North Star and Input Metrics
Use a shared measure without allowing it to flatten the product's value. · Data & Experimentation · Lesson 32 · 3 min
North Star and Input Metrics · 3 min
Situation
One number seems like a solution.
A team proposes total time in the app as its North Star. That may fit some entertainment experiences, but a tool that saves people time could get worse while the number rises.
A North Star is a chosen expression of delivered value. Its usefulness depends on the product's purpose and the behaviors behind it.
Mental model
Outcome above, controllable drivers below.
A useful North Star reflects a recurring valuable outcome, is measurable, and has a plausible relationship to sustainable business results.
Input metrics describe drivers the team can influence. For a collaborative planning product, completed weekly plans may depend on teams starting a plan, inviting the right people, and resolving outstanding assignments.
Example
Make quality part of the definition.
“Reports created” rewards output volume. “Teams using a report to complete their weekly review” moves closer to value, but may be harder to observe directly.
Choose the best practical proxy and name its limits. Validate it with research, downstream behavior, and quality measures rather than assuming the label makes it meaningful.
Failure case
Optimizing one side of a system.
In a marketplace, more bookings can coexist with poor supplier earnings or low customer trust. A single measure may hide an unhealthy balance.
Use guardrails or a small set of complementary outcomes when necessary. A North Star should coordinate thinking, not forbid evidence that complicates the story.
PM decision
Check the causal chain.
If the team raises an input metric, why should the valuable outcome improve? More invitations might create collaboration, or merely send unwanted emails.
Investigate whether invitees become active contributors and whether the team completes better work. A metric tree is a set of testable relationships, not a law of nature.
Remember this
Shared focus needs visible limits.
Choose a value-oriented measure, connect it to inputs, and watch for quality or ecosystem damage. Retire a proxy when it stops representing the outcome you care about.