Goodhart's Law

Goodhart's Law is a single, brutal observation: when a measure becomes a target, it ceases to be a good measure.

This isn't about dishonesty. It's about rational behavior. You learned in incentives that the incentive structure determines what people actually do. Goodhart's Law is the specific failure mode that emerges when you attach incentives to a measurement: people find the cheapest path to move the number, whether or not that path delivers the underlying Value Creation you wanted.

The measure was originally useful because nobody was trying to manipulate it. It passively reflected something real. The moment you make it a target - tie Commissions to it, set Hiring Targets around it, gate promotions on it - you've changed the game. The measure now reflects effort spent optimizing the measure, not the thing it used to track.

Every P&L is managed through numbers. You can't run Operations at scale by inspecting every transaction yourself - you pick measurements, set targets, and allocate resources based on what the numbers tell you. Goodhart's Law says this entire management layer is vulnerable to silent corruption.

The P&L impact shows up in three ways:

1. 1)Misallocated spend. If your measurements are distorted, your resource allocation decisions are built on lies. You'll pour Marketing Spend into a channel that looks productive but isn't. You'll staff up a team whose Throughput numbers are inflated.

2. 2)Hidden Churn. A metric looks healthy while the underlying Value Creation degrades. Customers leave, but your dashboard doesn't register the damage until it compounds into a Revenue miss.

3. 3)Broken Feedback Loops. Good Operators use Feedback Loops to learn what's working. Goodhart-corrupted metrics poison the loop - you get false positive signals, double down on the wrong things, and the real problems stay invisible.

The mechanics follow a predictable sequence:

Step 1: You pick a measure that correlates with the outcome you want. Close Rate correlates with sales effectiveness. defect rate correlates with engineering quality. CSAT correlates with customer happiness.

Step 2: You attach incentives to the measure. Bonuses, quotas, Hiring Targets, public dashboards, performance reviews - anything that makes hitting the number more attractive than missing it.

Step 3: People find ways to optimize the measure that don't require optimizing the outcome. This is the key insight. There are almost always cheaper ways to move a number than doing the hard underlying work:

• •Sales reps targeted on Close Rate stop pursuing large, uncertain deals. The rate improves but Pipeline Volume and Revenue shrink.
• •Engineers targeted on defect rate reclassify bugs as "feature requests." The defect count drops but product quality doesn't change.
• •Support reps targeted on CSAT cherry-pick easy tickets and dodge complex ones. Scores rise but hard problems go unresolved, driving Churn.
• •Recruiters targeted on Time-to-Fill push mediocre candidates through faster rather than finding better ones.

Step 4: The measure diverges from the outcome. You now have a number that looks great and a reality that's getting worse. And because you're managing by the number, you don't see the divergence until something breaks visibly - a Churn spike, a Revenue miss, a quality collapse.

The underlying math is straightforward. Every measure captures some dimensions of the outcome and misses others. When you incentivize the measure, rational actors exploit the gap between what the measure captures and what you actually want. The wider that gap, the worse the distortion.

Apply Goodhart's Law as a diagnostic lens in these situations:

When designing any incentive structure. Before you tie Commissions, bonuses, or targets to a metric, ask: How would a smart, rational person hit this number without delivering the outcome I actually want? If you can think of a way, so will they.

When a metric improves faster than expected. Sudden jumps in CSAT, Close Rate, Throughput, or defect rate should trigger suspicion, not celebration. Real operational improvement is usually gradual. Sharp moves often mean the measurement is being gamed, not that the underlying process improved.

When building a Scoring Model or Quality Gates. Any time you reduce a complex outcome to a number and gate decisions on that number, Goodhart applies. The more consequential the gate, the stronger the incentive to game it.

When two metrics diverge. If CSAT is up but Churn Rate is also up, one of those signals is lying. Goodhart's Law tells you which one to suspect: whichever one is being targeted.

Countermeasures:

• •Measure multiple dimensions of the same outcome (harder to game all of them at once)
• •Rotate which metrics are targeted so optimization pressure doesn't concentrate
• •Pair every efficiency metric with a quality metric (Close Rate paired with Expansion Revenue, for example)
• •Use auditing and Spot-Check processes on the measurement itself, not just the process being measured

Direct consequence of incentives - the specific failure mode when an incentive structure rewards a measurement instead of the outcome. Connects downstream to Quality Gates (gamed when the gate score carries consequences), Scoring Models (optimized for the score rather than the underlying quality), and Feedback Loops (corrupted metrics poison the signal). Apply when designing Commissions structures, Hiring Targets, or any resource allocation tied to a targeted metric.