Metrics 5 min read · · Last updated:
By Mark Ashworth · Founder, ChurnTools

Goodhart's Law in Retention: When the Metric Becomes the Target

When a measure becomes a target, it ceases to be a good measure. This is Goodhart's Law. It applies devastatingly to retention metrics. Once you incentivize NRR, teams optimize NRR at the expense of what NRR was supposed to indicate.

📊

Want a personalized score for your situation?

Take the free 60-second Churn Health Check

Score me →

When a measure becomes a target, it ceases to be a good measure.

This is Goodhart's Law. It applies devastatingly to retention metrics. Once you incentivize NRR, teams optimize NRR at the expense of what NRR was supposed to indicate.

How Goodhart's Law works

A metric starts as a proxy for something you care about. NRR is a proxy for "our existing customer base is growing revenue." NPS is a proxy for "customers feel good about us." Churn rate is a proxy for "customers keep paying."

Then you tie compensation, career advancement, or organizational status to the metric. Teams now have incentive to move the number.

Some of that movement comes from genuinely improving the underlying thing. Some comes from optimizing the measurement itself. Over time, the ratio shifts toward measurement optimization because it is cheaper and faster than real improvement.

Eventually the metric moves without the underlying reality changing. It has ceased to measure what it was supposed to.

Common cases in SaaS retention

1. NRR gaming

NRR targets are set. CS teams focus intensely on expanding the top 5% of accounts. Massive expansion in a few accounts pushes NRR above 110%. The base of smaller accounts, meanwhile, is churning at 8% quietly.

Reported NRR: 118%. Actual customer base health: declining. Two years later, the top-5% accounts consolidate their vendors and the growth story disappears.

2. NPS survey timing

NPS is a KPI. Someone realizes that surveys sent immediately after a positive product interaction get higher scores than surveys sent at random times.

Suddenly NPS is only sent after successful actions. Score improves from 32 to 51 over 6 months. Actual customer sentiment has not changed. The survey has just been re-timed.

3. Churn rate reclassification

Churn rate is being watched by the board. A cancellation this month would push the number above target. Someone convinces the customer to "pause" instead. The cancellation does not count as churn this quarter. It counts next quarter. Or it gets forgotten if the customer never comes back.

Reported churn: 3%. Actual customer loss: 4.5%. The definition of "churn" quietly expanded to exclude paused accounts.

4. Health score inflation

The CS team is measured on the percentage of accounts in "green" health status. Someone tweaks the health score rules to be more generous. More accounts turn green. CS performance appears to improve.

Reported healthy accounts: 78%. Actual account health: same as before. Just measured differently.

5. "Active user" definition expansion

Monthly active users is a growth metric. Someone expands the "active" definition from "logged in this month" to "logged in in the last 60 days." The number goes up.

Reported MAU: 45,000. Actual usage: unchanged. The window just doubled.

Signs your retention metrics are being gamed

1. Numbers are suspiciously clean

Real retention numbers wobble. If your NRR sits at exactly 108% for three quarters in a row, something is being managed. Real numbers oscillate; targeted numbers stabilize suspiciously.

2. Improvements coincide with metric definitions changing

A metric definition changes and the number improves in the same quarter. This is usually gaming, not improvement.

3. Uncorrelated metrics start correlating

NRR is improving. But gross retention is not. Product usage is not. NPS is not. Support tickets are not. If NRR is going up in isolation, it is probably being optimized at the expense of things it should correlate with.

4. Team behavior focuses on measurement, not customers

You start hearing: "How do we count this?" "Where does this show up in the metric?" "What if we redefined X?" These are the sounds of Goodhart in action. Teams are optimizing the measurement, not the outcome.

How to defend against Goodhart

Measure multiple correlated metrics

If NRR is a KPI, also track gross retention, product engagement, NPS, and support ticket volume. When they diverge, something is off. Multiple metrics that all should move together are harder to game than one.

Audit for gaming patterns

Regularly review whether metric definitions have shifted. Look for suspicious cleanness. Ask "what would the metric look like under the old definition?" and compare.

Rotate the KPI focus

If NRR is the KPI for two years and every team is optimizing it, switch the KPI to gross retention for a year. This forces broader retention work and prevents extreme optimization of any single metric.

Include qualitative context in reporting

Numbers alone hide gaming. Numbers with context (which accounts drove the change, what changed operationally, what customer feedback correlates) surface gaming.

Ensure metric owners do not control definitions

The person measured on NRR should not be the person who defines what counts as expansion revenue. That is a governance separation that most SaaS orgs skip.

The uncomfortable implication

Every metric in your retention dashboard is potentially subject to Goodhart. The better it looks, the more you should ask what is being gamed.

This is not paranoia. It is realism about how organizations respond to incentives. The teams with the healthiest retention metrics are not always the ones with the healthiest retention. Sometimes they are just the ones optimizing measurement most aggressively.

Real retention improvement is slower, messier, and harder to point at than metric gaming. That is why gaming is so common. And that is why healthy skepticism about your own numbers is a retention superpower.

Related concepts

To score whether your retention setup measures what you think it measures, take the 60-second Churn Health Check.

Free interactive tool

Score your retention setup in 60 seconds

8 questions. Get your tier (Critical to Best-in-Class), your weakest spots, and 3 specific things to fix next.

Take the Health Check

Frequently asked questions

Answers to the questions I get most often about this topic.

What is Goodhart's Law?

Goodhart's Law states that when a measure becomes a target, it ceases to be a good measure. Once you tie compensation, career advancement, or organizational status to a specific metric, people start optimizing the metric itself rather than the underlying outcome it was supposed to indicate. This distorts measurement over time.

How does Goodhart's Law apply to retention?

Common cases in SaaS: (1) NRR targets lead to concentrating expansion in a few large accounts while ignoring the base, (2) NPS targets lead to survey timing tricks and sample cherry-picking, (3) churn rate targets lead to reclassifying voluntary cancellations as "paused" or "restructured", (4) health score targets lead to score inflation through generous rules. All distort the underlying truth.

How do you avoid Goodhart's Law in retention?

Three defenses: (1) measure multiple correlated metrics so gaming one distorts the others visibly, (2) audit for gaming patterns regularly (unusually clean numbers, sudden improvements, timing coincidences), and (3) rotate which metric is "the target" so no single metric gets optimized to the exclusion of others. The core principle: never make one metric the sole KPI.

What are examples of retention metric gaming?

Real examples: reclassifying customer cancellations as "temporary pauses" to preserve retention numbers. Sending NPS surveys immediately after positive interactions to inflate scores. Redefining "active user" to include anyone who logged in during a longer window. Delaying churn recognition by shifting billing dates. Each of these is technically legal and destroys the metric's meaning.
MA

Written by Mark Ashworth

Founder of ChurnTools. I spend my time studying how SaaS companies lose customers and building tools to help them stop. Previously worked in SaaS growth and retention across multiple B2B products. I also write about growth and answer-engine optimization (AEO) at growthpigeon.com.

Ready to run your first retention experiment?

Browse 30+ proven playbooks for reducing churn across every stage of the customer lifecycle.

Browse Experiments →