Setup Time: The Hour That Never Appears in the Plan and Always Appears in the Capacity

In a weekly production planning meeting, what is instructive is not where the discussion concentrates but where it never does. Order sequencing, priority conflicts, key accounts, and material confirmations are debated at length, while the allowance entered into the plan for the transition between two different products is rarely questioned at all; the figure sitting in that column is round, inherited from last year, and in most cases identical regardless of product family. A month later, in the same meeting, the line is found to be behind plan, and the shortfall is attributed to material delays, unplanned breakdowns, or absenteeism. The shortfall is in fact the accumulation of transitions, none of which drew attention individually, whose aggregate reaches a visible fraction of the shift budget. Over the month the line stopped more often than the plan assumed, stayed down somewhat longer than the plan assumed on each occasion, and the difference never appeared in any record under its own name.

The same pattern surfaces on the cost side through a different aperture. In most costing systems, setup activity is not an independent line item; it sits inside the manufacturing overhead pool and is allocated on machine hours. That allocation base loads a ten-thousand-unit batch and a two-hundred-unit batch on identical logic, although the transition resource the two consume may differ by an order of magnitude. When the quoted price on the commercial side is derived from a unit cost built on cycle time multiplied by quantity, small orders reach the market at a structurally subsidized price. The consequence, over several quarters, is that the customer mix drifts on its own toward the orders that generate the most changeovers and leave the thinnest margin — a drift that occurs not as a commercial choice but as the natural output of the price list.

The tendency underlying this pattern is described in the operations literature as setup-time bias: the systematic assumption, in both planning and costing, that preparation and changeover durations are shorter than they are. Its origin is not inattention but the architecture of measurement. Cycle time is recorded continuously by the machine's own counter, without human intervention; setup falls into the interval between two counters and exists there only if somebody writes it down. When it is written down, what is captured is typically the intervention window alone — the minutes during which the die is removed and mounted. Searching for the fixture, retrieving it from the tool crib, dialing in the adjustment part, waiting for first-article approval, waiting for the quality signature, and the scrap generated during ramp-up all fall outside that window; some of it lands in the downtime log, some in the quality loss, and some nowhere at all.

A second source of the bias concerns which instance memory retains. Asked how long a given changeover takes, the answer supplied is not the average but the duration observed under the most favorable conditions: the transition completed by an experienced setter, with the fixture already staged, correct on the first attempt. Once that value enters the plan it functions as an anchor, and every subsequent revision is made around it within a narrow band. The fact that the true distribution is right-skewed — that rare but long transitions pull the mean upward — never reaches the planning assumption, because the plan demands a single number and the single number available is the most memorable one.

It is worth recognizing that this tendency is functional under specific conditions; otherwise the intervention is built in the wrong place. In a configuration characterized by low product variety, large batch sizes, and long delivery lead times, setup genuinely is a negligible item, and the administrative cost of measuring it separately exceeds the value of the information obtained; the shortcut is rational there. The difficulty lies not in the shortcut itself but in its persistence once the conditions have changed. Where the number of active SKUs has multiplied, where customers have reduced order sizes while increasing frequency, and where delivery commitments have compressed, setup ceases to be negligible and becomes the effective constraint on the line — yet the planning parameters remain calibrated to the period of low variety.

The first balance-sheet consequence of that persistence is capacity. Effective capacity appears higher than it is, in proportion to the transition loss omitted from the calculation, and this overstated figure becomes the foundation for sales commitments, promised delivery dates, and budget targets. When the commitments fail to hold, the diagnosis is usually a shortage of capacity, and the proposed remedy is additional equipment. Where the deficit originates in transition burden rather than cycle capacity, however, a new machine resolves nothing; it distributes the same transition burden across a larger capital base, raises unit fixed cost, and exposes the payback calculation to deviation from the first year forward. If the file arriving at the investment committee contains no source decomposition of the capacity gap, the decision has been taken on the available justification rather than the correct one.

The second consequence appears in working capital, and typically hides not in the inventory balance itself but in the question of why that balance has failed to move year over year. For a planner living with long and unpredictable changeovers, the most rational defense is to enlarge the batch; every enlargement rescues the weekly plan by reducing the number of transitions the line must absorb. That defense is correct in each individual week and, in aggregate, depresses inventory turns, extends the cash conversion cycle, and increases obsolescence and model-run-out risk. A measurement gap thus converts, two steps downstream, into a financing requirement — and by the time the conversation reaches the credit committee, the subject under discussion is no longer changeover duration but the working capital limit.

In capital projects the same tendency presents itself under different vocabulary. What corresponds to a production line's changeover is, on site, mobilization, access and scaffolding erection, isolation and lockout-tagout, permitting, hot work authorization, and the tie-in window preceding commissioning. Embedded inside activity durations, these items never appear in the schedule under their own names, carry no float of their own, and cannot be tested as independent elements in critical path analysis. In a configuration where several contractors work sequentially in the same area, transition duration is additionally sequence-dependent: which trade follows which materially alters total mobilization burden. Where that dependency never enters the scheduling logic, delay is attributed to contractor performance, although what has been produced is a structural burden arising from the sequence itself, with liquidated damages exposure accruing on top of it.

The third consequence emerges the moment the company enters a sale or partnership process. Among the first places a technical due diligence team will look is how availability loss within the equipment effectiveness calculation is explained; where the greater part of that loss sits in an undefined downtime category, the reasonable posture on the buy side is to anchor its analysis not to management's capacity projection but to the actual output of the trailing twelve months. The valuation expression of that posture is a compressed multiple, or the migration of the projection-dependent portion of consideration into an earn-out structure. What determines value at this table is not performance itself but the ability to demonstrate that performance is explicable and repeatable; an unrecorded changeover renders even genuinely strong performance indefensible.

The mechanism that neutralizes this tendency is not individual attention but an institutional design composed of four components. The first is the measurement boundary: setup is defined as the complete interval from last good part to first good part, and the waiting, approval, and ramp-up scrap contained within that interval are not permitted to migrate outside the line item. The second is the allocation base: setup is separated from the machine-hour pool and distributed on number of changeovers, so that the genuine burden of a small batch reaches the price list. The third is sequence dependency: instead of a single average, a family-to-family transition matrix is maintained, and production sequencing is performed against that matrix. The fourth is authority: it is established explicitly who may approve a schedule built on an out-of-band setup assumption, and that approval is recorded together with the reasoning behind the assumption.

BEIREK's intervention in this area begins not with a critique of the existing plan but with placing the parameter beneath the plan onto its own record. In the projects we manage, transition and mobilization items are not buried inside activity durations; they stand as independent lines carrying their own duration, their own responsible party, and their own float, and for sequence-dependent scopes the transition matrix is made part of the programme annexed to the contract. The assumption record is kept at the moment of proposal rather than the moment of approval, which means the question of on what basis, on whose representation, and against which observation window a given duration entered the plan ceases to be something reconstructed retrospectively once the plan fails. Where equipment expenditure is raised on capacity grounds, we apply a single test before FID: how much of the open capacity is cycle-driven and how much is transition-driven. Where the second component dominates, the subject of the file is not investment but the separation of setup work that can be performed while the machine runs from setup work that requires it to stop — a separation that typically produces the same result at a small fraction of the capital cost.

The real capacity of an operation becomes visible not in how fast the machine runs but in who measures the gap between two jobs, against what definition, and at what frequency. That gap does not vanish for want of a line item; its ownership merely becomes indeterminate, and every duration without a determinate owner eventually declares itself inside a capital decision, an inventory level, or a delay penalty.