A north star metric is the single measure a team agrees stands for the value its product delivers to customers, with a small set of input measures beneath it that the team's own work moves directly.
Sean Ellis is credited with the term, and it spread through the growth teams he built and advised after he coined the phrase growth hacking in 2010. The idea reached a wider product audience in December 2019, when Amplitude published a North Star Playbook written with the product coach John Cutler, which set the single measure inside a structure of inputs and gave the whole arrangement a name.
The problem it addresses shows up in any team large enough to split its work. A group of nine people with four measures on the wall spends the first twenty minutes of every planning session arguing about which measure this quarter's work is meant to serve. The argument never resolves, because all four measures are defensible and nothing says which one outranks the others. A north star answers that one question and nothing else.
The sections below say what a north star is meant to capture and how it differs from a revenue figure. They then show the inputs beneath one, worked through with arithmetic that reconciles. Finally they give the four properties a candidate has to pass and the parts of a business the measure leaves untouched.
What a north star metric is meant to capture
A north star counts a moment when a customer got what they came for.
Revenue records what the company captured. A north star records what the customer received, and the two figures move together only for as long as the product is doing its job. A company can raise revenue for three quarters by discounting hard, by selling annual contracts to people who will not renew and by charging for a module nobody opens. Every one of those shows up in revenue before it shows up as a problem, and none of them moves a measure of delivered value at all.
Two questions settle whether a candidate captures delivered value. If this number doubled and nothing else changed, would customers be better off. If it doubled, would the business be better off. A measure that earns a yes to both is a candidate. A measure that earns a yes to only the second one is a revenue proxy wearing a different name.
The inputs beneath a north star metric
A north star is usually too far from any one team's work to be moved directly, which is what the inputs are for.
Four hundred practices with an available vet, ten bookings each a week, nine bookings in ten attended and nineteen completed consultations in twenty written back into the practice record gives 3,420. Each input is small enough for one team to attack, and the four multiply out to the measure above them.
The product in that example lets small veterinary practices run remote consultations and writes the notes back into the practice record system. Its north star counts completed consultations each week, because a completed consultation with a written record is the moment a practice got the thing it pays for.
Four inputs multiply out to the star. The number of practices with a vet available for booking, the bookings each of those practices takes in a week, the share of bookings the vet actually attended and the share of attended consultations whose notes reached the practice record system. Multiplying those together gives 400 times 10 times 0.90 times 0.95, which is 3,420 completed consultations a week.
The inputs are what make the arrangement useful, because each one belongs to somebody. A team working on scheduling owns vet availability. A team working on the booking screen owns bookings per practice. A team working on call quality owns attendance. And a team working on the integration owns the share of records that arrive.
The arithmetic also prices the work. Raising attendance from 90 per cent to 93 per cent, with the other three inputs unchanged, gives 400 times 10 times 0.93 times 0.95, which is 3,534 consultations a week. That is 114 more consultations, or a rise of 3.3 per cent in the north star, from a three point improvement in one input. A team can decide whether three months of work on call reliability is worth 114 consultations a week before anybody starts.
The four properties a candidate has to pass
A measure becomes a north star by surviving four tests, and a candidate that fails any one of them costs the team a quarter before the failure is obvious.
- Understandable. A person who joined last week can say what the number counts in one sentence, without a definition document.
- Measurable. The product already emits the events the measure is counted from, or a team can make it emit them within a few weeks.
- Sensitive to the team's work. A release the team ships moves the figure inside a period the team can wait for, which usually means weeks.
- Hard to game. No cheap change moves the number without a customer being better off.
The gaming test is the one teams skip, and it is the one that costs most. Counting consultations started would pass the first three tests and fail this one, because a call that dropped after four seconds would be counted as progress. Counting completed consultations with a written record survives, because the only route to a higher number runs through a vet finishing a consultation.
A fifth consideration sits beside the four tests without being one of them. A north star should run ahead of revenue. A measure that only moves after the money has arrived reports on a quarter that has already closed. No team can act on a report about a closed quarter.
What a single measure cannot describe
A north star describes delivered value and describes nothing else, which is worth saying out loud before a team pins one to the wall.
Cost sits outside it. The veterinary product could raise completed consultations by routing every call through a human operator, and the north star would rise while the margin collapsed. Support load sits outside it. Churn concentrated among the twelve largest practices sits outside it, because those practices are a rounding error in a weekly count of consultations and most of the revenue. Risk sits outside it entirely.
A single measure also breaks at the wrong altitude. A company running three products for three different kinds of customer cannot hold one north star between them without choosing a number so general that no team can move it. In that situation each product carries its own, and the company holds its financial measures separately.
Time sets a third limit. A north star rewritten every quarter is a quarterly goal wearing a grander name, and it loses the property that made it worth having, which is that several teams and several quarters point the same way. A north star that has survived two years of pressure without being rewritten is doing its job.
Deriving a measure this way takes a workshop, several arguments and a week of checking what the product actually emits. Published frameworks take the other route and hand a team a ready made set of categories to fill in, and the next page takes the two a product manager meets most often.