Skip to content
All notes

Why metrics lie without context

·3 min read

Early in my PM career, I walked into a quarterly review with a slide that said "DAU up 23%." The room was happy. The VP nodded. We moved on.

Key Takeaway

Three weeks later, we discovered the increase came from a bug in the session tracking that was double-counting users who switched between tabs. Real DAU was flat. We'd spent the quarter optimizing for a metric that was lying to us.

The NumberWhat happened"DAU up 23%"The MechanismWhy it happenedCounter-MetricWhat you're payingDecisionAll three requiredpush notifications3x opt-out ratecaused by → traded off by

Every metric needs three things

The number is what happened. DAU is 23% up. Revenue grew 15%. Support tickets dropped 40%.

The mechanism is why it happened. DAU grew because we launched push notifications. Revenue grew because we raised prices. Support tickets dropped because we removed a feature.

The counter-metric is what you're paying for it. Push notifications increased DAU but also tripled opt-out rates. Price increase grew revenue but churn went from 3% to 7%. Removing the feature cut tickets but NPS dropped 12 points.

The dashboard trap

I've worked with teams that had 30+ metrics on their main dashboard. They checked it every morning. They could tell you the exact number for any KPI at any time.

But when I asked "which of these metrics would change your next decision?", the answer was usually two or three. The rest were vanity metrics — numbers that went up, which made people feel good, but didn't inform any actual choice.

The best dashboard I ever built had four numbers on it. One for acquisition, one for activation, one for retention, one for revenue. Each one had a target, a trend line, and a one-sentence explanation of why it was moving. That's it.

Three questions I ask every time

When someone presents a metric, I ask three questions:

1

"What's the mechanism?"

If they can't explain why the number moved, the number is noise.

2

"What's the counter-metric?"

If there's no trade-off, either we got lucky or we're not looking hard enough.

3

"What would you do differently if this number were 10x worse?"

If the answer is "nothing," then we're tracking the wrong thing.

These questions aren't popular in meetings. Nobody wants to hear "your 23% growth might be a bug." But the alternative is building strategy on sand.