When The Measure Becomes The Mission: Goodhart’s Law and The Sustainability Trap

I have spent a reasonable portion of my career sitting in rooms where someone eventually says “we need to define what success looks like” and everyone nods solemnly, as if the act of nodding is itself a measure of how seriously they are taking the question before someone opens a spreadsheet.

I remember one such room vividly, though I will describe it in terms general enough that nobody gets called, because that is not the point and also because, selfishly, I still want to be invited to future rooms. I was part of a function whose job was to create value between various stakeholders, like a pattern connector, within a large organisation, which sounds clean on paper and is, in practice, the kind of remit that is genuinely difficult to measure without thinking carefully about it first. And when organizations are under pressure, thinking carefully is the last thing on their minds, and knee jerk reactions are a norm.

The instinct in that room, as in most rooms of this kind, was to count things. We threw amazing words around, albeit in hindsight, I thought I was unstoppable at that age. Yikes! These fancy jargons included, amongst others, number of sales meetings initiated, number of proposals brought forward, number of business cards collected. I hesitated, not because I had a better answer ready, but because something in my gut said that counting meetings was going to teach us to have more meetings, not better ones, and that those were not the same thing. I was, as they say, outvoted by the spreadsheet and that was a beautiful experience in my life, that taught me when to open my mouth and when to keep it shut. But that silence had brewed a storm we didn’t anticipate what was to come, because as you guessed it, the metric became the target. The target drifted quietly away from the goal it was supposed to represent, which was value created for clients or customers; this was easier said than done, I agree! And I did not have a name for any of this until much later, when I stumbled across a sentence that I wish someone had taped to the wall of every room I have ever sat in; when a measure becomes a target, it ceases to be a good measure and the outcome becomes the measure.

Don’t You Hate It When You Learn Something Very Late

Goodhart’s Law, articulated by British economist Charles Goodhart in 1975 in a paper on monetary policy, though the underlying idea had been lurking in human behaviour long before anyone gave it a Latin-sounding name. The law states, in its most quoted form, that any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes. Translate the academic mumbo-jumbo to plain vernacular, it states that the moment you start measuring something and attaching consequences to the number, people stop optimising for the thing and start optimising for the number.

I mention this not to sound clever but because I genuinely did not know this was a named concept for an embarrassingly long time, and when I finally encountered it, my first reaction was more perturbed and of a ignorant facepalm moment. There is always a time to learn what you already knew. Funny to you, but for a know-it-all geek like me, that is equivalent of flat-lining on an exam.

The examples are everywhere once you start looking, and some of them are genuinely magnificent in how badly they went wrong. The British colonial government in India, attempting to reduce the cobra population in Delhi, offered a bounty for dead cobras. Enterprising locals responded by breeding cobras specifically to kill them for the bounty money. When the government discovered this and cancelled the programs, the breeders released their now worthless snakes, making the cobra problem significantly worse than it had been at the start. The metric, dead cobras delivered, became the target, and the target stopped having anything to do with the goal, which was fewer cobras in the streets. A similar program in French colonial Vietnam offered bounties for rat tails and produced, predictably, a thriving ecosystem of tailless rats being released back into the sewers to breed more tails for future bounty collection. Bureaucrats have been reinventing this particular mistake across centuries and continents with impressive consistency.

This would be funny if it stayed in the realm of colonial pest control. Oh lord are we in for a laugh!

From Story Points To Sustainable Development Goals: The Same Mistake, Bigger Consequences

I wrote a while back about the Story Point Conundrum in product management, where story points were invented to measure complexity and ended up being used to measure productivity, which turned every sprint planning meeting into a subtle competition to see who could estimate the highest without getting challenged, which is Goodhart’s Law in a Jira board. That particular dysfunction costs organisations time and money and the occasional good engineer who gets tired of the theatre.

[embed]The Story Point Conundrum Did you know, that story point only meant the complexity of the work and had nothing to do with actual estimation of…www.linkedin.com

But consider what happens when you apply the same dynamic to problems of genuinely civilizational scale.

The United Nations Sustainable Development Goals, adopted in 2015 and running to 2030, represent humanity’s most ambitious attempt to define what a better world looks like. 17 Goals, 169 Targets and 232 Indicators; the sheer numerical architecture of the SDG framework is, depending on your disposition, either a triumph of multilateral consensus or a masterclass in what happens when a committee tries to measure everything simultaneously and ends up measuring nothing with particular conviction. I will leave that to your disposition. What I will say is that the SDG Report 2026 found that meaningful gains have been made across the goals but remain uneven and insufficient in the face of escalating conflicts, slowing economic growth, climate change, rising debt burdens and declining official development assistance.

Somewhere in Geneva, there is almost certainly a person whose full-time job is to count whether the one hundred and sixty-nine targets are being counted correctly. They are probably also exhausted, and they are probably also being measured on something that is not quite the thing they are actually trying to achieve.

The structural flaw in the SDG framework, and I say this with enormous respect for the ambition behind it, is that it assumes all seventeen goals can be pursued simultaneously regardless of where a country sits on the most basic rungs of human development. A country cannot meaningfully pursue SDG 14, Life Below Water, if its population does not have reliable access to SDG 2, which is Zero Hunger, or SDG 6, which is Clean Water and Sanitation. Maslow understood in 1943 that human beings do not self-actualise while they are worried about their next meal. The SDGs, written by a committee of nations that largely have functioning food systems, occasionally forget that sequencing matters enormously, and that a metric which looks like progress at the aggregate level can conceal the fact that the countries most in need of the goal are the ones least able to pursue it while also surviving.

Carbon Credits: Goodhart’s Law Wearing A Green Hat

The carbon credit market is where Goodhart’s Law has had its most expensive and consequential recent outing, and it deserves its own paragraph rather than a footnote.

The logic of carbon credits is sound in principle. Companies that reduce emissions beyond what is required should be able to sell that reduction as a tradeable asset to companies that are struggling to reduce theirs, creating a market incentive for emissions reduction across the economy. The metric, verified emissions reductions expressed as credits, was designed to represent actual climate benefit. It became a target. And then it became a commodity. And then it became a creative accounting exercise.

More than 68% of DAX40 companies that purchased carbon credits ended up supporting projects with no real climate impact, with the Max Planck Institute finding that 84% of carbon credits are high risk. An oil and gas company can now purchase enough carbon credits from a wind farm in Rajasthan to declare itself carbon neutral on its annual report while continuing to extract petroleum at full capacity, which is Goodhart’s Law at planetary scale dressed up in a press release. The urgency of addressing greenwashing has reached a new peak in 2025, with HSBC’s decision to exit the Net-Zero Banking Alliance citing concerns over greenwashing exposure, and the UK’s Competition and Markets Authority beginning large-scale public enforcement of its Green Claims Code from Autumn 2025. Regulators are waking up to what Goodhart predicted in 1975 wherein, the metric stopped describing the reality a long time ago.

The ESG framework broadly has the same problem. The initial ESG boom from 2020 to 2023 functioned less as an environmental revolution and more as an exercise in re-labelling existing index funds to capture higher fees, with fossil fuel extractors remaining staples in supposedly sustainable portfolios because exclusion would have forced genuine index deviation and reduced fee income. The measure became the target, the target became the product and surprisingly, the product was sold to people who thought they were buying the goal.

Who Measures The Measurers, And Why Nobody Has Sorted This Out Yet

Here is where the argument gets interesting, and where I want to propose something that may or may not already exist in some form, because I checked and the answer is complicated.

The 2008 financial crisis has many fathers, but one of the most structurally important and least discussed is the rating agency problem. Moody’s and Standard and Poor’s were rating mortgage-backed securities during the housing bubble while being paid by the banks that were issuing those securities. Their metric was deal volume, because more deals meant more rating fees and hence their target became deal volume. And their ratings, which were supposed to measure risk, stopped measuring risk and started measuring the fee-generating potential of the relationship with the issuer. The result was a global financial system built on assets that were rated AAA and were, in practice, constructed from mortgages issued to people with no income, no job, and no assets. We have all seen the movie and unfortunately, some of us lived through the version that did not have a soundtrack.

The problem was not that rating agencies existed but who was paying them and what they were therefore incentivised to produce.

The same structural problem exists in impact measurement wherein when an organisation measures its own impact, it is, to put it gently, not a disinterested party. When a funder, usually a family office or a philanthropist, requires an impact report from the organisation it funds, the organisation has a clear incentive to produce a report that justifies continued funding, which means it will gravitate toward the metrics that look best rather than the metrics that are most true. For foundations, NGOs, and donors, data now plays a central role in philanthropy, with growing pressure to show results, count outcomes, and prove that funding is making a difference, but an important question is often left unasked; What if the way we measure impact is rooted in the same systems of inequality we are trying to change?

What does not yet exist at meaningful scale is a philanthropy-funded, operationally independent impact measurement body with the mandate, the methodology, and crucially the incentive structure to measure outcomes on behalf of the people funding the work rather than the people doing it. The Gold Standard exists for carbon projects, IRIS Plus exists for impact investors and the SDG Impact Assessment Tool exists for organisations wanting to self-report their contribution to the goals. But none of them have the structural independence, the enforcement teeth, or the funding model that would make Goodhart’s Law significantly less likely to eat them alive.

The parallel is not the rating agencies before 2008, rather what the rating agencies should have looked like, which is bodies funded by and accountable to the investors who needed accurate information, rather than the issuers who needed favourable ratings, with published methodologies, rotating mandates to prevent capture, and consequences for producing assessments that bear no relationship to underlying reality.

This is not a solved problem, but it is a solvable one, and the fact that it remains largely unsolved while ESG assets under management approach forty trillion dollars globally suggests that the people with the most to gain from better measurement have not yet decided that better measurement is worth paying for. The metric the investment industry is optimising for is assets under management, not actual impact per dollar deployed, and until that changes, the measurement framework will continue to serve the metric rather than the mission.

A Closing Note, A Confession, And An Offer Of Coffee

I want to be honest with you about something. I checked how many views my earlier posts was getting approximately three times while writing the section about how metrics corrupt outcomes. Goodhart’s Law ate the author in real time, and I did not even notice until the fourth reread. Please do not judge, views are technically a metric and I am apparently not immune.

The serious version of that confession is that none of us escape Goodhart’s Law by understanding it. We escape it, to the extent we escape it at all, by designing measurement systems with enough humility to know that any metric will eventually be gamed, building in review cycles that question the metric rather than only the performance against it, and occasionally being willing to sit with the discomfort of measuring something difficult rather than reaching for the nearest countable proxy.

If the cobra farms felt like a historical curiosity and the carbon credit market felt like a finance problem, consider what the same dynamic looks like when it is happening right now, at scale, inside organisations spending serious money on AI and measuring entirely the wrong things to justify it.

Enterprise AI spending hit an average of $85,521 per month per organisation in 2025, a 36% increase from the previous year, and yet only 51% of organisations can confidently say whether their AI investments are delivering any return at all. The other half, to borrow a phrase, are flying blind and burning millions while hoping something sticks. PwC’s 2026 CEO Survey found that 56% of CEOs reported neither increased revenue nor decreased costs from AI in the last twelve months, with only 12% reporting both, which is a striking result for a technology that has consumed more boardroom attention and capital expenditure than anything since the cloud migration wave. The metric being optimised in most of these organisations is adoption, measured by number of tools deployed, number of users onboarded, and volume of AI-generated output produced. MIT’s research found that more than half of generative AI budgets go toward sales and marketing tools because success there is visible and revenue attribution looks good in board presentations, while the highest ROI consistently comes from back-office automation that nobody wants to put on a slide. The metric, visibility of AI adoption, became the target, and the target drifted away from the goal, which was actual business value. S&P Global found that 42% of companies abandoned most of their AI projects in 2025, more than double the rate of the prior year, and that money, unlike a released tailless rat, does not come back. Goodhart would have recognised every single one of these budget post-mortems. He might even have found them funny, in the way that things are funny when they are also entirely predictable and entirely avoidable and yet somehow keep happening anyway.

The regenerative economy that the sustainability conversation is slowly building toward, the one that replaces extraction with restoration and measures wellbeing rather than just output, will not arrive by accident but only because enough people in enough rooms decide that the harder measurement question is worth the discomfort of asking it, and that the spreadsheet is a tool rather than an answer.

If you are working on any of this, in impact investing, in ESG governance, in philanthropic measurement, in public policy, or in a meeting room where someone has just proposed measuring something countable in place of something important, I would genuinely love to hear from you. I am always up for a coffee conversation about metrics, measurement, and the quietly consequential business of deciding what counts. Fair warning: I will probably arrive with opinions, ask inconvenient questions, and leave without a clean answer, because that is roughly what this topic deserves. But the coffee will be good and the conversation will be better.

Reach out. Let us count something that matters.

This post is not YAAP (Yet Another AI Post) nor AI authored. I love to research and write about topics like these, which stemmed from some event in my life. If you like what I wrote, and my style, follow for more random ramblings in the future.

This post is part of an ongoing series on sustainability, technology, and the systems we build to make sense of the world. This post appears both on Substack and Medium, so you can read it behind paywalls for free, should you be interested.