Campbell's Law: How HR Metrics Corrupt the People Processes They Try to Measure

60-Second Summary

Campbell's Law (1976): high-stakes metrics not only stop measuring — they distort the underlying process.
Original context: education. Standardized testing distorts curriculum (Atlanta cheating scandal, 2009).
HR analogs: stack-ranking distorts hiring; engagement targets distort honest feedback; DEI dashboards distort how identity is recorded.
Difference from Goodhart: Goodhart says the measure rots; Campbell says the people and processes rot too.
The fix is structural: low-stakes measurement, broad outcome metrics, and human judgment in the loop.

Campbell's Law is the meaner version of Goodhart. Goodhart says the number lies. Campbell says the process underneath the number warps to fit. If engagement is tied to bonuses, your managers don't just coach the score — over time, they avoid hard conversations, hide attrition risk, and reshape their leadership behavior around the survey calendar.

What Campbell actually wrote

“The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor.”
— Donald T. Campbell, 'Assessing the Impact of Planned Social Change' (1976)

The education-policy origin

Campbell's evidence base was US education policy. He noted that once standardized test scores were used to evaluate teachers and schools, three things happened: curriculum narrowed to tested subjects, teaching became test-prep, and outright cheating emerged. The 2009 Atlanta Public Schools scandal — where 178 teachers and administrators systematically erased and corrected student answers — was Campbell's Law in textbook form.

Where it shows up in HR

HR system	Original intent	Campbell distortion
Stack ranking (Welch-era GE)	Surface true top and bottom performers	Managers hire weak players to protect their team's distribution
DEI representation dashboards	Track inclusion progress	Quiet reclassification + retention drops as people are added to hit numbers
Engagement bonus targets	Improve workplace climate	Suppressed honest dissent, coached survey responses
Time-to-hire dashboards	Reduce friction in recruiting	Lowered bar, skipped reference checks
NPS/CSAT for support agents	Improve customer experience	Begging customers for 10s, deflecting hard cases
Promotion velocity targets	Career growth visibility	Inflated titles, role-creep without scope

How it differs from Goodhart

Goodhart vs. Campbell

Goodhart's Law

About the measure
Says the number loses meaning
Fix: rotate or decouple metric

Campbell's Law

About the people and processes
Says the underlying behavior distorts
Fix: lower stakes, broaden judgment, allow human override

What to do instead

Use metrics as one input to human judgment, never as the decision itself.
Keep performance metrics low-stakes for individual managers — never one metric tied to one bonus.
Triangulate: any decision that uses a metric should also use peer signal and qualitative evidence.
Explicitly name the Campbell risk in dashboard design — 'this metric will be gamed if we tie it to comp' belongs in the dashboard documentation.
Audit for second-order effects: did stack-ranking actually surface top performers, or did your best people quit while your weakest survived by political alignment?

DEI dashboards

DEI metrics are uniquely vulnerable to Campbell distortion because the underlying construct (belonging) is harder to measure than the proxy (representation). When the proxy becomes the target, hiring distorts but retention numbers reveal the gap — usually 18 months later.

FAQ

Frequently asked questions

Should we abandon all HR dashboards?

No — measure widely, report internally, but keep stakes low and human-in-the-loop. Dashboards as diagnosis, not as verdict.

Can we incentivize on outcomes instead of metrics?

Yes, but Goodhart-Campbell still apply if the outcome is narrowly defined. Broad, hard-to-game outcomes (e.g., revenue per FTE over a 2-year horizon) survive longer.

How do I know my metric has been corrupted?

Three signs: sudden non-linear improvement after the metric was tied to stakes; underlying outcomes (attrition, customer outcomes) diverging from the metric; or front-line whistleblower stories about 'how we hit the number'.

Takeaways

Tying a measure to stakes warps not just the measure but the underlying social process.
DEI, engagement, and hiring dashboards are uniquely vulnerable.
Low-stakes diagnosis + human judgment is the only system that survives long-term.