The dilemma
Illustrative scenario
Priya is editor-in-chief of a recipe and home-cooking site that earns money from display ads and a paid membership. Last quarter, leadership tied the editorial team's bonus to monthly page views. The team delivered: headlines turned into teasers ("The one mistake ruining your pasta"), and long recipes were split into ten-page slideshows. Page views climbed every week.
Priya's other reports tell a different story. Fewer visitors now scroll far enough to reach the actual recipe. Membership sign-ups have not moved. The ad sales lead says advertisers are paying less for impressions on the slideshow pages, and reader comments complain about being made to click through.
The bonus review is on Friday. The team hit its target fairly, by the rules it was given. Does Priya renew the page-view bonus for next quarter, or change the target and risk the team feeling the goalposts moved?
The short answer
Goodhart's law is usually quoted as "When a measure becomes a target, it ceases to be a good measure." That compact wording comes from the anthropologist Marilyn Strathern (1997), who generalized an argument the economist Charles Goodhart made in 1975 about monetary policy: a statistical relationship that looked stable tended to break down once the authorities leaned on it for control purposes.
How strong is the evidence?
Adage or reasoning rule
A memorable saying or reasoning rule, not an experimental finding. Treat it as a prompt to check, not as evidence.
This is an adage: a well-argued observation from economics and organizational life, not a measured effect with a known size or a replicated experiment.
The mechanism is easy to state. A metric is chosen because it moves with something you care about. Once pay, rankings or approval depend on it, people look for the cheapest way to move the metric itself: selecting easy cases, redefining what counts, or cutting quality that nobody measures. The metric keeps rising while its link to the goal weakens.
The distortion tends to be worst when the reward is large and tied to a single number, when the number can be moved without moving the outcome, when the people who move it also report it, and when the real outcome shows up much later. It is not inevitable: metrics that sit close to the outcome and are hard to fake (audited revenue, verified deliveries) hold up better, and some targets do focus effort usefully. Goodhart's original point was about monetary statistics; applying it to marketing KPIs is a generalization, which is how most writers now use it.
Investigating an example
WeCaseStudy analysis
Sales goals and the Wells Fargo unauthorized accounts
In 2016 the US Consumer Financial Protection Bureau fined Wells Fargo $100 million. According to the regulator, bank employees, "spurred by sales targets and compensation incentives," covertly opened deposit and credit card accounts without customers' knowledge or consent.
WeCaseStudy analysis: we read this as a Goodhart dynamic. A count of accounts opened is a sensible proxy for deeper customer relationships until people are pressured and paid to raise the count; at that point the number can keep rising while the relationships it was meant to signal are being damaged. This interpretation is ours. The regulator's findings concern illegal sales practices, not measurement theory.
Limits: this is an enforcement case, not a controlled study. Culture, management pressure and weak oversight were part of the story, so it cannot tell you how often targets backfire or by how much.
Our inference about a real, sourced event. The source does not itself name this law.
Choose your move
You are in the dilemma above. Which move would you make? Open each one to see its trade-offs.
A Renew the page-view bonus unchanged: the team met the target fairly.
Morale holds and traffic keeps climbing, but you keep paying for slideshow splits and teaser headlines. If ad rates and reader trust erode, the damage shows up quarters later, after more bonuses have been paid on the number.
B Replace page views with membership sign-ups as the single new target.
Closer to revenue, but it is a new single target with its own cheap routes: aggressive paywall prompts and discount-led sign-ups that cancel quickly. Editors also control only part of the sign-up journey, so the reward becomes noisy.
C Keep page views as one input, add guardrails such as recipe-reach rate and revenue per thousand visits, and announce the change a quarter ahead.
Harder to explain and calculate, and someone has to judge trade-offs. It makes gaming visible and gives the team notice, but too many metrics can blur what actually matters.
No move wins in every business. The law tells you what to check, not what to choose.
How to apply it
- Name the goal behind the number. Write one sentence on what the KPI is a proxy for (for example, "page views stand in for engaged readers who see ads and may subscribe").
- List the cheap routes. Run a short pre-mortem: how could someone hit this number next quarter without improving the goal? Each route you find needs a check.
- Add one or two guardrails per target. Choose counter-metrics that would worsen if a cheap route were taken: clicks paired with purchases per visitor, leads paired with sales acceptance, handle time paired with repeat contacts.
- Separate moving from auditing. Have someone outside the rewarded team sample records, headlines or leads each month and confirm definitions have not drifted.
- Recheck the link every quarter. Compare the KPI with the business outcome it was chosen to predict. If the relationship weakens, revise the target before the next reward cycle.
What to measure
Track each target next to its guardrails, the ratio between them (sign-ups per thousand page views, revenue per lead) and the quarter-by-quarter relationship between the target and the outcome it stands for. Guardrail: if the target rises while outcome per unit falls for two periods in a row, pause rewards tied to it and investigate.
Worked example
| Metric | Before bonus | After bonus | Change |
|---|---|---|---|
| Monthly page views | 4.0 million | 5.6 million | +40% |
| Visitors who reach the recipe | 62% | 41% | −21 points |
| Ad revenue per 1,000 page views | $9.00 | $7.20 | −20% |
| Monthly ad revenue | $36,000 | $40,320 | +12% |
| Membership sign-ups per month | 2,100 | 2,050 | −2% |
Illustrative scenario, not data from a real publisher. Ad revenue = page views × revenue per 1,000 ÷ 1,000. The headline metric rose 40% while revenue rose 12% and reader-quality signals fell; whether that trade is worth it depends on longer-term effects a single quarter cannot show.
Where it breaks
- Not every target collapses. Metrics close to the outcome and hard to fake hold up better than easy proxies such as clicks or form fills.
- No targets is not the cure. Without numbers, effort drifts and budgets follow the most persuasive story.
- Guardrails can be gamed too, and a dashboard with twenty "key" metrics leaves nobody sure what matters.
- It is a generalization. Goodhart wrote about monetary statistics; the marketing version is an analogy, useful but not a tested law.
- Ethics: when a target pushes staff to act against customers, the fix is incentive design and oversight. Blaming individuals for a pressure the system created repeats the problem.
What if you apply it incorrectly?
A team learns Goodhart's law as "metrics are bad" and drops targets altogether. Quarterly reviews turn into storytelling contests, and the most confident narrative wins budget. A year later nobody can say which campaigns worked, so leadership reinstates one blunt KPI with no guardrails, which is even easier to game than the original.
Myth versus evidence
The social-video version
Any metric you track becomes useless the moment you track it.
What the evidence supports
A metric tends to lose its signal when strong rewards attach to it and it is cheap to move; guardrails, audits and outcome-close metrics reduce the distortion.
Five-minute field assignment
Pick the KPI your team is rewarded on. Write down three ways someone could raise it without improving the business. For each, find one guardrail in last quarter's data (conversion after the click, refund rate, repeat visits) and check: did the KPI and the guardrail move in the same direction?
Checkpoint
Two questions: one about the idea, one about a decision. Answer both correctly to master this law. Questions are drawn from a larger bank and graded on our server; a retry brings different questions.
Evidence ledger
Every factual claim in this lesson, where it comes from, and how far it has been checked.
-
Goodhart's 1975 argument concerned monetary statistical relationships breaking down once used for control purposes.
Limit: Original wording and publication details to be checked against the paper.
-
The phrasing "When a measure becomes a target, it ceases to be a good measure" is Strathern's 1997 generalization.
Limit: Confirm exact wording and page in the article.
-
On 8 September 2016 the CFPB fined Wells Fargo $100 million; it said employees, "spurred by sales targets and compensation incentives," covertly opened unauthorized accounts.
Limit: Regulator findings; the link to Goodhart's law is WeCaseStudy inference.
-
The recipe publisher and the page-view table.
Limit: Teaching scenario; no real company.
Further reading
- Goodhart, C. A. E. (1975). Problems of monetary management: The U.K. experience.
- Strathern, M. (1997). 'Improving ratings': Audit in the British university system. European Review, 5(3).
- Consumer Financial Protection Bureau. (2016, September 8). Consumer Financial Protection Bureau fines Wells Fargo $100 million for widespread illegal practice of secretly opening unauthorized accounts [Press release].
Which other law could explain this decision?
Editorial review: Ali Dezhami, November 8, 2026. Last substantive update: November 8, 2026. Our methodology