StockSignal
How optimizing for measurable targets distorts the metrics themselves.
Introduction
Metrics serve a dual purpose in business: they describe the current state of the business and they provide targets for management to optimize against. The tension between these two purposes is the structural problem that Goodhart's Law describes. When a metric is used purely as a descriptor — revenue growth, customer satisfaction, employee retention — it provides reliable information about the business's condition. When the same metric becomes a target that determines compensation, capital allocation, or strategic direction, the people responsible for the metric develop incentives to improve it through any available means, including means that improve the number without improving the underlying reality.
A call center that measures performance by average call duration creates an incentive for agents to rush calls, resolving them quickly but not thoroughly. A hospital that measures performance by patient mortality rates creates an incentive to avoid treating the most severely ill patients. A company that measures performance by quarterly earnings per share creates an incentive to manage earnings through accounting choices, share buybacks, and short-term cost cuts. In each case, the metric that was intended to capture quality becomes a target that degrades quality while appearing to improve it.
Understanding Goodhart's Law structurally means examining why the transformation from measure to target creates distortion, how this distortion manifests in business contexts, and what it implies for the reliability of the metrics most commonly used to evaluate companies.
Core Concept
The distortion occurs because any metric is a simplification of the reality it represents. Customer satisfaction as a number cannot capture the full complexity of customer experience. Revenue growth as a percentage cannot capture the quality and sustainability of the growth. When people optimize for the simplified metric, they find ways to improve the number that do not require improving the complex reality. These optimization strategies exploit the gap between the metric and the reality — the aspects of reality that the metric does not capture become the areas where quality is sacrificed to serve the number.
The problem is structural, not behavioral. Many managers believe they are genuinely improving performance when they optimize metrics. But the metric and the reality are not the same thing. Any effort that moves the metric without moving the reality creates a divergence that accumulates over time. The longer the metric remains the target, the more the divergence grows, and the less reliable the metric becomes as a description of reality.
Multi-metric approaches reduce but do not eliminate the problem. A company that measures both revenue growth and customer satisfaction is harder to game than one that measures only revenue growth, because optimizing one at the expense of the other is more visible. But any finite set of metrics leaves gaps that can be exploited, and the addition of more metrics creates complexity that can itself be gamed — managers learn to optimize the combination of metrics in ways that do not reflect the holistic performance the metrics collectively attempt to capture.
The temporal dimension amplifies the distortion. Short-term metrics are easier to manipulate than long-term ones because the time horizon for validation is shorter. A quarterly earnings target can be met through actions that harm next quarter, but the harm is not visible until the current target has been achieved. This structural bias toward short-term optimization at the expense of long-term health is one of the most consequential manifestations of Goodhart's Law in business.
Structural Patterns
- Metric Decay — The reliability of a metric as a descriptor declines the longer it is used as a target. Early in its use, the metric may accurately reflect performance. Over time, optimization strategies erode its informational content as managers learn to improve the metric without improving the reality.
- Proxy Gaming — When the actual desired outcome cannot be directly measured, proxy metrics are used. The proxy's correlation with the actual outcome may be strong initially but weakens as optimization strategies exploit the gap between the proxy and the outcome.
- Campbell's Law Extension — The more important the consequences attached to a metric — compensation, promotion, capital allocation — the greater the incentive to distort it. Metrics with high-stakes consequences are more aggressively gamed than metrics used for informational purposes only.
- Cobra Effect — Incentive systems designed to solve problems can create new problems that are worse than the original. Named after the colonial-era bounty on cobras that led to cobra breeding, this pattern appears when the metric-improvement strategy has perverse side effects that the metric does not capture.
- Teaching to the Test — When evaluation criteria are known in advance, optimization focuses narrowly on satisfying the criteria rather than developing the broad capability the criteria were meant to assess. The result is performance that meets the criteria but does not reflect genuine competence.
- Informational Degradation Spiral — As metrics become less reliable due to gaming, decision-makers add more metrics or more complex metrics, which creates additional gaming opportunities and further reduces informational quality. The system produces more data and less insight.
Examples
Earnings per share management is the most consequential manifestation of Goodhart's Law in corporate finance. When executive compensation is tied to EPS growth, management has structural incentives to increase EPS through any available means — not just through genuine business improvement, but through share buybacks that reduce the denominator, through accounting choices that shift expenses across periods, and through short-term cost cuts that boost current earnings at the expense of future competitiveness. The EPS number improves, but the improvement may not reflect any change in the business's underlying earning power.
Customer satisfaction scores demonstrate the law in operational contexts. When bonuses depend on survey scores, frontline employees learn to influence the measurement process — requesting high ratings from satisfied customers, discouraging dissatisfied customers from completing surveys, or providing small concessions that improve scores without resolving underlying service issues. The measured satisfaction rises while the actual customer experience may remain unchanged or even deteriorate in areas the survey does not capture.
University rankings illustrate the law in institutional contexts. When a university's reputation and funding depend on its ranking, administrators optimize for the specific metrics the ranking system uses — faculty-to-student ratios, spending per student, alumni giving rates. Resources flow toward improving ranking inputs rather than toward educational outcomes that the ranking may not capture. The ranking improves, but the education may not, and the resources diverted to ranking optimization are unavailable for their original purposes.
Risks and Misunderstandings
A common error is assuming that better metrics solve the problem. While better-designed metrics can reduce some gaming opportunities, any metric that becomes a target will eventually be optimized in ways that exploit the gap between the metric and the reality. The structural problem is not the quality of the metric but the transformation of a measure into a target.
Another misunderstanding is treating metric gaming as exclusively dishonest behavior. Much gaming occurs through legitimate means — choosing among permissible accounting methods, timing discretionary expenditures, prioritizing activities that are measured over those that are not. These choices are rational responses to the incentive structure, and they occur even among managers with genuine integrity. The distortion is a property of the system, not just of the individuals within it.
It is also tempting to abandon metrics entirely in response to Goodhart's Law. While the law reveals the limitations of metric-based evaluation, the alternative — evaluation without metrics — introduces its own problems of subjectivity, inconsistency, and bias. The appropriate response is not to eliminate metrics but to use them with awareness of their limitations, to rotate or evolve metrics before they decay, and to supplement quantitative measures with qualitative judgment.
What Investors Can Learn
- Scrutinize metrics that determine compensation — When management is compensated based on specific financial metrics, those metrics are structurally likely to be optimized in ways that may not reflect genuine business improvement. Examine the means by which the metrics are being achieved, not just whether they are being achieved.
- Look for divergences between related metrics — When one metric improves while related metrics do not — for example, revenue growing while cash flow stagnates, or EPS growing while total earnings decline — the divergence may indicate that the improving metric is being gamed rather than genuinely earned.
- Assess the stability of metric relationships over time — If the historical relationship between a metric and business outcomes weakens over time, the metric may be losing its informational value due to optimization pressure.
- Value qualitative assessment alongside quantitative metrics — Metrics provide useful information but are incomplete descriptions of business reality. Qualitative factors — management candor, organizational culture, strategic coherence — provide information that metrics cannot capture and are harder to game.
- Be wary of companies that emphasize a single metric — Companies that organize their narrative around a single performance metric create concentrated incentives to optimize that metric. Businesses that present a balanced view of performance across multiple dimensions are less likely to be distorting any individual measure.
Connection to StockSignal's Philosophy
Goodhart's Law reveals a structural limitation in metric-based evaluation — the act of measuring changes what is being measured when the measurement carries consequences. Understanding this dynamic is essential for interpreting the financial metrics that form the basis of business analysis. This focus on the reliability of the information itself, and the structural forces that can degrade that reliability, reflects StockSignal's approach to understanding businesses through awareness of the systemic dynamics that shape what is observable and how reliably it represents the underlying reality.