Where has Goodhart’s law turned your well-intentioned metric into a liability? If you lead a sustainability program, a reporting function, or an executive team that sets targets, the honest answer is probably: somewhere on your board slide already. You just haven’t found it yet.
The law
In 1975, Charles Goodhart, an economist advising the Bank of England, observed that any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes. The anthropologist Marilyn Strathern later sharpened his point into the sentence everyone quotes: “When a measure becomes a target, it ceases to be a good measure.”
Goodhart was writing about monetary policy. He could have been writing about sustainability, a field that runs on measures turned into targets: percentages, intensities, scores, and claims. And in sustainability, the collapse of the measure is no longer just an embarrassment. Since September 27, 2026, in the European Union it can be an illegal claim.
The percentage fell; the pile grew
Consider a company that reduces the virgin plastic in each unit it makes by 30% over ten years, and reports the reduction as a triumph. The report is accurate. The conclusion is not.
Do the multiplication nobody did. At 5% annual growth, ten years compounds to 1.63 times as many units. Per-unit plastic fell to 0.70 of its starting point. Multiply the two: total virgin plastic is up roughly 14%. The percentage fell. The pile grew.
Notice that nobody lied, and nobody gamed anything. The team hit its target honestly. The target was per unit, and the denominator moved. The denominator is the count underneath the percentage, the number of units the per-unit figure divides by, and at 5% growth it does not sit still. This is the form of Goodhart’s law that should worry executives most: not the cynical kind, where people cheat the number, but the innocent kind, where the number is accurate, audited, and faithfully reported, while the thing it was meant to describe goes the other way.
The target was per unit. The planet counts units.
The measure was never the point
A recycled-content target asks how much recycled polyester is in the jacket. A better question asks what the jacket is for. A designer who did the arithmetic on cost per wear, a €30 jacket worn five times costs €6 per wear; a €300 jacket worn three hundred times costs €1, reframed the problem entirely: this jacket needs to survive 300 wears. One question optimizes an input. The other examines the assumption underneath it.
Or take offset-based “carbon neutral” claims. The metric performed beautifully for a decade: targets set, offsets purchased, claims printed. The assumption underneath, that paying for reductions elsewhere is the same as reducing your own emissions, was never examined. The EU has now banned the claim outright. A target can be met, reported, and audited for years, and still end as a liability, because the belief it rested on was never true.
And there is no final, incorruptible metric waiting at the end of this search. Cost-per-wear can be gamed the moment it becomes a target, just as recycled content was. That is the law’s real lesson. The answer is not a better measure. The answer is examining what your measures assume.
Goodhart is the symptom; the belief is the disease
The standard telling of Goodhart’s law blames gaming, people optimizing the number instead of the work. In nearly fifty years inside this field, I have found the gaming to be the least of it. The measure usually fails earlier, at the moment it is chosen, because it is chosen inside a belief nobody has examined.
The per-unit target assumed the denominator would sit still. The recycled-content target assumed the material was the problem. The offset claim assumed someone else’s reductions were interchangeable with your own. Every metric on your dashboard is a belief wearing a number, and the belief, not the number, is where the risk lives.
This is also why the problem is about to get faster. The AI systems now drafting strategies, disclosures, and sustainability reports were trained on our reporting, decades of it, built on the same unexamined assumptions. Ask a model to help you hit the target and it will optimize against the gamed measure fluently, confidently, and at machine speed. Goodhart’s law used to operate at the speed of an annual report. It now operates at the speed of generation.
Audit your own metrics: five steps
- List the headline metrics. The handful of numbers on the board slide, the ones your sustainability performance is judged by.
- Name the denominator. For every percentage, ratio, and intensity figure, ask what it assumes will hold still: unit volumes, product mix, revenue, headcount. Denominators move. That movement is where the gap hides.
- Ask what the target rewards. Every target rewards some behavior and quietly punishes some other. If your people are rewarded for the number, what are they being punished for noticing?
- Follow the material. Compare each metric with the physical quantity it claims to describe: units, tons, wears, years of use. When the metric and the material disagree, believe the material.
- Ask the belief question. For each metric: what had to be assumed for this to be the right measure? Who examined that assumption, and when?
Steps one through four find discrepancies. Step five finds the cause. It is the step almost everyone skips, and the only one that prevents the next metric from failing the same way.
The diagnostic
Step five, done properly, is a discipline rather than a checklist. It is the work of the Belief Gap Diagnostic: a structured examination of the beliefs underneath an organization’s claims, targets, and AI-supported decisions, the distance between what the organization believes and what it can actually substantiate. That distance is the belief gap, and it is where regulators, journalists, and competitors will eventually look.
Your metrics are not lying to you on purpose. They are answering, faithfully, the questions they were built from. The question worth asking, before someone else asks it for you, is the one this post began with: what had to be assumed for that answer to be true?
*Ken Alston has spent nearly fifty years inside sustainability: twenty years at SC Johnson, seventeen alongside William McDonough and Michael Braungart, and nine years in independent practice. He keynotes on the assumptions hiding inside executive decisions, and runs the diagnostic that finds them before a regulator, a journalist, or a model does. Book an exploratory conversation →