Goodhart's Law and "Lowest Common Consensus"

Consensus cultures, paralyzed by the complexity of agreeing on nuanced trade-offs, often default to single, easily-gamed metrics, making them uniquely vulnerable to Goodhart's Law.

Goodhart’s Law in Brief

Goodhart’s Law is best summed up by: “When a measure becomes a target, it ceases to be a good measure or good target.”

Any single metric you use to track progress is useful only as long as it remains an observation. The moment you incentivize people to improve that metric, they will find the shortest, easiest path to increase the number, often by destroying the underlying value the metric was supposed to represent.

So are metrics bad?

No. The standard solution to Goodhart’s Law is counterbalancing metrics. You never set a single target; you set multiple targets that trade off between them. You force the system to maintain equilibrium rather than maximizing a single variable to infinity.

• The Example: A classic case is a call center.

  • Single Metric (Vulnerable): If you only measure Average Handle Time (AHT) to increase efficiency, agents will just hang up on difficult customers to keep their averages down.

  • Counterbalance (Resilient): You pair AHT with Customer Satisfaction (CSAT). Now, an agent cannot just hang up (hurting CSAT) nor can they spend an hour on a simple issue (hurting AHT). They must balance speed with quality.

The Consensus Complication

Counterbalancing works well when a decisive org can pick this healthy tension. But what happens in a consensus-based culture, where every decision requires broad stakeholder buy-in? Aside: Note that more-consensus or more-decisive is not better or worse, just better or worse fit for different kinds of problems. Now back to the consensus culture trying to pick a mix of metrics to counterbalance:

It breaks down due to the sheer “informational load” required to get everyone to agree.

In a consensus culture, getting approval for one metric is already a grueling process of meetings, socialization, and sign-offs. You have to convince every stakeholder that Metric A is the right thing to focus on.

If you try to introduce a counterbalancing Metric B, you don’t just double the difficulty—you exponentially increase it. You aren’t just asking for agreement on A and agreement on B; you need consensus on the interaction between them. And the debate between all the possible pairs of A and B.

• “How much efficiency are we willing to sacrifice for satisfaction?”

• “What is the acceptable exchange rate between AHT and CSAT?”

• “Who decides when the tension is out of balance?”

This combinatorics problem is paralyzing. The complexity of agreeing on the trade-offs is too high for a committee to process efficiently.

The Lowest Common Denominator

Faced with this paralyzing complexity, consensus cultures inevitably retreat to simplicity. They need something everyone can say “yes” to without weeks of debate.

They default to a single, easily understood metric—the lowest common denominator. It’s the only thing that can survive the friction of the consensus process. People get tired of the debate, and you get a metric version of “nobody is fired for hiring IBM”.

Because they are structurally incapable of efficiently agreeing on the complex counterbalancing measures required to keep a system honest, consensus cultures almost always end up with single, unprotected targets. And as Goodhart warned us, those are the targets that get gamed the fastest.