Back to notes

August 12, 2026 · 5 min read

How a North Star Metric Got Gamed Within One Quarter

We picked a clean, simple North Star metric. The team optimized it exactly as designed, and the business got measurably worse.

PostmortemGrowth

The North Star metric was weekly active users. Clean, easy to explain, tied conceptually to the product's value. Within one quarter, weekly actives were up 18%, and the business was in worse shape than before — because nobody had set a guardrail against the specific way teams found to move that number.

What 'optimizing' actually looked like

The fastest way to move weekly active users, it turned out, was a notification strategy that pulled users back into the app for shallow, low-value sessions — a badge count, a reminder, a nudge with no real new information. It worked exactly as measured: weekly actives went up. Retention cohort quality, time-to-value for new users, and NPS all declined in the same quarter, because the team was optimizing the metric, not the thing the metric was supposed to represent.

Every metric, given enough optimization pressure, will eventually be gamed in exactly the way it's easiest to game. A North Star without a guardrail metric is an invitation, not a safeguard.

What we added after catching it

  • Paired the North Star with explicit guardrail metrics (notification unsubscribe rate, session depth, cohort retention) that would flag exactly this kind of gaming.
  • Required any initiative claiming North Star impact to show it didn't move the guardrails in the wrong direction before being counted as a win.
  • Redefined the North Star itself from raw weekly actives to a qualified version — sessions with a minimum depth threshold — that was much harder to move with low-value engagement.
  • Reviewed the metric definition quarterly specifically asking 'what's the cheapest way to move this number that we wouldn't want' and treated a good answer as a design flaw to fix.

The team that built the notification strategy wasn't acting in bad faith — they were given one number to move and moved it efficiently. That's the actual lesson: a single metric, no matter how well chosen, needs guardrails from day one, because the easiest way to move most metrics and the right way to grow the business are rarely the same path.