How do you measure your north star metric correctly?

A north star metric is only useful if it measures the moment a user gets real value from your product — not a proxy that flatters you. Most founders pick something easy to count rather than something meaningful to customers, then wonder why the number moves without the business moving with it. The fix is to define value delivery precisely, instrument it directly, and treat any gap between that metric and revenue as a diagnostic signal rather than a success story.

Define value delivery before you define the metric

Before you open any analytics tool, write one sentence that completes this prompt: 'A user has gotten real value from our product when they have ___.' The blank cannot be 'signed up,' 'logged in,' or 'completed onboarding.' It must describe an outcome the user cares about — a file shared with someone who opened it, a recurring booking made, a piece of code that passed tests. That moment of value is the only honest candidate for a north star.

This matters because measurement follows definition, not the other way around. If you define the metric first and then retrofit a story about why it represents value, you will find yourself celebrating numbers that precede churn. Paul Graham's observation in 'Do Things That Don't Scale' that early Airbnb's survival depended on direct, hands-on user engagement illustrates how fragile the gap between surface activity and genuine value delivery is — especially early on, when the temptation to read weak signals as strong ones is highest.

Once you have the definition, pressure-test it with three questions: Does this moment predict that the user will still be here in 90 days? Does it correlate with word-of-mouth referrals? Does it lead, not lag, revenue? If the answer to any of these is unclear, your definition is still too loose. Tighten it until you can point to a specific event in your data infrastructure and say 'that row, that timestamp.'

Instrument the metric at the event level, not the session level

Aggregate dashboards lie by omission. If your north star is 'weekly active users who completed a core action,' the session-level view will show you a healthy-looking curve while hiding the fact that 80% of those sessions involve users who completed the action once six weeks ago and are now just browsing. Instrument at the event level: every occurrence of your value-delivery moment should fire a distinct, named event with a user ID, a timestamp, and enough context to segment it later.

From that event stream you want three derived numbers updated daily: the raw count of value-delivery events, the count of unique users who experienced them, and the ratio of those users who experienced the event more than once in the past 28 days. The third number is your real north star. Single-event users are often curiosity; repeat-event users are forming habits. The gap between the first and third number tells you whether your activation is converting to retention.

Keep the instrumentation minimal but exact. One well-defined event tracked reliably beats a dozen loosely defined ones tracked inconsistently. If your engineering team disagrees about which user action fires the event, the definition is wrong — go back to the prose statement of value delivery and get alignment there first.

Separate leading from lagging, and act only on leading signals

Revenue is the lagging confirmation that your north star is real. The north star itself should be a leading indicator — something that moves weeks or months before revenue does. The test: if the metric drops today, should you panic today, or will you only feel it in the P&L three months from now? A true north star gives you enough lead time to diagnose and correct before the damage is financial.

This means you need to know the average time lag between a rise or fall in your north star and the corresponding movement in revenue. Measure it empirically: pull two time series, offset one by various intervals, and find where the correlation is highest. That offset is your leading window. If it's two weeks, a dip in north star events today is a warning you can act on. If it's six months, you need an even earlier-leading metric upstream of the north star — something like 'users who completed the value-delivery event within 72 hours of signup,' which predicts whether the north star cohort will be healthy.

Avoid the trap of using cohort averages to hide dispersion. A cohort where 20% of users hit the north star 10 times and 80% hit it zero times will look identical in the average to a cohort where every user hits it twice. Those two cohorts have completely different futures. Always look at the distribution, not just the mean.

Run weekly root-cause reviews, not monthly metric check-ins

The most common failure mode is reviewing the north star on a monthly cadence and treating any upward trend as permission to move on to other work. Weekly reviews force you to explain movement — both increases and decreases — at a level of specificity that monthly reviews never demand. If the metric rose 12% last week, you should be able to say exactly which user segment drove it, which acquisition channel those users came from, and whether their value-delivery events are the kind that predict retention or the kind that don't.

For any week where the metric moves more than one standard deviation in either direction, run a root-cause analysis before discussing solutions. What changed in the product? What changed in the acquisition mix? What changed in the macro environment? The goal of a root-cause review is to avoid the same class of problem recurring — precisely the framing Garry Tan's engineering documentation emphasizes for bug analysis, and it applies equally well to metric anomalies. A metric that moved for an unknown reason will move again for the same unknown reason.

Build a simple decision log alongside your metric dashboard: date, metric value, hypothesized cause, action taken, result observed four weeks later. This creates the institutional memory that lets you distinguish a real trend from noise, and it prevents the organization from rediscovering the same explanations every quarter.

Avoid the three most common north star mistakes

The first mistake is measuring supply when you should be measuring demand. If you run a marketplace, tracking listings created is a supply metric; tracking successful transactions is a demand metric. Only demand metrics tell you whether users are getting value. Supply metrics tell you whether your growth team is doing its job, which is a different and less important question at most stages.

The second mistake is picking a metric that cannot be gamed internally. If your north star is 'daily active users' and your team can goose it by sending more push notifications, you will eventually find yourself sending more push notifications instead of building a better product. A good north star is hard to move without actually delivering value. If your team can think of an obvious way to inflate it without improving the user experience, add a constraint: 'daily active users who completed a core action without being prompted by a notification in the past 7 days.'

The third mistake is keeping the north star constant as the business evolves. Your north star in the first 100-customer phase may legitimately be different from the right metric at 10,000 customers, because the value delivery moment may shift as the product matures. Revisit the definition at each major inflection point — new product line, new customer segment, new business model — and be willing to replace it rather than stretch it to cover territory it was never designed to measure.

“Almost all startups are fragile initially... the big danger is that you'll dismiss your startup yourself.”

— Paul Graham, source

The one thing to do

Write the single sentence that describes when a user has received real value from your product, find the exact event in your data that marks that moment, and make that event's 28-day repeat rate your north star — everything else is a supporting metric.

Frequently asked questions

How many north star metrics should a startup have?

One primary north star, plus one or two leading-indicator metrics that predict it. More than three metrics at this level splits attention and makes it impossible to know which number to optimize when they conflict.

What is the difference between a north star metric and a KPI?

A north star metric measures the moment customers get core value from your product; KPIs measure the health of specific functions or channels. KPIs should serve the north star, not compete with it. If a KPI is moving in the right direction while the north star is flat, the KPI is probably measuring the wrong thing.

How do you know if your north star metric is actually predictive of long-term success?

Run a cohort analysis: segment users by whether they hit your north star event in their first week, then track 6-month retention and lifetime value for each group. If users who hit the event retain and pay at materially higher rates, the metric is predictive. If the difference is small, your definition of the value moment is wrong.

Should your north star metric change as you scale?

Yes, at major inflection points. The right metric for validating that your product delivers value to 100 early adopters is often not the right metric for optimizing growth across 100,000 diverse users. Revisit the definition whenever your core customer profile or primary use case shifts significantly.

Sources

More playbook answers · Growth Prophet home