Suboptimization: When Every Indicator Reads Green and the Consolidated Result Stays Red

There is a recurring scene in the monthly operations review. Procurement presents an improvement in unit cost against the prior quarter, manufacturing reports record equipment utilization, logistics has moved order fill rate upward, and quality has driven the defect rate down. Every indicator on every slide points in the favorable direction, and no unit has behaved incorrectly with respect to its own mandate. Yet when the consolidated view opens at the end of the same session, the delivery date has slipped by a quarter, working capital has expanded, an unbudgeted air freight charge has been expensed, and a claim discussion has opened on the customer side. Nobody in the room owns that outcome, for the sufficient reason that the outcome does not appear in any unit’s indicator set.

The same pattern presents itself more sharply in capital-intensive project work. A procurement line that closes the main transformer order at a lower price but against a longer delivery window has behaved correctly with respect to its own savings target; the schedule owner encounters that trade only later, when the switchyard erection sequence has to be rebuilt around a delayed arrival, and at that point the incremental crane mobilization together with a second pass at site setup can absorb several times the savings originally captured. Two pieces of information were never present on the same table at the moment of decision: the price differential is measured on the procurement line, the cost of the resequencing accrues on the construction line, and no record connects one to the other.

The name of this pattern is suboptimization — the degradation of whole-system performance that results when each subunit optimizes the indicator assigned to it. The core of the mechanism lies not in measurement but in where the boundary is drawn. Every indicator is a proxy measure, and the validity of a proxy holds only within the boundary on which it was defined. The boundaries of an organization chart, however, are drawn along lines of controllability rather than along the path value actually travels; because a manager can be held to account only for variables under that manager’s control, the indicator is narrowed to precisely those variables. The measurement architecture therefore fails to see transfers between units, not by accident but by design.

That narrowing is not in itself a defect. Under a defined set of conditions it operates as a rational shortcut that lowers coordination cost. In a business where interdependence is low, capacity is abundant, and lead times are predictable, the local indicator makes delegation possible, renders accountability legible, and removes the need to escalate every trade to a higher tier. The difficulty lies not in the existence of the shortcut but in its persistence after the conditions that justified it have changed: once a bottleneck forms in capacity, once supply narrows to a single qualified source, or once schedule float begins to erode, interdependence between units rises by an order of magnitude, and the distance between the local optimum and the system optimum widens in the same proportion.

What keeps that widening invisible is the divergence between the direction in which cost travels and the location at which measurement occurs. Local optimization rarely destroys value; more typically it relocates value from one line to another, and because the relocation goes unrecorded, it registers as a saving on the first line and as an unexplained variance on the second. Production planning that increases batch size to spread setup time across more units simultaneously expands work-in-process inventory and warehouse demand; a quality function that tightens acceptance criteria generates delivery slippage and price revision on the supplier side. In both cases the indicator of the deciding unit improved, while the consideration accumulated in a line item belonging to a neighbor.

The first surface on which the institutional cost becomes visible is the balance sheet rather than the income statement. Taking a procurement discount against a larger lot slows inventory turns and lengthens the cash conversion cycle; the saving appears as a single line in the quarterly report, while the cash committed is concealed in the level of the inventory account relative to the prior year. The same structure produces a consequence on the financing side as well: expanding working capital raises revolver utilization, narrows headroom under covenant headings, and, in a cyclical contraction, pulls the renegotiation table forward in time. Not one link in that chain appears on the procurement scorecard.

In project-based work the cost migrates instead to the contract surface. The schedule float a package owner consumes in order to hold a cost target is in fact the shared reserve of the entire programme; once it has been spent inside a single package, subsequent packages are left to compress their own durations, add shifts, or pay acceleration premiums. The contractual expression of this appears, predictably, under three headings: approach toward the liquidated damages cap, growth in change-order volume, and the opening of interface responsibility to dispute between the parties. Where an interface has no owner defined in the contract, the owner becomes, in practice, whichever party produced the delay last, and that assignment rests on negotiating position rather than on technical merit.

The third surface is valuation. An acquirer or an investment committee seldom accepts a target’s reported unit-cost savings at face value; the quantities actually normalized are cash conversion cycle, inventory turns, rework rate, and the variance of delivery performance. A business with strong functional indicators and weak end-to-end indicators produces two foreseeable outcomes at the diligence table: a portion of the price is shifted into an earn-out structure or a post-closing adjustment mechanism, and integration risk is priced through a discount to the multiple. What determines valuation, in the end, is not performance itself but the demonstrable claim that performance is reproducible independently of any single function.

This tendency is neutralized through institutional architecture rather than individual resolve, and the intervention has four separable components. The first is end-to-end metric ownership: the elapsed time from order to delivery, or from FID to commercial operation, together with the cash committed across that interval, requires a single owner whose indicator is not partitioned across functional scorecards. The second is the interface register — a written entry, made at the moment of the decision and carrying an estimated magnitude, recording which decision transfers which burden onto which adjacent line. The third is holding the decision record at proposal rather than at approval, since the trade is made when alternatives are eliminated, not when the surviving option is signed. The fourth is a review rhythm with cross-functional authority present: no improvement in one unit’s indicator is approved while the representative of the affected unit is absent from the table.

The BEIREK intervention on this problem begins by establishing a single schedule and a single cost ledger for the project, with package owners maintaining their own programmes as derivatives of that master structure rather than the reverse. On that foundation an interface matrix is operated: the delivery, data, site access, and approval dependencies between each pair of packages are named, each dependency carries a date and an owner, and no interface left without an owner is treated as having advanced in the programme. Trades among the procurement, engineering, and construction lines are held in a separate trade-off log, so that when a unit-cost improvement is proposed, the schedule float or committed cash assumed in exchange, and by which line, is written on the same page.

The second layer is rhythm. When the monthly review ceases to be a ceremony in which units present their indicators in sequence and becomes a session devoted exclusively to interfaces and trades, the subject of discussion shifts from performance to transfer. On every decision touching the critical path, a written opinion is sought from at least one adjacent line other than the proposing unit; where that opinion is unfavorable the decision does not halt, but it is recorded together with its reasoning and compared, in the following quarter, against the outcome actually realized. Over time that comparison generates the organization’s own trade-off calibration, and institutional memory becomes independent of individual tenure.

For a management team the operative question is not whether the indicators are good, but which indicator is measured on whose line and settled on whose line. Where that question cannot be answered in writing, the organization is in all likelihood producing transfer rather than improvement, and the invoice for transfer is typically presented at the point noticed last — in the closing schedule, in working capital, or in the multiple applied at the diligence table.