--- title: "Why System Throughput Falls While Every Function Improves Its Own Metric" description: "Functional optimization degrades end-to-end performance wherever the measurement boundary coincides with the departmental boundary, because the gain each unit books against its own metric reappears next door as inventory, queue time or rework, and is never carried as a separate line in any account. The neutralizing mechanism is not individual awareness but a distinct authority and record regime governing interface decisions." url: https://www.beirek.com/en/blog/functional-optimization-local-optima-cost canonical: https://www.beirek.com/en/blog/functional-optimization-local-optima-cost published: 2026-01-25 modified: 2026-01-25 category: "Operations & Supply Chain" category_url: https://www.beirek.com/en/blog/category/operations-supply-chain language: en-US reading_time_minutes: 7 publisher: BEIREK LLC publisher_url: https://www.beirek.com license: "© BEIREK LLC — citation with attribution and link permitted" keywords: ["functional optimization","end-to-end performance","working capital and inventory turns","delivery reliability","interface governance in capital projects"] topics: ["Operating metric architecture and attributability","Working capital consequences of unit-cost decisions","Interface risk in package and contract design","Due diligence treatment of margin repeatability"] alternate_language_url: https://www.beirek.com/tr/blog/functional-optimization-local-optima-cost --- # Why System Throughput Falls While Every Function Improves Its Own Metric > **In short:** Functional optimization degrades end-to-end performance wherever the measurement boundary coincides with the departmental boundary, because the gain each unit books against its own metric reappears next door as inventory, queue time or rework, and is never carried as a separate line in any account. The neutralizing mechanism is not individual awareness but a distinct authority and record regime governing interface decisions. *Procurement lowers unit input cost, manufacturing raises asset utilization, logistics cuts cost per ton shipped; delivery reliability nonetheless deteriorates and working capital expands. The pattern originates where the measurement boundary coincides with the boundary drawn on the organization chart, and its price accumulates not in the income statement but on the balance sheet.* --- In the monthly operating review of a multi-site industrial group, all four functions report improvement: procurement, having consolidated its supplier base and moved to larger purchase lots, has pulled unit input cost down; manufacturing, running longer campaigns and cutting changeover frequency, has raised asset utilization; logistics, consolidating shipments, has reduced cost per ton moved; quality has brought the reject rate lower. In the same session the commercial side reports that on-time-to-promise performance has slipped, that two anchor accounts have reduced order size, and that requests for expedited shipment have multiplied. Finance notes, separately, that inventory turns have slowed and the cash conversion cycle has lengthened. None of the reported figures is wrong; each has been measured correctly, computed correctly, and presented correctly. The same pattern surfaces a second time at the table where capital budget is allocated, where every improvement business case brought by a function is defensible on its own terms, payback periods are plausible, and savings lines are concrete; the sum of those business cases, however, never materializes in the following year's income statement at anything approaching the promised magnitude. No separate account interrogates whether a saving declared by one function has reappeared in the cost base of its neighbor, because corporate accounting attributes the decision that produced the saving to a named center while dissolving the burden that decision created into overhead, cost of production or working capital. The saving thereby becomes a named line item and the offsetting load an anonymous residual. This behavior carries the name functional optimization — the improvement of performance at the level of individual units rather than the end-to-end system — and its mechanics rest on the coincidence of the measurement boundary with the boundary drawn on the organization chart. A metric is deemed manageable only if it can be attributed to one accountable owner, and attributability naturally selects magnitudes that stay inside a function while excluding magnitudes born at the interface between functions. That selection is not an error; it is rational to the extent that it lowers coordination cost, keeps accountability legible, and spares managerial attention from fragmentation. The difficulty lies not in the shortcut itself but in the persistence of the measurement architecture after the conditions that justified it — stable demand, short lead times, a narrow product range — have changed. The operational expression of the mechanism appears along three characteristic lines. In procurement, the distant source accepted in exchange for a lower unit price, together with high minimum order quantities and extended lead times, gives the price differential back through inventory carrying cost and diminished responsiveness to demand shifts. In manufacturing, long campaigns reduce changeover loss and lift utilization while narrowing mix flexibility, thereby enlarging work-in-process and lengthening the order queue. In logistics, consolidation lowers cost per shipment by increasing the number of loads waiting to be dispatched, while widening the dispersion of delivery dates and, in exceptional cases, forcing expedited freight. The three lines share one structure: a buffer is never removed, only relocated; and the place it now occupies is typically a unit that does not measure it. To this must be added the arithmetic of capacity. To the extent that the output of a production or supply chain is set by the capacity of its narrowest section, an efficiency gain achieved at a point that is not that section raises no system output whatsoever; it merely increases the quantity of material accumulating in front of or behind that point. The distinction between efficiency and effectiveness resides precisely here: doing the work correctly is an achievement measurable at unit level, whereas doing the correct work can only be defined at system level. Because the functional metric regime rewards the former and assigns the latter to no accountable owner, the organization drifts predictably toward the former. Incentive structure hardens the persistence of the tendency. Where bonus, budget allocation and promotion decisions are tied to functional metrics, the individual return to a decision-maker who withdraws a choice that generates cost at the interface is negative, since the withdrawal surrenders a gain recorded on his own dashboard in order to reduce a burden recorded on none. Institutional memory reinforces the same asymmetry: the history of the saving sits in the presentation archive, while the history of the offsetting load, kept nowhere, remains knowledge that must be rediscovered in the following period. The matter is therefore not weakness of individual will but absence of a record regime. The corporate price accumulates first on the balance sheet. While the unit-cost gain occupies a line of the income statement, the inventory build, the extended cash conversion period and the enlarged working capital requirement produced by the same decision convert into financing cost; in a company carrying leverage, this directly compresses utilization of the working capital facility and, with it, the room available under covenant headings. Secondary items arising from weakened delivery reliability follow: expedited freight, overtime, exposure to liquidated damages, and the lengthening of the sales cycle that follows erosion of customer confidence. What these items share is that each is individually small and readily absorbed into overhead, while in aggregate they can exceed the unit-cost differential earned by a multiple. The second price emerges once the company enters a transaction or a credit process. The question posed by the acquirer and by the credit committee during due diligence concerns not the level of margin but the conditions under which that margin is repeatable; in a structure where delivery performance oscillates and inventory levels are determined by lot size rather than by demand, that question cannot be answered affirmatively. The foreseeable consequences are a working capital normalization deducted from price, representations and warranties extended toward operational undertakings, a higher escrow ratio, and an earn-out trigger anchored to an operational indicator such as delivery performance rather than to profit. In capital projects the same pattern appears in tender design: reducing price per package by increasing the number of packages leaves interface responsibility in the gap between contractors, and the saving returns as programme delay and claim volume. The mechanism that neutralizes this tendency comprises four distinct components. The first is the placement, alongside functional metrics, of at least one end-to-end measure that no single function can improve unilaterally — on-time-to-promise, order-to-cash duration and total delivered cost being its three typical constituents. The second is a trade-off record: each functional improvement proposal quantifies, at the moment of proposal rather than the moment of approval, the inventory, waiting time or lost flexibility it creates on the other side, and any proposal lacking that declaration is not tabled. The third is separation of authority — optimization contained within a function remains free, while parameters that alter the interface, among them lot size, lead time, buffer location and package boundary, belong to the system owner. The fourth is cadence: those parameters are reopened on a review calendar aligned with the rate at which the demand profile changes, not with the annual budget. The intervention BEIREK conducts in capital projects and multi-site industrial groups is built on converting these four components into a governance architecture. Package and contract boundaries are drawn according to interface risk rather than cost category; every line at which two contractors intersect in the responsibility matrix is assigned a distinct owner, and any interface left unclaimed is named before contract award. The decision record is maintained so that the burden created on the opposite side is written on the same page at the moment the saving is declared; programme, cash and inventory effects are consolidated into a single table so that the same three magnitudes sit side by side in every review session. Ahead of changes to critical parameters, a pre-mortem session describing the scenario in which the proposal fails is run, and the counter-argument role is assigned permanently to a named individual. The difference this architecture produces owes less to the quality of decisions than to their visibility; once it becomes written which metric improves while which account deteriorates, the argument over whether a functional gain constitutes a system gain closes before the board in a single table. The end-to-end performance of an organization is not the sum of its functions but the design quality of its interfaces; and for as long as the owner of the interface remains undefined, the improvements every unit reports accurately will continue to cancel one another silently. The operative question is this: among the improvement proposals approved this quarter, for which of them is there a record of which item it enlarged on the other side, and by how much. ## Key Points - Functional metrics institutionalize quickly because they are measurable and attributable to a single owner, whereas cost created at the interface remains off the books, since no account owns it. - Efficiency gained at a point that is not the binding constraint does not raise system output; it produces intermediate inventory and waiting time instead. - A unit-cost gain appears on one line of the income statement, while its price accumulates in inventory turns, cash conversion days and expedited freight expense. - Buyers and credit committees confronting weak delivery reliability price not the level of margin but its repeatability, which directly enlarges escrow ratios and widens earn-out coverage. - The effective intervention is a decision record that compels each functional improvement proposal to declare the burden it creates elsewhere at the moment of proposal rather than at the moment of approval. ## Questions ### What is functional optimization, and why can it be harmful? Functional optimization is the improvement of performance at the level of individual units rather than the end-to-end system. As each unit raises its own measurable metric, the counterpart of that gain reappears next door as inventory, queue time or lost flexibility. What proves harmful is not the improvement itself but the fact that the burden created at the interface is never carried as a separate line in any account, so the net effect is never computed at all. ### Why does lowering unit cost not always produce a gain for the company? A lower unit price is typically obtained in exchange for a high minimum order quantity, an extended lead time, or a distant source. Those three elements act directly on inventory levels, cash conversion duration and responsiveness to demand shifts. The price differential appears in the income statement while its counterpart accumulates on the balance sheet as working capital; once financing cost and expedited freight are added, the net effect can turn negative. ### How does delivery performance fall while departmental metrics improve? Departmental metrics measure magnitudes attributable to a single owner, whereas delivery performance forms at the intersection of several units and lies within the unilateral control of none. Long production campaigns, shipment consolidation and large purchase lots each improve their own metric while lengthening the order queue. The buffer does not disappear; it moves to a unit that does not measure it, and its effect surfaces in the dispersion of delivery dates. ### Which institutional mechanism prevents this tendency? Not individual awareness but decision architecture. An end-to-end measure that no single unit can improve unilaterally is placed alongside functional metrics; each improvement proposal declares the burden it creates elsewhere at the moment of proposal; interface parameters such as lot size, lead time and buffer location are decided by the system owner rather than by the function; and those parameters are reopened on a calendar aligned with the rate at which the demand profile changes. --- Source: https://www.beirek.com/en/blog/functional-optimization-local-optima-cost Publisher: BEIREK LLC — https://www.beirek.com