MRP as Central Planning
MRP as Central Planning
Mises in the Factory — No. 2
Why the production plan fails for the same reasons Gosplan did.
Monday, eight o’clock. The MRP plan came out overnight: production orders by department, purchase requirements by item, delivery dates lined up day by day. On screen it is perfect.
At ten a press stops because of a damaged die. A supplier calls to say the batch of gaskets will arrive Thursday, not Tuesday. The main customer pulls an order forward by two weeks. None of this information was in the plan, and none of it could have been when the plan was calculated.
By eleven the department head already has a spreadsheet open. At two in the afternoon the planning office reruns the MRP, which moves three hundred orders. By Friday nobody looks at the plan anymore: everyone works from the expedite list.
Anyone who has spent a few years on a shop floor recognizes the scene. Less well known is that its mechanism was described decades before MRP existed, in connection with a far bigger problem: the planning of an entire economy.
What MRP promises
MRP (Material Requirements Planning) was born in the 1960s, with Joseph Orlicky’s work at IBM and the spread of the first business computers. The idea is elegant: starting from forecast demand for finished products, through the bill of materials and lead times, you calculate backwards what to produce and buy, in what quantity and when.
The technical merit is real. MRP replaced reorder-point replenishment, which ignored the dependency between components, with a coherent calculation of requirements. For a company with thousands of items, exploding the bill of materials is work that no office could do by hand.
But the implicit promise goes beyond calculation. It is the idea that, once all the data is gathered in a single system, a center can decide for every department what to do, and that the factory’s problem is mainly a computational one. It is the same idea that underpinned Soviet central planning.
Gosplan on the shop floor
Gosplan, the Soviet Union’s state planning committee, did on a grand scale what MRP does on a small one. It set final production targets, derived from them the requirements for steel, coal, machinery and labor, and assigned quotas to every plant. On paper the plan balanced.
To work, MRP rests on three assumptions that are rarely stated.
- Demand is known. The calculation starts from a forecast and treats it as a fact. Every forecast error is amplified along the bill of materials.
- Lead times are fixed. Each item has a throughput time recorded in the master data, independent of the load on the departments. In reality lead time depends precisely on load: the more the plan pushes orders into a department, the longer the queue.
- Capacity is infinite. Classic MRP calculates what is needed, not whether it can be done. The capacity check comes afterwards, if it comes at all.
These are the central planner’s assumptions: a known end, stable transformation times, resources available when the plan calls for them. When one of the three fails, and on a shop floor at least one fails every day, the plan becomes a wrong instruction executed with discipline.
Hayek at the terminal
In 1920 Ludwig von Mises argued that an economy without private ownership of the means of production cannot calculate: without market prices for capital goods, the planner does not know which use of resources is the most economical. In 1945 Friedrich Hayek, in The Use of Knowledge in Society, shifted the emphasis from calculation to knowledge.
Hayek’s thesis is that the economic problem is not how to process data that has already been collected, but how to use knowledge that never exists in concentrated form. It is knowledge “of the particular circumstances of time and place”: scattered among thousands of people, often tacit, and above all changeable. Its value lies in the moment it appears.
On the shop floor this knowledge has precise faces. The operator knows the die is showing signs of wear. The warehouse clerk knows the incoming batch has a packaging problem. The shift leader knows that with two people on vacation, line 3 cannot hold the pace. None of this enters the master data, and by the time it does, it is already stale.
MRP doesn’t fail because it calculates badly. It fails because it calculates well on data that, by the time it is needed, no longer describes the factory.
System nervousness and shadow spreadsheets
The planning literature has a name for what happens when the plan meets reality: system nervousness. A small change downstream, an order pulled forward or a scrap event, propagates through the bill of materials and shifts hundreds of orders upstream. Every MRP rerun produces a new plan, coherent and already obsolete.
The organization’s response is predictable, and it is the same one observed in planned economies. Alongside the official plan, a parallel economy grows: the department head’s spreadsheet, the expedite list, the direct phone call to the supplier, the stock kept “just in case” in a corner of the warehouse. It is the factory’s black market: the people who hold the local knowledge work around the plan to make things run.
The paradox is that these shadow systems are often what keeps deliveries standing. But they are invisible to management, which keeps measuring adherence to a plan nobody follows. The result is a double cost: you pay for the central planning structure, and you pay for the informal work that corrects it.
Under Gosplan, enterprises inflated their requirements to build in a margin and hid capacity so quotas wouldn’t be raised. On the shop floor the phenomenon has a familiar name: “prudent” lead times in the master data, which stretch the plan, increase inventory and make the system even slower to react.
Where the analogy stops
A serious objection has to be faced. Every firm is, by definition, an island of planning. Ronald Coase explained it in 1937: inside the firm, resources move not by following prices but by following the decisions of those who direct. If planning were always wrong, firms would not exist.
The Austrian answer, formulated by Murray Rothbard, is that the firm can plan internally precisely because it is surrounded by a market. It buys factors and sells products at prices it doesn’t set, and those prices tell it whether its internal planning is creating or destroying value. The limit to the size of the firm arrives when, for part of the chain, those external prices no longer exist.
That is why Mises’s calculation argument does not apply to MRP as it stands: the company knows the cost of its materials and the price at which it sells. What applies fully is Hayek’s argument about knowledge. MRP’s problem is not the absence of prices, but the pretense of concentrating in a center knowledge that by its nature stays dispersed across the departments.
The distinction is useful in practice. Planning at the aggregate level, where coherence and a long horizon are what matter, is legitimate. Prescribing from the center what each department must do each day, ignoring what only the department knows, is the point where planning becomes Gosplan.
What to do on Monday morning
There is no need to switch off the MRP. What it needs is to be given back the job it is good at, and relieved of the one it cannot do.
- Use it for the horizon, not for execution. Bill-of-materials explosion, requirements for long-lead-time materials, aggregate capacity planning: here the central calculation adds value.
- Decentralize execution with pull signals. On the shop floor, the kanban of Episode 1 transmits actual scarcity instead of forecast scarcity. The signal originates where the knowledge is.
- Decouple with buffers placed deliberately. Approaches like Demand Driven MRP (DDMRP) put strategic stock at chosen points in the chain, so that a change downstream does not propagate through the whole bill of materials.
- Bring the shadow spreadsheets into the light. Every parallel Excel file signals knowledge the system cannot see. Instead of banning it, ask what the department head knows that the master data doesn’t.
- Measure flow, not adherence to the plan. Actual lead times, on-time delivery to the customer, work in process. An adherence metric rewards whoever executes a wrong plan.
- Leave the departments an explicit margin of decision. Whoever sees the problem first must be able to change the sequence without waiting for the next overnight rerun.
The plan and the order
The production plan is not the enemy. It becomes one when it claims to replace the order that grows out of the shop floor: local signals, decisions by the people who see the problem, continuous adjustments that no center could prescribe.
The Soviet Union did not collapse for lack of computers. Many factories don’t struggle for lack of ERP processing power. In both cases the limit is the same: the knowledge that matters cannot be gathered into a center.
In the next episode we go down to the gemba, where that knowledge lives: Hayek at the Gemba. Knowledge of the circumstances of time and place lives on the shop floor, not in the executive suite.
References: Ludwig von Mises, “Economic Calculation in the Socialist Commonwealth” (1920); Friedrich A. Hayek, “The Use of Knowledge in Society”, American Economic Review (1945); Ronald H. Coase, “The Nature of the Firm” (1937); Murray N. Rothbard, Man, Economy, and State (1962); Joseph Orlicky, Material Requirements Planning (1975); Carol Ptak and Chad Smith, Demand Driven Material Requirements Planning (2016).
Write a comment