MRP vs demand forecasting: bu ne anlama geliyor
These two are often conflated, because both appear to answer “how much will we need”. They answer different questions, and the order matters.
Demand forecasting: independent demand
Demand forecasting predicts independent demand — the finished goods a customer asks for directly. It is a statistical estimate: it derives the future from past sales, seasonality, trend and promotional effect. It carries uncertainty and is properly a probability distribution.
MRP: dependent demand
MRP (Material Requirements Planning) calculates dependent demand. It is arithmetic, not prediction: if you will build 1,000 finished units and the bill of materials specifies four screws each, you need 4,000 screws. There is nothing to forecast.
MRP takes three inputs: the master production schedule, the bill of materials, and current stock. Its outputs are purchase and production orders.
The chain
Demand forecast → Master Production Schedule → MRP → Purchase and production orders
Forecasting sits at the head of the chain. MRP consumes its output.
The cost of confusing them
The most common error is treating MRP output as if it were a forecast. MRP produces exact numbers — 4,000 screws — and that precision is misleading: its input is an uncertain forecast, so its output cannot be certain.
If the demand forecast is 20% wrong, MRP multiplies that error down through the bill of materials. At component level the error grows rather than shrinks.
A second error is placing safety stock at the wrong layer. Uncertainty originates in finished goods; holding safety stock at component level buys the same protection far more expensively.
MRP’s silent assumptions
Because MRP is a calculation it looks exact, but it rests on three assumptions that break routinely:
Fixed lead time. MRP uses one lead time per component. In reality that is a distribution, and MRP cannot represent one.
Infinite capacity. Classic MRP ignores production capacity; it computes requirements and assumes capacity will be found. Capacity is checked afterwards, in a separate step — and conflicts found there send the whole plan back.
Accurate bills of materials. If an engineering change has not been processed, MRP orders the wrong component in exactly the right quantity.
Nervousness: small change, large wave
MRP has a well-known behaviour: a small change in the master schedule produces large, oscillating order changes at lower levels. This is “MRP nervousness”.
The cause is that lot-sizing rules behave as thresholds: 99 units of requirement is one lot, 101 is two. When the forecast shifts slightly every week, purchase orders are reissued every week and supplier trust erodes.
The usual remedy is to freeze part of the horizon: the first 4-6 weeks are fixed, beyond that is open.
The practical conclusion
Do not use MRP as a forecasting tool. It is a calculator and it is only as good as its inputs. Invest in the quality of those inputs — forecast accuracy and master data hygiene.
Both can be wrong at once
MRP can run correctly and produce a wrong answer: if the forecast feeding it is wrong, the calculation flawlessly derives a wrong requirement list.
The distinction matters because blame lands in the wrong place. The planner corrects the MRP output, the correction is not recorded, and next period the same error recurs identically.
Where they have to meet
A forecast carries uncertainty; MRP does not. A forecast saying “400 ± 80” enters MRP as 400, and the band disappears.
Converting that band into safety stock is the point where the two systems join. In most installations the conversion is a fixed percentage — meaning the uncertainty was measured and then not used.