In a production planning meeting, a customer advises that one order will be pulled forward by two weeks; the planning engine runs overnight, and the schedule that reaches the table the following morning has moved not only that order but the dates of dozens of work orders sharing neither a material nor a line relationship with it. Purchasing discovers that a set of supplier releases placed for three months out has been advanced; the production manager finds that the line plan he circulated last week as settled is now void; suppliers, having received a third consecutive revision notice, quietly stop anchoring to the requested date and begin scheduling against their own capacity instead. The pattern recurs and is largely predictable in its shape: there is no proportionality whatsoever between the magnitude of the change entering the system and the magnitude of the change leaving it.
The same pattern appears well beyond the factory floor, showing up with equal regularity in capital-intensive project schedules. A one-month slip in an equipment delivery, once recalculated inside an integrated programme, displaces the dates of line items that sit nowhere near the critical path, and the weekly schedule issued to the field differs from the previous week's in several hundred rows. What then happens on site follows a familiar sequence: the official programme continues to be distributed, while the actual order of work is governed by a separate list the superintendent maintains himself. Once the link between a plan remaining in circulation and a plan producing decisions is severed, the planning function ceases to operate as a coordination instrument and settles into a reporting ritual.
The behavior has a name — **schedule nervousness**, the tendency of a small input change to produce a disproportionate and propagating rearrangement of planning output. Its mechanism follows directly from how the planning system is constructed. An engine operating on material requirements logic does not treat the existing schedule as a starting position to be preserved; it accepts the order pool, inventory position, lead times, and lot-sizing rules as inputs and regenerates the schedule from them. Because that derivation is deterministic, a deviation anywhere in the input set travels downward through a multi-level bill of materials, compounding at each level as it interacts with lot consolidation and lead-time offset rules, until an order for a component three levels down — a component with no visible connection to the original change — lands in an entirely different week.
The second layer of the mechanism lies not in the arithmetic of the calculation but in the incompleteness of the objective function. The system searches, in principle, for the best available schedule, yet nothing in that search assumes change itself carries a price. Moving a date is a free operation inside the model, whereas in practice it generates a fresh notification to a supplier, a round of internal recoordination, a tooling changeover, a shift-plan revision, and, more often than acknowledged, an erosion of credibility. So long as the cost of change is priced at zero, the system will find it rational to rewrite the entire schedule in pursuit of a marginal improvement, on every run; the difficulty is not that the system is working incorrectly, but that it is executing the objective it was given rather too well.
The third layer is organizational, and it proves more determinative than anything technical. In a culture where every signal arriving from the demand side — a sales representative's verbal note, a customer intention not yet confirmed, a regional manager's forecast adjusted upward against a year-end target — is admitted to the plan as input on arrival, the instability of the system is essentially a mirror of the instability of the organization. These reflexes are not errors; in the short run they represent the cheapest available way to demonstrate responsiveness to a customer, to evidence flexibility, and to keep the commercial path clear. The problem resides not in the shortcut itself but in its persistence once conditions shift — once lead times lengthen, capacity fills, and the supplier base narrows.
The institutional cost surfaces first in the working capital cycle. When what the schedule will demand in a given week becomes unforecastable, the purchasing function manages that unpredictability with the only instrument available to it, namely inventory; safety stock levels are raised not against a calculated service-level target but against the accumulated memory of past surprises. Inventory turns slow, though the slowdown appears on the balance sheet as a level rather than an explanation, and the unanswered question is never the line item itself but why the line item sits above where it sat a year earlier. The same unpredictability generates a buffer on the supplier side as well, and that buffer enters price: a quotation extended to a buyer with a high revision frequency runs structurally above one extended to a stable buyer for identical volume.
The second cost category consists of the expediting charges distributed across the income statement. In a schedule under continuous rewrite, a delivery date that looked comfortable a week ago can turn critical this week, and the response is air freight, a weekend shift, a partial shipment, an additional installation crew. Individually these items are modest and are rarely reported alongside one another; because logistics expense, overtime, and outsourced labor sit in separate accounts, the aggregate annual cost of planning instability never presents itself as a single line in any management report. A cost that cannot be seen cannot be defended in a budget discussion, and what cannot be defended in a budget discussion is not, in practice, reduced.
The third and most durable cost is the plan's loss of legitimacy. Where the schedule the system produces changes materially from week to week, the people at the working face and along the supply base develop an entirely rational accommodation: they read the new schedule and decline to act on it, determining the real sequence from their own experience instead. Beyond that point, knowledge of how the operation actually runs resides not in the system but in the heads of a few individuals, and the company has, without deciding to, converted institutional memory into personal memory. In a due diligence process this typically surfaces through a single question — what is the adherence rate between planned sequence and executed sequence — and a low ratio reads as the most direct available evidence that delivery performance is not repeatable independently of a particular operations manager.
What neutralizes the tendency is not a sharper forecast but a decision architecture, and that architecture separates cleanly into three components. The first is the **frozen horizon**: a window, its length set by the longest critical lead time and setup duration, inside which the schedule is closed to alteration, with any intervention within it routed to the approval of a named individual rather than processed automatically. The second is the **change threshold**: a tolerance band defined per material and per work item, below which a deviation triggers no regeneration at all, preventing the system from rewriting a whole schedule in pursuit of a marginal gain. The third is **replanning cadence**: the moment of update is fixed by calendar rather than by event, so that a schedule published on a stable weekly or fortnightly cycle queues every intervening signal to the next cycle.
For those three components to function, a fourth and frequently omitted layer is required — a record capturing who requested a change, on what grounds, and at what moment, maintained at the point of request rather than at the point of approval. The discipline BEIREK installs on the programmes it manages is organized around exactly this record: every plan revision is tied to a source, a reason category, and a cost estimate, and the record forms the opening agenda item of the weekly schedule meeting. Within a few cycles the origin of revisions becomes visible, and in most cases the source of instability turns out to be not uncertain demand but a commercial habit of transferring unconfirmed information into the plan as though it were confirmed. The appropriate response is not to manage a conflict between sales and planning but to write an admission rule specifying the maturity a signal must reach before it may enter the plan.
The second line of intervention involves splitting the plan itself into two layers. Inside the frozen window the schedule constitutes a commitment and is communicated in identical form to supplier, field, and customer; beyond that window it is explicitly labeled indicative and enters circulation on the stated assumption that it will move. The operational effect is immediate and concrete: the supplier knows which date to reserve capacity against, the field team knows which week to staff, and the rationale for maintaining a second informal sequence disappears. The same separation also renders delivery performance measurable, since variance is now computed against a fixed commitment recorded at the moment of freeze rather than against a reference that never stops moving.
The instability the planning system generates has usually been erased entirely from the picture an investment committee receives; what reaches the committee is inventory level, on-time delivery rate, and margin, not the mechanism sitting behind those three indicators. Yet the measure of how governable an operation actually is has little to do with how optimistic its plan looks and a great deal to do with how frequently the plan is departed from and whether those departures are recorded anywhere. In assessing operational maturity, the question worth asking is not how accurate the forecast proved to be, but how much of the system a missed forecast is permitted to rewrite.
What ultimately warrants evaluation, then, is not the accuracy of the plan but its stability; and stability is obtained not by removing uncertainty from the environment, which no operation has ever managed, but by fixing institutionally the gate through which uncertainty enters the plan and the rate at which it is allowed to pass.