In short
A north star metric is the single number that best represents customers receiving the value your product promises, chosen so it cannot rise while the customer experience deteriorates. Good ones measure delivered value — messages sent, nights booked, projects completed — rather than activity or revenue. The test is whether you could improve the number by making the product worse.
What it is for
A north star metric exists to align decisions. When a team of twenty is making choices weekly, a shared definition of better prevents everyone optimising for something different.
It is not a target, a KPI set, or a revenue forecast. It is one number chosen because it goes up when customers are getting what they came for.
The test that eliminates most candidates
Could you improve this number by making the product worse?
If yes, it is the wrong metric.
Pageviews. Improved by splitting content across more pages, breaking search, and adding pagination. All worse for users.
Time in app. Improved by making things harder to find. Fine for entertainment, terrible for a tool where efficiency is the value.
Signups. Improved by removing information people needed to decide, producing users who leave.
Revenue. Improved by raising prices, adding fees, and making cancellation difficult — for a while.
A good north star fails this test cleanly. You cannot increase nights actually stayed or messages received by someone who replied by degrading the product.
What good ones have in common
They measure delivered value, not activity. Not documents created but documents shared. Not searches run but searches that ended in a result being used.
They include a quality condition. "Messages sent" can be spam. "Messages sent that received a reply" cannot easily be.
They are countable frequently. A metric you can only measure quarterly cannot guide weekly decisions.
They are influenceable. The team must be able to move it through work they control.
They lead revenue rather than lagging it. Revenue is an outcome; the north star should be the thing that produces it, so it moves earlier and tells you something in time to act.
Choosing one
Start from the promise. What does someone hire your product to accomplish? Then find the countable event closest to that happening.
For a booking product it is completed stays, not bookings — a booking that gets cancelled delivered nothing. For a design tool it might be projects handed off. For a support product, tickets resolved without escalation.
Then apply the test. Then check whether you can actually measure it, since the perfect metric you cannot instrument is worse than a good one you can.
Counter-metrics are not optional
Any single metric can be gamed, usually without anyone intending to.
So pair it with two or three counter-metrics that would reveal the damage. If the north star is completed orders, watch refund rate and support contacts. If it is documents shared, watch whether recipients open them.
A north star without counter-metrics is a hill people will climb by any route available, and the routes that damage the product are frequently the easiest ones.
Where teams go wrong
Choosing revenue. It lags, it is influenced by things outside the team, and optimising it directly encourages extraction rather than value.
Choosing something nobody influences. If the team cannot move it through their work, it is a report rather than a guide.
Changing it frequently. The value comes from a stable shared definition. Reconsider annually, not quarterly.
Treating it as the only thing. It guides prioritisation; it does not replace judgement, and some important work — security, accessibility, technical debt — will never move it.
Setting it as a target. A metric used to evaluate people stops being a measurement, which is the reliable failure mode of any number attached to performance.
When one number is the wrong idea
Be honest that this does not fit everything.
Marketplaces have two sides with different health, and a single number hides one of them.
Early products have too little data for any metric to be stable, and chasing one substitutes for talking to the handful of users you have.
Agencies and services businesses deliver value per engagement rather than through repeated product use — what to measure there looks different.
Products serving distinct segments may need one per segment.
In those cases a small dashboard beats a forced single number.
Making it useful
A north star only works if people encounter it. Put it somewhere visible, refer to it when prioritising, and be explicit when a decision is expected to move it.
And break it into inputs. "Completed projects" is composed of projects started, completion rate, and time to complete — those are the things a team can actually work on, and connecting daily work to the top-level number is what makes it more than a poster.
If your team is measuring several things and agreeing on none of them, book a call.
Common questions
What is a north star metric?
The single number that best represents customers receiving the value your product promises, used to align decisions across a team. It is not a target or a revenue forecast — it is chosen because it rises when people are genuinely getting what they came for.
How do you know if a north star metric is wrong?
Ask whether you could improve it by making the product worse. Pageviews improve by splitting content across more pages; time in app improves by making things harder to find; signups improve by withholding information people needed. A good metric fails that test cleanly.
Should revenue be a north star metric?
No. Revenue lags the value that produced it, is influenced by factors outside the team's control, and optimising it directly encourages extraction — raising prices, adding fees, obstructing cancellation. The north star should be the thing that leads revenue.
Why do you need counter-metrics?
Because any single metric can be gamed, usually unintentionally. Pair the north star with two or three numbers that would reveal damage — refund rate and support contacts alongside completed orders, for example. Without them, a north star is a hill people will climb by whatever route is easiest.
When is a single north star metric the wrong approach?
For marketplaces, where one number hides the health of one side; for very early products with too little data for stability; for services businesses delivering value per engagement rather than through repeated use; and for products serving genuinely distinct segments, which may need one metric each.
