What stood out to me about these metrics is that each captures something valuable, but each could also improve while the product gets worse.
Airbnb: nights booked:
Nights booked captures whether guests and hosts actually have transactions. It also accounts for stay length, so that a long booking counts more than a one night stay. That makes it a useful proxy for Airbnb activity and value delivered to both sides. But a team could inflate it with unsustainable discounts or by pushing longer stays regardless of what guests need. This would result in reduced trust over time for consumers.
Facebook: 7 friends in ten days:
The early growth team found that making friends correlated with retention. The number represented finding enough connections to experience Facebook’s value before leaving. A team could game this by pressuring users to add strangers. They’d hit 7 faster, but end up with a less relevant feed and fewer reasons to return.
Slack: 2,000 messages:
The 2000 messages signals that a team had really tried communicating through the product. A team could manufacture messages through bots or unnecessary prompts to game this. That would increase volume while making actual work less efficient.
This is the problem Goodhart’s Law describes. Optimizing a metric may weaken its connection to the outcome we care about. I’d pair each metric with checks on satisfaction and retention.
