--- title: "When a Measure Becomes a Target: The Quiet Degradation of Performance Indicators" description: "Once a measure is attached to an incentive, the organisation locates the cheapest route to improving it, and that route rarely runs through the underlying performance. The link between indicator and reality therefore weakens after a target is set. The neutralising mechanism is not individual integrity but paired counter-measures, a documented evidence chain, and periodic recalibration of the indicator itself." url: https://www.beirek.com/en/blog/goodharts-law-in-corporate-metrics canonical: https://www.beirek.com/en/blog/goodharts-law-in-corporate-metrics published: 2025-04-24 modified: 2025-04-24 category: "Organisational Psychology" category_url: https://www.beirek.com/en/blog/category/organisational-psychology 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: ["Goodhart's law","KPI degradation","performance measurement governance","incentive design","due diligence evidence chain"] topics: ["Performance indicator design and incentive structures","Governance architecture for measurement systems","Valuation impact of unverifiable KPI series","Project-level performance reporting discipline"] alternate_language_url: https://www.beirek.com/tr/blog/goodharts-law-in-corporate-metrics --- # When a Measure Becomes a Target: The Quiet Degradation of Performance Indicators > **In short:** Once a measure is attached to an incentive, the organisation locates the cheapest route to improving it, and that route rarely runs through the underlying performance. The link between indicator and reality therefore weakens after a target is set. The neutralising mechanism is not individual integrity but paired counter-measures, a documented evidence chain, and periodic recalibration of the indicator itself. *An indicator that has served a company reliably for years begins, from the moment it is tied to a bonus, a promotion, or a budget allocation, to reflect not the thing it measures but the behaviour directed at it. The drift is slow, quiet, and invisible in the reporting pack; the cost typically accumulates during the period in which the indicator looks its best.* --- In a board presentation, an operational indicator that has improved steadily for three years is usually met with a single question: whether the improvement is sustainable. What often goes unasked is whether the process behind the indicator has improved at all, since over the same period the actual output of that process — the quality of delivered work, the volume of customer complaints, the rework rate — may show no corresponding movement. The divergence between the indicator and the reality it purports to represent is not produced by a single decision; it accumulates through dozens of small operational choices, each of them defensible in its own context. A sales team pulls an order forward by a week to protect the quarter-end close; production planning releases low-margin stock early to preserve the turnover ratio; a field team logs a complex request as a separate ticket so that average resolution time remains intact. None of these breaches a rule, and all of them conform precisely to the definition of the measure. What these behaviours share is that the indicator itself has been converted into a target. When a measure is first installed, it is generally a passive instrument of observation: it watches a process from outside and determines no one's behaviour directly. At that stage the link between the indicator and the underlying performance holds firm, for the simple reason that no one has any motive to prise the two apart. The link begins to loosen at the moment a consequence is attached to the measure — a bonus, a promotion, a budget allocation, a decision to wind down a unit — because from that point there are two routes to improving the number, and one of them is materially cheaper than the other. Goodhart's law — the proposition that a measure ceases to be a good measure once it becomes a target — names this mechanism. At its core the law carries no moral claim; the claim it carries is entirely economic. Any indicator is a narrowed representation of the reality it attempts to capture, and that narrowing is not a defect but a precondition: a representation that did not narrow could not be operated. When an incentive is fastened to the indicator, the organisation discovers the gap between representation and reality and moves through it rationally. Degradation follows not from institutional stupidity but from the opposite — an accurate reading of the incentive structure as it has actually been written. Recognising the conditions under which this tendency remains functional matters, because a remedy built on the wrong diagnosis will be installed in the wrong place. Where the reality represented by the measure is narrow and well defined, where the process is short-cycle, and where the cost of manipulating the indicator exceeds the cost of genuinely improving performance, attaching a target works efficiently; first-contact resolution in a service centre or defect density on an assembly line typically falls into this category. The difficulty arises where the indicator is bolted onto a multi-layered, long-horizon process whose output is realised with a lag — project profitability, customer lifetime value, maintenance discipline, institutional reputation. In these domains the distance between measure and reality is wide, and the return on the effort required to close that distance appears long after the measurement window has closed, which makes the shortcut both cheaper and, in the short run, risk-free. The first layer of institutional cost accumulates not in the indicator's own line item but in the ones adjacent to it. An inventory policy tied to a turnover target will improve the stock line while inflating expedited procurement premiums on the supply side, split-shipment costs in logistics, and unfulfilled demand on the commercial side; the sum of the three frequently exceeds by several multiples the amount recovered in inventory, yet because the three sit on different lines under different managers, the aggregate is never displayed anywhere. The same pattern surfaces as schedule compression in a construction programme tied to a progress-billing target, as a lengthening tail in the receivables ageing table in a commercial structure tied to sales volume, and as the artificial splitting of files in an approval process tied to average handling time. The second layer is the erosion of the measure's own credibility, and for management this layer is the more expensive of the two. A degraded indicator does not merely convey wrong information; it conveys wrong information in a reassuring form, since the series is smooth, the trend is favourable and the reporting is disciplined. The board allocates capital against that series, sets capacity on the strength of it, and runs cross-unit comparisons through it, so that an entire sequence of decisions comes to rest on a compromised signal. The resulting decline in decision quality does not manifest as one bad call but as a chain of mutually reinforcing ones, and unwinding that chain back to its origin generally consumes something close to a full budget cycle. The third layer becomes visible on the capital markets side. What determines a company's valuation is often not the indicator series itself but whether that series can be corroborated by independent evidence. What a buyer's team actually does in diligence is not to accept the KPI but to test, on a sampled basis, the link between the KPI and the underlying transaction record; instances where the definition shifted mid-period, where scope was quietly narrowed, where the exception category was widened, or where the definition was never documented at all tend to surface quickly under that test. The outcome is rarely abandonment of the transaction. The outcome is a discount applied to the valuation multiple, an earn-out structure deferring consideration past closing, an expanded scope of representations and warranties, or a higher escrow percentage. Put differently, a degraded measure exacts its price not in the company's operations but in the economics of the share purchase agreement. The structural intervention is not built by removing the indicator, nor by demanding greater personal integrity from managers; this tendency is governed by institutional architecture rather than individual will. A workable intervention has four components. The first is counter-measure pairing: alongside every indicator tied to an incentive sits a second indicator that the shortcut behaviour would necessarily damage, with the award conditioned on both being satisfied together. The second is the definition record: the measure's definition, scope, exclusions and calculation method are fixed in writing, and any change to that definition is permitted only through a versioned decision taken at the start of a period. The third is the evidence chain: the link between the indicator and the raw transaction record that generates it is tested by regular sampling. The fourth is the recalibration rhythm: each indicator is reassessed at defined intervals against the question of whether it still measures what it was installed to measure, and retired where it does not. BEIREK's intervention in capital-intensive projects consists of establishing this architecture at the project level. When a project's performance indicators are being defined, the behaviour each indicator will reward and the line on which that behaviour will generate cost are mapped in advance; progress-billing tempo, schedule advancement, quality rejection rate and supply delivery performance are positioned on a single table as counter-measures to one another, so that any improvement on one line reveals within the same reporting cycle whether it has been transferred from another. A definition record is maintained for every indicator, and no mid-period change of definition is implemented without being logged together with its rationale and its effect; that record later becomes the primary document carrying the verifiability of the series when lender reporting or a sale process comes onto the agenda. The second mechanism is the placement of a calibration session inside the decision rhythm itself. In the project management office's periodic review, the indicators are on the agenda alongside their values: which measure improved unexpectedly in the last period, what operational change accounts for that improvement, and whether the account given is consistent with the evidence observed in the field. In that session the role of asserting that a measure has degraded is assigned explicitly to a named individual, and the role rotates. Absent an institutionally protected counter-argument position, the individual cost of questioning a series that looks good remains systematically higher than its benefit, and it is therefore not questioned. The managerial difficulty with measure degradation is that it is least visible precisely when it most requires attention: when an indicator deteriorates, everyone examines it, whereas when it improves, a request to examine it reads as a declaration of distrust. What breaks that asymmetry is converting the examination from an outcome-triggered reaction into a calendar-driven routine. The real question to ask about a company's measurement system is not whether its indicators are currently correct, but whether a mechanism exists that would notice the moment they began to stop being so. ## Key Points - The moment an indicator is attached to an incentive it ceases to function as an instrument of observation and becomes an object of negotiation; the degradation arises from the structure of the incentive rather than from bad faith. - Measure degradation typically begins during the period in which the indicator is improving, which is precisely why a well-behaved KPI should not be demoted in audit priority. - An incentive system anchored to a single indicator will, predictably, transfer value out of the domains the indicator does not cover — maintenance discipline, quality, customer relationships, institutional memory. - In diligence, the distance between a reported metric and its underlying evidence chain feeds directly into the valuation multiple; a KPI series that cannot be substantiated produces a discount rather than a rejection. - Neutralisation does not mean removing the indicator; it means embedding a paired counter-measure and a recalibration rhythm into the governance architecture itself. ## Questions ### What is Goodhart's law, and how does it appear inside companies? A measure ceases to be a good measure once it becomes a target. Inside companies this appears after a bonus, promotion or budget allocation is attached to an indicator, at which point the cheapest route to improving that indicator stops running through the underlying performance. The observable result is an improving series accompanied by real output that remains flat or deteriorates. ### How can it be established whether a KPI has degraded? The most reliable signal is an improvement that cannot be attached to an independent operational explanation. Where the measure improves while adjacent items — complaint volume, rework rate, the tail of the receivables ageing table — deteriorate, the improvement is most likely a transfer of value rather than a gain. A second signal is any quiet mid-period change to the definition, the scope, or the exception category. ### How does measure degradation affect company valuation? A buyer's diligence team does not accept the KPI series; it compares that series, on a sampled basis, against the raw transaction record that produced it. Where the definition shifted mid-period or the evidence chain cannot be reconstructed, the consequence is typically not termination but a discount to the valuation multiple, an earn-out structure, an expanded scope of representations and warranties, or a higher escrow percentage. ### Can this degradation be prevented without abandoning performance targets? Four components are generally sufficient: pairing every incentivised indicator with a counter-measure that the shortcut behaviour would damage; recording the measure's definition in versioned form and barring mid-period amendment; testing the link between the indicator and the raw transaction record through regular sampling; and reassessing each indicator at defined intervals against whether it still measures the intended thing. --- Source: https://www.beirek.com/en/blog/goodharts-law-in-corporate-metrics Publisher: BEIREK LLC — https://www.beirek.com