paperwork and manual entry, so the master scheduler plans next week from data that misstates last week. Harmony AI is an AI-native MES, a real-time operational layer that connects machines, existing planning software, and paperwork into one live record with no rip-and-replace, giving the MPS honest inputs: demonstrated output by item and line, as it happens, plus AI agents that flag a bucket tracking short while there is still a week to react instead of a variance report after the month closes. Deployment is white glove and in person, Harmony AI's team maps how production actuals flow today, paper and all, and wires them into one source of truth alongside your planning stack. The CLS case study shows the shift from binders to a live record, and our ROI calculators and tools can put a number on what stale actuals cost.