PTEN
Blog
B2B Marketing Measurement

Goodhart's Law in marketing: when a metric becomes a target, it stops measuring what matters

When a metric becomes a target, it stops measuring what matters. How Goodhart's Law distorts industrial B2B marketing and how to avoid the trap.

Marcelo Costa
Marcelo Costa 7 min read

Translated from the Portuguese original. Read in Portuguese

“This is starting to look like a pet shop.” According to Vitor Peçanha’s own account, co-founder of Rock Content, published in the Peça-chave newsletter in March 2023, the line came from his business partner, Diego, in a comment about the company’s Instagram. Peçanha opened the profile and confirmed it: photos of dogs walking around the office took up a good share of the posts. And he was the person in charge of marketing at one of the largest content marketing companies in the world.

The explanation he gives is anything but anecdotal. The social media team had an engagement target: likes, comments and shares. Every time the number was below expectations at the end of the month, there was a reliable way to close the gap. Post a puppy. Engagement went up, the target was hit, and none of it brought Rock Content any closer to a new customer or to a position as an authority in marketing. The metric worked perfectly. The business didn’t.

That is the cleanest picture I know of a principle every marketer should have pinned to the wall: Goodhart’s Law.

What Goodhart’s Law says

British economist Charles Goodhart formulated the idea in 1975, in the context of monetary policy. The version that became famous, later popularized by anthropologist Marilyn Strathern, is blunt: “When a measure becomes a target, it ceases to be a good measure.”

The logic goes like this. A metric is useful because it works as a proxy for something that truly matters but can’t be measured directly. Engagement is a proxy for genuine audience interest. Traffic is a proxy for qualified reach. Number of leads is a proxy for real demand. As long as nobody is deliberately trying to maximize the proxy, the correlation between it and the real outcome stays reasonable.

The moment the proxy becomes a target, with bonuses, pressure or reports tied to it, people start optimizing the number, not the phenomenon. And there is almost always a shorter path to moving the number than to moving the phenomenon. Posting a dog is faster than producing content that positions the company as a technical authority. The shortcut appears, the correlation breaks, and the metric keeps climbing while the outcome it was supposed to represent stalls or gets worse.

Why marketing is especially vulnerable

Marketing is a field where the outcome that matters, qualified demand and influenced revenue, tends to be slow, hard to attribute and outside the team’s direct control. Proxies, on the other hand, are fast, visible and easy to move. That asymmetry is the ideal ground for Goodhart’s Law to take hold.

Add to that the culture of monthly reporting. When a team has to present numbers every month and the real outcome only matures over quarters or longer, the pressure pushes toward whatever moves within the reporting cycle. Traffic, impressions, followers, raw lead volume. These are the so-called vanity metrics, and they are not useless by nature. The problem is never that the metric exists. The problem is when it becomes the goal.

It’s worth separating two things that are often confused. A misused vanity metric is one that becomes a target without being connected to the outcome. The same metric, read as a diagnostic signal within a context, can be perfectly legitimate. Engagement says something about how content resonates. Rock Content’s mistake wasn’t looking at engagement. It was turning engagement into an isolated target, without tying it to what the content needed to do for the business.

How this shows up in industrial B2B marketing

I work in industrial technology marketing, in an automation and instrumentation segment, and the same trap shows up here in different clothes. The logic applies to the industrial market in general, from mining and steel to industrial pharma, food and beverage, and pulp and paper, in any operation where the buying cycle is long and the decision goes through technical specification.

The most common example I see is organic traffic treated as a goal in itself. A team under pressure to “grow blog traffic” finds the predictable shortcut: producing generic, top-of-funnel content on high-volume terms that attract any visitor. Sessions go up, the chart looks great in the report, and the specifying engineer who decides on the purchase of a pressure transmitter for a superheated steam line never gets there, because that content doesn’t answer their question. Traffic became a target and stopped measuring qualified reach.

The same happens with raw lead volume. When the target is number of leads, with no qualification cut, the short path is to lower the barrier: a shallow piece of content behind a form, a lure that captures contacts from people who would never specify the product. The number goes into the report, sales gets a list that doesn’t convert, and the gap between the indicator and revenue grows with every month of “target achieved.”

This distortion is even more dangerous now because the point of contact has moved. As I argue in the article on GEO for industrial marketing, much of today’s technical decision-making starts with a query to a generative AI, months before any click on the website. A team chasing traffic as a target will keep optimizing for the sessions chart while the authority criterion migrates to a place that metric can’t even see.

How not to fall into the trap

No metric is immune. Goodhart’s Law isn’t solved by swapping a bad indicator for a better one, because any new indicator becomes a target as soon as it gets a goal attached. What reduces the risk is changing how you measure, not what you measure. Some practices I apply:

Tie each metric to the business outcome it represents, and review that link when the number goes up. If traffic grew, the next question isn’t “do we celebrate?”, it’s “did qualified pipeline grow with it?”. The instant the metric rises and the outcome doesn’t follow, it’s a sign that it has become a target and lost its validity.

Work with metrics in pairs, where one counterbalances the other. Lead volume alongside qualification rate. Traffic alongside session depth or conversion by source. A quantity indicator paired with a quality indicator makes the shortcut visible: you can inflate one of the two, but it’s much harder to inflate both at the same time without moving the real phenomenon.

Treat the number as the start of a diagnostic conversation, not as a verdict. The right question in the face of an engagement spike is what caused it, not whether the target was met. Was it content that brought the company closer to a buyer, or was it the equivalent of the puppy?

And, perhaps the most uncomfortable in practice, resist the temptation to promise stakeholders the easy number. Much of the pressure that turns proxies into targets comes from a monthly report that needs a line that always goes up. Arguing that the right indicator matures over quarters, not weeks, is part of the job of anyone managing marketing in a long-cycle market.

What stays

The pet shop story is good because it’s embarrassing in just the right measure. It wasn’t incompetence; it was a smart team doing exactly what the target asked for. That’s how Goodhart’s Law works: it doesn’t punish mistakes, it rewards obedience to the wrong indicator. The more efficient the team, the faster the proxy detaches from its purpose.

For those doing industrial marketing, where real results take time and proxies are tempting, the defense isn’t distrusting the numbers. It’s remembering that every indicator, once it becomes a target, starts drifting away from what it was supposed to represent. The job is to measure that drift before it grows too large to reverse.

Sources