The Goodhart Trap: When Metrics Lose Meaning

Measurements allow us to determine the dimensions of tangible objects, and repeated measurement authenticates these results. Nearly every aspect of life can be measured. In modern commercial and social settings, progress is often assessed through quantifiable data. 

Our reliance on measurements extends to evaluating whether sacred numbers align with our buildings, homes, and places of worship.  Manipulating these measurements is often viewed as a serious transgression. As a result, measurements remain a non-negotiable standard in society. 

Measurements are deeply ingrained in our culture and daily lives. Our familiarity with certain numbers leads us to quick conclusions, and we often shape our actions and judge outcomes based on these metrics, viewing them as indicators of success or failure.

When looking at measurements, it is important to consider sub-measurements that feed into the final measurements. What are those that are vital to our business, vital to our programming, or vital for our long-term sustenance? The main metric consists of sub-metrics. These sub-metrics alone will give a wider picture of what is being constituted as the final measurement. So, is it the final number that has emerged alone that becomes the opinion creator?  Or, is it the sub-metrics that also need attention?  We must also draw attention to many intangible measurements that give the final metric its status. This makes the final single metric that many of us consider vital, sceptical, too. 

Social scientists note that people often optimise their actions to achieve specific metrics, focusing on the target rather than the underlying standard. As a result, the desire to 'win' can overshadow the true value of the measurement. 

One of the most interesting observations in measurement was made by the British economist Charles Goodhart in 1975. He observed that “When a measure becomes a target, it ceases to be a good measure.” Simply put, the more we push ourselves toward the final metric, the more we may ignore the macro-world of many sub-metrics that might contribute to it. Goodhart's Law teaches us that any metric used as a target may be gamed, manipulated, tunnelled, or distorted, undermining its usefulness as a genuine measure of success. 

This issue arises in company valuations as well. A retail investor cannot rely solely on the final balance sheet to understand a company's true condition. In a data-driven world, even advanced tools like AI can produce distorted results due to a narrow focus. Recently, the Securities and Exchange Board of India (SEBI) announced caps on mutual fund expense ratios to reduce investor costs. However, as Goodhart's Law suggests, once expense ratios become the target, fund houses may shift costs to other areas such as distribution fees, advisory charges, or hidden operational expenses. As a result, the expense ratio may no longer accurately reflect the investor's true costs. 

Banks dress their NPAs (non-performing assets) and CARs (capital adequacy ratios) through loan restructuring and regulatory arbitrage. The net result is that the bank’s performance becomes cloudy to stakeholders (shareholders) and the regulators. AI, which often relies on neural networks and back-propagation to minimise output errors, does not guarantee convergence to correct solutions. As a result, AI can produce inaccurate conclusions in areas such as credit scoring and customer service data analysis. 

Falling into the Goodhart Trap is common. Metrics are powerful tools, but when turned into rigid targets, they often lose their meaning. It’s a principle that resonates across economics, business, education, and even personal goal-setting. Just because we can measure something does not give us a clear picture of what is being measured. 

FINE