What gets measured is not the same as what matters, and rewarding the measure corrupts it. Not everything that counts can be counted. Extrinsic reward for measured performance crowds out the intrinsic motivation that made people good at the work in the first place

How metrics trigger gaming

We all remember how Wells Fargo set aggressive cross-selling quotas for its bankers. Facing unpaid overtime and termination for missing targets, roughly 5,300 employees resorted to fraud, opening credit cards and online accounts for customers who never authorized them.

A Baltimore narcotics detective, Ed Burns spent months building a case against a major drug lord. His superiors preferred he arrest low-level street dealers instead, since five arrests a day produced better arrest statistics than one slow, high-value investigation.

Under Robert McNamara, the Pentagon used enemy body counts to measure progress in Vietnam. Field commanders routinely exaggerated the numbers to improve their chances of promotion, even though many considered the metric close to meaningless.

New York State published surgeon-level mortality data for coronary bypass surgery. Cardiac surgeons became less willing to operate on severely ill patients, and some referred their riskiest cases to the Cleveland Clinic, keeping those failures out of the state’s numbers entirely.

College graduation rates factor into rankings and funding, so institutions faced pressure to pass more students. Courses became easier, standards loosened, and adjunct instructors who failed too many students risked their contracts not being renewed.Airlines reported strong “on-time performance,” but part of the improvement came from padding the scheduled flight time itself, so planes arrived within a longer, easier window without actually flying any faster.

The moment measurement gets tied to reward and punishment for people, gaming, cheating, and goal-distortion become rational responses to a badly built system.

1. Rewarding the measure corrupts it

This is Campbell’s Law and Goodhart’s Law in plain English: any number that becomes a target stops being a good measure, because people change their behavior to hit the number rather than to achieve the thing the number was supposed to represent.  If beating the target gets rewarded, then people set the target low enough to beat it comfortably. Right?

Build multi-signal evaluation before you build the dashboard

2. Metrics don’t measure tacit skill 

Muller draws on the difference between “technical knowledge” like recipes, formulas, and things you can write down & “practical knowledge” like the tacit skill of a surgeon’s hands, a teacher’s feel for a classroom, a general’s sense of a battlefield. Metric fixation assumes only the first kind is real knowledge. AI tools are extremely good at generating countable outputs (lines of code, drafts, tickets closed).

Careers that survive and thrive in the AI age will be the ones built on judgment, taste, trust, the ability to know which problem is worth solving. This is exactly what metrics don’t measure.

A counterinsurgency advisor in Afghanistan, discovered that the local price of exotic vegetables ie vegetables grown outside the district and hauled in at risk, was a more honest signal of a district’s actual peace and security than official violence statistics. After all, traders only transport risky goods when they trust the roads. Such a metric only surfaced because someone with deep local judgment was paying attention to.

Complex work is driven by mastery, mission, and professional pride – not money

Every conversation about “incentive alignment,” badges, leaderboards, and bonus-linked KPIs is a bet that extrinsic reward beats intrinsic drive. Pay a child for every book they read, and you’ll often get a kid who reads short books fast to earn money and reads fewer books once you stop paying. 

Healthcare work is full of naturally satisfying moments like relieving pain, solving a diagnostic puzzle, being trusted by a patient. Turning all of it into a financial transaction can quietly poison those satisfactions.

“Effective governance demands preserving this “strategic memory,” anchoring KPIs in a narrative clarifying intended outcomes and beliefs. Linking metrics to a shared strategic picture enables businesses to validate hypotheses, preventing optimization of numbers over true outcomes, and ensuring data informs strategy, not substituting for it, especially amidst abundant AI-driven analysis.”

Read this <What sits upstream of the KPI>

Talent Management metrics need to change

Recruiting and hiring. Talent acquisition teams are rewarded for time to fill and cost per hire.. This will lower the bar on harder searches. Hiring managers skip the deeper conversations that actually reveal fit or long-term potential.

Measure quality of hire at the twelve-month mark instead, using manager ratings, retention, and ramp-to-productivity time as the real signal.

Onboarding. New hire experience gets reduced to training completion rates and orientation attendance, since those are easy to track in an HRIS. An employee can finish every module and still feel unsupported, disconnected, and ready to leave within six months, because completion says nothing about belonging or clarity.

Track whether a new hire’s manager held unprompted check-ins and whether the employee reports feeling set up to succeed at 30, 60, and 90 days.

Performance reviews. Standardized ratings reward employees who hit the same numeric targets regardless of the difficulty of their assignment, client roster, or team conditions. A manager handling a struggling team or an unfair client mix looks worse than one handling an easy account, even when the harder manager is doing the better work.

Measure performance relative to starting conditions and degree of difficulty, not just the raw outcome number.

Promotion and forced ranking. Stack ranking rewards individuals for outscoring their peers, which feels rigorous and defensible to leadership. It quietly punishes cooperation, since helping a colleague can lower your own relative standing, and it damages morale even in companies that use it.

Measure documented instances of mentoring, knowledge sharing, and cross-team support as part of promotion criteria, not just individual output.

Employee engagement surveys. Organizations reward high participation rates and rising engagement scores as proof the culture is healthy. Employees who suspect their honest answers will be traced back to them give safe, compliant answers instead of real ones, so the score improves while trust quietly erodes.

Protect genuine anonymity in a subset of channels and track whether people raise hard issues through informal routes, which is a better signal of real trust than the survey score itself.

Learning and development. L&D teams are often measured on courses completed and training hours logged, since these numbers are simple to report upward. Employees complete the modules to check the box, without necessarily building the skill the training was meant to develop.

Measure applied behavior change on the job, through manager observation or performance shifts in the weeks after training, rather than completion alone.

Retention and turnover. Leadership tracks overall turnover rate as the headline health metric for the organization. This number treats the departure of a weak performer and the departure of a top performer as identical, so a company can lose its best people while the turnover rate still looks acceptable.

Segment exit data by performance tier and track regretted attrition specifically, since that is the number that actually predicts damage to the business.

Diversity hiring. Many organizations reward recruiters and hiring managers for meeting representation percentages at the point of hire. This can produce creaming at the entry level while doing nothing to fix promotion rates, pay equity, or retention further up the pipeline, so the headline number improves while the underlying problem persists.

Measure representation and outcomes at every stage of the pipeline, including promotion velocity and voluntary attrition by group, not just the hiring number.

The measure of a measure

More measurement usually means more administrative cost, more time spent reporting instead of doing, and more subtle gaming that management never sees. Judgment was never meant to be replaced by measurement. It was meant to be informed by it.