Choosing a North Star Metric
One number that represents delivered value, moves in weeks, and cannot be faked easily.
Leadership wants one metric on the wall. Someone proposes MAU: Monthly active users: how many distinct people used the product at least once in a month.. Someone else says revenue. A third person says NPS. All three are defensible, all three are wrong in a specific way, and the argument goes around the room twice before someone suggests tracking all of them, which is how you end up with a dashboard nobody reads.
A North Star metric: The single measure a team agrees best represents the value customers get. is a proxy for the value your customers get, chosen because you believe it leads revenue. That belief is the whole thing. If you cannot explain the causal chain from the metric to money, you have picked a vanity number with good PR.
Four tests, and the one that is usually skipped
Four tests for a candidate metric
- Does it represent value received?
- The user got something they wanted. Signups do not qualify. A completed booking, a delivered message, a reconciled month all do.
- Does it lead revenue?
- You should be able to show that accounts with more of this convert or retain better. If you have never checked, check before you commit.
- Does it move within a planning cycle?
- A metric that only responds over a year cannot steer a team over six weeks. It belongs in the board deck, not on the wall.
- Is it hard to game?
- Ask how you would move it in a week if you were cynical. If the answer is easy and harmful, someone will eventually do it accidentally.
Weak candidates
- Monthly active users: counts arrival, not value
- Page views: rewards friction and confusion
- Revenue: lags by a quarter and is moved by pricing and sales, not product
- NPS: slow, noisy, and driven by who responded
Stronger candidates
- Nights booked (marketplace)
- Messages delivered to an active recipient (communication)
- Weekly reconciled accounts (finance tooling)
- Shifts published on time (workforce tooling)
Worked example
Hypothetical: Marrow, a meal planning app
Marrow starts with monthly active users at 120,000. The number is flat, and nobody can tell whether that is good. The team looks for an action that predicts retention and finds that in a sample of accounts, those that cooked from a plan at least twice in their first month were still active at month three far more often than those that opened the app the same number of times without cooking. So the candidate becomes weekly plans cooked from. It sits at 31,000 a week, which is a smaller and more uncomfortable number, and that discomfort is the point: it excludes people who browse recipes and never cook, which is exactly the population MAU was hiding. Note the guard rail needed here. You could inflate plans cooked from by prompting people to mark a plan as cooked, so the team pairs it with a check that marked plans correlate with grocery list usage.
Break the North Star into inputs, otherwise it is a scoreboard nobody can play. For Marrow, weekly plans cooked from decomposes into active planners, plans created per planner, and share of plans that get cooked. Each team can own one input. This is where the metric stops being a poster and starts being a way to divide work.
One structural warning. A single North Star works when the business has one engine. If you run a two sided marketplace or a genuine platform, one number will systematically hide one side, and you are better off with a small paired set plus an explicit statement about which one you optimise when they conflict.
Be ready for the North Star to feel like a demotion when you first publish it. Going from 120,000 monthly active users to 31,000 weekly plans cooked from is a worse headline in every meeting where someone wants a big number, and a PM who cannot defend that trade will quietly keep reporting both until the old one wins. Defend it by showing the retention gap that justified the switch. A smaller number that predicts revenue beats a larger number that predicts nothing, but you have to prove the first half of that sentence, not assert it.
Set the target separately from the definition, too. Choosing what to measure is an analytical question and choosing what number to hit is a planning question, and teams that decide both in the same meeting tend to pick whichever metric makes the target look achievable.
Pair whatever you choose with a counter metric: A metric watched alongside a target to catch the damage that hitting the target might cause. that would catch the damage if you pushed too hard. Volume paired with quality, speed paired with error rate, growth paired with retention. The counter metric is not decoration. It is the thing that lets you notice you have started winning the wrong game.
Quick check
Why is monthly active users a weak North Star for most products?
The takeaway
A North Star measures value delivered, predicts revenue, responds within a quarter, and travels with a counter metric.
Try this tomorrow
Write the causal chain from your current headline metric to revenue in four steps. If any step is a guess, go and check it this week.
Answer the check above, then bank the day.
Where this comes from
- The North Star Playbook, Amplitude
- Measure What Matters, John Doerr
- Hacking Growth, Sean Ellis and Morgan Brown
Metrics is one of six tracks. These lessons summarise and build on the work above, they do not reproduce it. Buy the books, they are better.