When the Incentive Manufactures the Problem: The Cobra Effect and the Architecture of Measured Behaviour

In most companies, a final ten days of the quarter in which sales volume runs materially above the monthly average is recorded as ordinary end-of-period activity; yet where the same window shows discount rates climbing, payment terms lengthening, and order flow in the first three weeks of the following quarter falling below trend, what is being observed is not demand but the calendar of the commission threshold. The pattern repeats regardless of how disciplined the sales organisation is, because the quantity being measured is volume, and the cost of pulling volume forward — some margin surrendered, some receivable extended — remains lower than the cost of missing the accrual. The manufacturing equivalent is familiar to anyone who has set a scrap-rate target and then watched not the scrap itself change but the manner in which scrap enters the record: the part is routed to rework, rework is tracked under a separate line, and the target indicator holds. In both configurations the system has received exactly what it asked for, and exactly what it did not want has occurred.

The institutionally expensive version of this behaviour appears wherever the indicator the incentive measures feeds directly into the problem the incentive was created to solve. Where a maintenance function is assessed on the number of fault tickets closed, the return on root-cause analysis falls, and recurring failure becomes, in effect, a revenue line operating in the function's favour rather than a defect it is paid to eliminate. Where a collections function is assessed on the value of receivables restructured, generating restructurable receivables — which is to say, selling to counterparties likely to encounter payment difficulty — becomes the behaviour the system rewards, and the credit committee's caution becomes a friction the sales organisation learns to route around. Wherever the magnitude of the problem and the magnitude of the reward are wired in the same direction, the complete elimination of the problem serves no one's interest inside the system.

The mechanism carries a name — the cobra effect, describing an incentive structure that measurably enlarges the problem it was designed to reduce — and it operates through two steps that are worth separating. The first is that no incentive system can measure the objective itself: profitable and sustainable growth, safe operation, durable customer relationships are not directly countable, so every system substitutes a proxy — revenue booked, tickets closed, average handling time, incidents reported, restructured value. The second is that from the moment the proxy is defined, it acquires an existence independent of the objective and becomes improvable along two distinct routes, one running through genuine improvement of the objective and the other running through improvement of the proxy alone. Where the second route costs less than the first, the organisation takes the second route with high predictability, and that choice reflects the cost logic embedded in the system rather than any individual moral disposition.

Recognising that the tendency is not an error is the precondition for managing it. Using a proxy is a rational answer to a genuine tension between the cost of measurement and the speed of decision-making; running a sales organisation weekly against customer lifetime value is not practically feasible, whereas running it against booked revenue is. The problem lies not in the shortcut but in the silent change of the condition under which the shortcut held. Revenue is a sound proxy while the relationship between revenue and contribution remains stable; the moment discount authority is devolved to the field, the product mix broadens, or competition sharpens on price, that relationship breaks and the same indicator begins to measure the inverse of what it measured a year earlier. What determines the useful life of an incentive scheme is therefore not the elegance of its design but the duration over which its underlying condition stays constant.

The institutional cost accumulates first in an unexpected part of the balance sheet. Where a commission structure built on booked revenue coexists with discount and payment-term authority held by the same population, the consequence surfaces not in the income statement but in receivable turnover and the cash conversion cycle; the company tightens on cash while it is growing, the tightening is read as a financing requirement, and the causal link between the commission plan and the working-capital facility never appears side by side in any management report. On the production side, the scrap target migrates into indirect labour as rework hours; on the health and safety side, an incident-reporting target expresses itself not in the insurance premium but, several years later, in a single claim file arising from an event that was never reported. What these line items share is that they reach multiples of the incentive budget while leaving no trace within the accounts that can be attributed back to the incentive.

The second layer of cost accumulates in the quality of revenue rather than its quantity. A revenue base composed of discounted, extended-term transactions concentrated at quarter-end may appear healthy in aggregate while remaining weak in repeatability, because the customer decided on the basis of a pricing window rather than a relationship, and when the window closes the demand closes with it. When an acquirer examines that picture, the intra-quarter shape of monthly revenue, the periodic volatility of the discount rate, and the trend in collection days point to the same conclusion from three independent directions. Revenue of this character is not corrected on the valuation desk through normalisation adjustments; either the multiple itself is pulled down, or a portion of consideration is moved into an earn-out structure conditioned on demonstrating repeatability after closing. The commission plan is itself a data-room document, and experienced diligence reads it not as a compensation schedule but as a behavioural map of how the revenue was produced.

A third layer accumulates in institutional memory, and it is the layer least visible from the reporting pack. An organisation that has spent years optimising a proxy becomes dependent, over time, on the individuals who know how the proxy is optimised — which customer to approach in which week, which cost belongs in which record, which threshold is met by which route — and that knowledge sits nowhere in writing, residing entirely in people. Because it cannot be transferred, the departure of a key salesperson or a key production manager produces a deterioration in the indicator that no one can explain from the documentation. A meaningful share of situations diagnosed as founder dependency turn out, on inspection, to be dependency not on the founder's presence or judgement but on an incentive system that has been administered person by person rather than by rule.

The mechanism that neutralises this tendency is built through measurement architecture rather than individual awareness, and it has four components. The first is the paired counter-indicator: no incentive metric is ever defined alone, and each is accompanied by a second metric that deteriorates precisely when the first is being gamed — revenue paired with gross margin, volume paired with days sales outstanding, tickets closed paired with the reopening rate on the same equipment, handling time paired with first-contact resolution. The second is the extension of the measurement window to the period over which the behaviour's consequences mature, so that where part of the commission attaches to cash collected rather than invoices issued, and a further part to second-year renewal, the economic attraction of pulling volume forward disappears on its own. The third is the separation of authority, under which discount approval, term approval, and commission accrual do not converge in the same person. The fourth is a revision cadence, under which the conditional assumptions underlying the plan — above all the assumption that the proxy still represents the objective — are tested once a year at a table independent of the one that designed the plan.

The way BEIREK addresses this layer in capital-intensive projects and portfolio transformations begins by removing the incentive structure from the compensation file and placing it inside the project control architecture, where it belongs. When progress payments in an investment programme are tied to physical completion percentages, the cheapest optimisation route available to a contractor is to front-load the work packages that are easiest to measure; the result is a project that tracks against the schedule on paper while the critical-path items required for commissioning fall behind, and the divergence becomes visible only when commissioning is attempted. For that reason we construct the payment structure on mutually checking indicator pairs rather than a single progress metric, tie payment thresholds to deliverables sitting on the critical path, and track the divergence between the works programme and the cash curve in a separate record maintained independently of the liquidated-damages cap in the contract.

The second intervention concerns the moment at which reasoning is captured, which is the point of proposal rather than the point of approval. When an incentive structure or a payment schedule is proposed, a single-page record states which proxy represents which objective, under what condition that representation holds, and at what threshold the structure will be reopened should the condition change; because the record is written before the plan produces results rather than afterwards, it cannot be rationalised retrospectively to fit whatever the results turn out to be. A fixed question is then added to the monthly project rhythm: among the indicators that improved this period, which may have improved without the objective it represents having improved at all. In most months the answer comes back empty, and in the month it does not, the amount preserved comfortably exceeds the cost of maintaining the rhythm for the year.

An incentive system is the most honest sentence an institution addresses to itself, and where statements of intent and the payment schedule contradict one another, the organisation listens, with entire predictability, to the payment schedule. The operative question is therefore not whether a given plan is fair, but whether the indicator it measures still represents the objective it was chosen to stand for, and by whom, at which table, the expiry of that representation is being monitored.