The calculation that decides what to buy and make, when — and the three inputs that decide whether it can be trusted.
MRP is the calculation that turns demand into supply suggestions. It takes what you need to deliver, explodes it through the bill of materials, subtracts what you already have or have on order, and proposes purchase and work orders offset by lead time. It is arithmetic, not intelligence: the output is only as good as the BOM, the stock figure and the lead times it reads.
Without MRP, purchasing is done from memory and a spreadsheet, which works until the product range or the order book grows. The characteristic failure is not running out — it is running out of one component while holding six months of another. MRP replaces that judgement with a calculation that can be re-run whenever demand changes.
Firm sales orders plus a forecast, together forming the master production schedule. Garbage here produces garbage everywhere downstream.
Each finished item is broken into components, level by level, multiplying by quantity per and allowing for scrap.
Current stock, open purchase orders and open work orders are subtracted. Only the shortfall becomes a suggestion.
A suggestion is dated backwards from when the material is needed, using the item's purchasing or manufacturing lead time.
MRP proposes. A planner reviews, consolidates and converts. Systems that auto-release MRP output without review tend to generate noise.
A planner opens the MRP output and sees a suggestion to order 340 of a bracket, due in eleven days. The instinct is to distrust it. The right move is to peg it.
The arithmetic unwinds cleanly. Gross requirement is 500, from two works orders for the parent assembly — 300 for one, 200 for the other. On hand is 220, of which 60 are already reserved against an earlier order, so available is 160. Net requirement is 340. Supplier lead time is eight days and the first works order starts in eleven, so the order date is three days out, not today. Nothing is wrong; the number simply had four inputs and the screen showed one.
Pegging is what turns MRP from an oracle into a tool. A planner who can see those four lines will act on the suggestion. A planner who cannot will override it — and once planners routinely override MRP, the system is generating work rather than removing it. That is the real failure mode, and it is a visibility problem rather than an algorithm one.
| Term | What it means |
|---|---|
| Gross requirement | Total demand for a component before anything on hand or already ordered is taken into account. |
| Net requirement | What is actually needed after stock, reservations and open orders are netted off. This is what MRP turns into planned orders. |
| Lead time offset | Working backwards from the date something is needed to the date it must be ordered. Wrong lead times are the most common reason MRP output looks absurd. |
| Lot sizing | The rule that turns a net requirement into an order quantity — lot-for-lot, fixed quantity, or a minimum plus multiples. |
| Pegging | The trace from a planned order back to the demand that caused it. Without it, nobody can answer why MRP is suggesting something. |
MRP nets against the stock balance. If the balance is wrong, every suggestion below it is wrong. Cycle counting is a prerequisite, not an optional extra.
Lead times entered at implementation and never revisited are the most common cause of chronically early or late ordering.
A forecast is an estimate. Running MRP on an unreviewed forecast commits real money to a guess.
A weekly MRP run on a business whose demand changes daily will always be reacting to last week.
| Metric | What it tells you | What distorts it |
|---|---|---|
| Planned order override rate | How much planners trust the output. A high rate means the inputs are wrong, not the planners. | Overrides recorded as new orders rather than as changes, which hides them. |
| Lead time accuracy | Whether quoted supplier lead times match actual receipts. Everything downstream depends on this. | Averaging across suppliers; one unreliable supplier can be invisible in the mean. |
| Schedule adherence | The share of works orders completed in the planned period. | Rescheduling before measuring, which makes adherence perfect by definition. |
| Inventory turns | Whether netting is actually reducing stock or simply reordering it on time. | Seasonal demand; compare like periods, not consecutive ones. |
We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.