1. What MRP is, and what this system is

The problem

You make things out of other things. A bicycle is a frame, two wheels, a drivetrain, brakes, a cockpit; a wheel is a hub, a rim, thirty-two spokes and a tyre; a frame is tube, dropouts, paint. Some of those you buy, some you make, and every one of them has to be there — in the right quantity, at the right time — or the bicycle does not get built. Six hulls in the warehouse and no rudder bushings is not a bicycle problem, but it is exactly this problem.

The naive answer is to keep plenty of everything. That ties up money, fills shelves, and still fails, because "plenty" is a guess and the one part you run out of is the one you didn't guess about.

Material Requirements Planning

MRP is the calculation that replaces the guess. It was worked out in the 1960s on early IBM computers, written up by Joseph Orlicky in Material Requirements Planning (1975), and spread through the 1970s by APICS as something of a crusade. Its central insight is a distinction:

• Independent demand — what customers want — has to be forecast. Nobody knows how many bikes will sell in March.
• Dependent demand — what the bikes will need — should never be forecast. It can be calculated, exactly, from what you have decided to build.

Decide to build six bikes in March and you know you need twelve wheels, three hundred and eighty-four spokes, six frames, and so on down every level of every bill of materials. That is the explosion. Subtract what is on hand and what is already on order, and you have net requirements. Push each requirement back by its lead time — spokes take thirty days from Switzerland, so the order must go in early February — and you have a plan of when to order and when to start making. Run that every week as reality changes, and you have MRP.

Three inputs, in Orlicky's words:

1. The master production schedule — what you intend to make, and when.
2. The bill of materials — what everything is made of, level by level.
3. Inventory records — what you have and what is coming.

And one hard truth that the MRP literature repeats until it is tedious: the calculation is only as good as the records. A bill of materials that is 90% right and stock counts that are 90% right produce a plan that is wrong somewhere in nearly every product. The practitioners of the 1970s learned to insist on record accuracy above 95%, kept honest by cycle counting — counting a few locations every day rather than everything once a year. This system takes the same view: the stock ledger is never edited, only added to, and the reconcile tool exists to prove the cached totals against it.

MRP II

By 1980 the calculation had grown a company around it, and Oliver Wight coined Manufacturing Resource Planning — MRP II — for the result. The original MRP plans materials and quietly assumes infinite capacity: it will happily schedule forty wheels on Tuesday whether or not anyone can build them. MRP II closes that loop:

• Routings and work centres: every made item has a sequence of operations, each at a kind of workstation, each taking time. Now the plan can be checked against capacity — how many hours the wheel stands actually have — and adjusted.
• Shop floor control: work orders released, started, reported complete, with materials consumed as they are used, so the inventory records the plan depends on stay true because of production rather than despite it.
• Purchasing as part of the same plan, with suppliers, lead times and open orders feeding back into what is available when.
• Costing and finance on the same data, so a standard cost rolls up a bill of materials and the company's books and its plan agree.
• The whole thing run as a closed loop, top to bottom, sales and operations planning down to the bench.

Later the same idea, extended to everything a company does, was renamed ERP by Gartner in 1990. That is the world of SAP and Oracle; it is not where this system is going.

What this system is

Foudra is a stock control and production system with MRP II's structure and, today, a reorder-point brain. That is a deliberate sequence, not a limitation being hidden.

What is here, and mirrors the MRP II model:

What replenishment uses today is the older, simpler logic: a reorder point per part — order when on hand plus on order falls to the minimum. It is honest (it counts what is on order) and it is what most small manufacturers actually run on. What it does not do is look ahead: it cannot say "you will be short of bushings on the 14th because of the six hulls due on the 20th". That requires the master production schedule and the time-phased explosion, and the tables are shaped for exactly that addition.

The one MRP II idea already fully present is the most useful one for a small shop: buildable quantity with its binding constraint — how many of something you could assemble from the shelf right now, and which component stops you making more. That is the explosion, run backwards, for today.

Words that matter here

A handful of terms the rest of the manual uses without explanation:

• Part — anything with a number: bought, made, or sold. A version is a separate part row.
• Ledger — the append-only record of every stock movement. On hand is a cached total of it.
• Standard vs actual — the planning figure (standard cost, BOM quantity) versus what a specific order paid or a specific job consumed. Both are kept, deliberately apart.
• Phantom — a part built inside a work order and consumed by the next step without ever sitting on a shelf. Rare.
• Backflush — issuing a step's materials from stock automatically when the step is reported complete, instead of issuing them by hand beforehand.
• Group and station — a capability ("wheel truing") and one instance of it ("stand 2").
• Reorder point — the minimum stock below which a part should be reordered.

On to setting up.