SAP MRP

MRP nets demand (sales orders, forecasts, dependent demand from BOMs) against supply (stock, open orders) and, where there is a shortfall, proposes replenishment: purchase requisitions for bought materials and planned orders for manufactured ones. It plans the whole multi-level supply chain to ensure materials are available when needed, without excess.

• Demand & supply elements that MRP nets.
• Lot-sizing and safety stock.
• Planned orders → production orders / purchase requisitions.
• MRP run (classic MRP or S/4HANA MRP Live on HANA).

MRP is four steps per material, run level by level down the bill of materials, and knowing the order explains every result it produces.

Net requirements. Take the demand in each period, subtract the stock on hand and the receipts already scheduled. What remains is a shortage. Safety stock is treated as unavailable, so it raises the shortage rather than covering it.

Lot sizing. Turn the shortage into an order quantity. Lot for lot orders exactly what is short. A fixed lot size orders in multiples. Period lot sizing groups a week or a month together. The procedure sits on the material master, and leaving it at a default produces either constant tiny orders or enormous ones.

Procurement type. Decide what kind of proposal. Externally procured materials get a purchase requisition. In-house produced materials get a planned order, and the bill of materials explodes so components acquire dependent requirements. That explosion is what makes MRP a cascade rather than a calculation.

Scheduling. Work backwards from the requirement date using lead times to get a start date. If that start is in the past, MRP still creates the proposal and raises an exception message saying so.

Then it moves to the next level down and does the same for the components.

Half an hour, and it is the exercise that makes planning make sense.

1. Open MD04 for a finished material and note what is there.
2. Create a demand, either a sales order or a planned independent requirement.
3. Run MD02 for that material with a planning mode that regenerates. Read the log.
4. Return to MD04. A planned order now covers the requirement, and the available quantity returns to zero or to safety stock.
5. Open the planned order and look at its components. Each is now a dependent requirement on that component's own MD04.
6. Read the exception messages. They are the useful output: start date in the past, order proposal outside the planning horizon, excess stock. Each names a decision somebody needs to make.

The habit worth building is reading exceptions rather than proposals. The proposals are arithmetic; the exceptions are where the plan disagrees with reality.

The MRP type on the material decides whether any of the above happens at all, and picking it carelessly is the single most common planning error.

Deterministic planning nets against actual demand, which is what has been described so far. Right for anything with a bill of materials or real customer demand behind it.

Consumption-based planning ignores demand entirely and reorders when stock falls below a point. Right for cheap, fast-moving consumables where forecasting individual demand is not worth the effort. Reorder point planning is the common form, and it can be manual or automatically calculated from consumption history.

Forecast-based planning plans against a forecast derived from history rather than against orders.

No planning is a legitimate choice for materials procured directly against a project or an order.

The mistake to avoid is using reorder point planning for a component with a real bill of materials above it. Demand then arrives in lumps that the reorder point cannot anticipate, and the material is either constantly short or permanently overstocked. See the bill of materials for what drives dependent demand.

MRP ties PP to MM (purchase requisitions for components) and sales/forecasts (demand). It reads BOMs (dependent requirements) and stock, and its output drives procurement and production. In S/4HANA, MRP Live runs on HANA for far faster planning. Accurate master data (BOMs, lead times, stock) is essential for good MRP results.

Planning is also where master data quality becomes visible. Lead times, lot sizes, safety stock and scrap factors all enter the calculation, and none of them are maintained by planners after go-live unless somebody owns that. A plan nobody trusts is nearly always a master data problem wearing a planning problem's clothes. See PP configuration for the settings behind the run.

The strategy group on the material decides how demand and supply meet, and it changes the behaviour of everything above.

Make to stock plans against a forecast. Production runs to build inventory, and a sales order consumes stock that already exists. The forecast is entered as planned independent requirements, and customer orders consume them rather than adding to them, which is what stops demand being counted twice.

Make to order plans against actual orders. Nothing is produced until a customer asks, and the resulting stock is assigned to that order rather than being general. That gives you certainty and a lead time the customer has to accept.

Assemble to order sits between them: components are stocked to a forecast and the final assembly waits for the order, which is how a configurable product can be delivered quickly without holding every variant.

Consumption logic is the detail worth understanding. Forward and backward consumption periods decide how far a customer order looks for a forecast to consume. Set too narrowly, orders and forecast both stay in the plan and production builds twice what is needed.

The symptom of a wrong strategy is a plan that looks reasonable and produces the wrong inventory.

How MRP is executed matters as much as how it calculates.

Regenerative planning replans everything in scope, ignoring what was planned before. Thorough and slow, and it is the right choice after a significant master data change.

Net change planning only replans materials with a relevant change since the last run, which is what makes a nightly run over a large catalogue feasible. It depends on the planning file being accurate, and a corrupted planning file produces materials that are silently never replanned.

Net change in the planning horizon narrows further, replanning only within a defined window.

The planning horizon itself has to be longer than the longest cumulative lead time, or proposals appear too late to act on. That is a calculation rather than a preference, and it is frequently set to a round number that happens to be shorter than the supply chain.

Frequency is the other decision. Nightly is normal. More often gives responsiveness and produces plan churn, where a planner acts on a proposal that has disappeared by the afternoon.

• Bad master data (BOMs, lead times), wrong plans.
• Ignoring safety stock/lot-sizing settings.
• Not converting planned orders, plans not executed.

• Ignoring exception messages. They are the part of the output that needs a human, and a planner who only reads proposals is doing the computer's job rather than their own.
• Running MRP without a planning horizon that matches lead times. Proposals appear too late to act on.
• Safety stock used to hide unreliable lead times. See capacity and confirmations for what happens after the plan.

Running MRP is one transaction, and setting the MRP types, lot sizes and horizons so the plan is executable is the part you do in the course.