Every team I've worked with has wanted a north star metric at some point. Usually it shows up after a planning session where everyone presented a different chart and nobody could agree what "doing well" meant.
The idea is good. One number that reflects the value users get, which the whole team can rally around. In practice, picking the number is the easy part. Making sure it doesn't quietly push the team in the wrong direction is harder.
What makes a decent north star
I look for three things:
- It goes up when users get real value, not just when they show up.
- The team can influence it through the work they actually do.
- It moves often enough to learn from. A yearly number won't guide a sprint.
Revenue usually fails the first test. It's important, but it's a result of value, and it lags. Daily active users often fails too, because people can open an app and get nothing out of it.
Amplitude's north star playbook frames it as the metric that best captures the core value your product delivers, supported by a small set of input metrics the team can work on directly. 1
Some examples from products I know
These are illustrative, not the official metrics of any company I worked at.
| Product type | Possible north star | Input metrics | Guardrail |
|---|---|---|---|
| Subscription streaming | Weekly viewers who watch a meaningful amount | Successful play starts, discovery to play, new subscriber first-week viewing | Play failure rate, cancellation rate |
| Travel booking | Completed bookings that aren't cancelled or disputed | Search to checkout, payment success, confirmation clarity | Refund requests, support contacts per booking |
| B2B AI content tool | Generated outputs accepted and used without heavy edits | Weekly active teams, requests per team, acceptance rate | Review time, error reports |
| B2B marketplace | Repeat transactions between buyers and suppliers | Quote responses, time to quote, order completion | Disputes, supplier drop-off |
Notice that each one has a guardrail. A north star with no guardrail is how teams end up optimising something that looks great and annoys everybody.
Input metrics do the daily work
Nobody can work on "weekly meaningful viewers" directly on a Tuesday. They can work on making the play button start video faster on a particular TV platform, or on making the first row of the home screen more relevant.
So I break the north star into two to four input metrics, and those are what show up in sprint goals and experiments. The north star is how we check, every month or so, that the inputs are adding up to something.
Signs your north star is misleading you
A few things I've learned to watch for:
- It went up and nobody can explain why. Could be seasonality, a tracking change, or one big customer.
- It's easy to game. If sending more notifications lifts it, someone will eventually send more notifications.
- Teams argue about the definition every quarter. That usually means the metric is doing too much.
- It moves the opposite way from what support is hearing. Trust the complaints and dig in.
When any of these show up, I don't throw the metric out right away. I go back and check the definition, the denominator and the guardrails.
Keep it small and written down
The definition should fit in two sentences, with a note on what counts and what doesn't. For example: "A meaningful viewer is a subscriber who watches at least N minutes of long-form content in a calendar week. Trailers and autoplay previews don't count."
Pick N with the team and write down why you picked it. If the reason changes later, change the number openly.
A north star helps you say no to work that won't move it. That's most of its value, and it only works if people trust the number. I'd spend the extra week getting the definition right before putting it on any slide.
If you're thinking about how this connects to planning, the outcome-based roadmap example shows how I'd tie a goal like this to actual work.