North Star metrics: the one number that actually moves a business

Sean Ellis popularised the term, but the idea predates it by decades: every healthy business has one number that, when it grows, reliably signals that real customers are getting real value. Airbnb tracks nights booked. Spotify tracks time spent listening. WhatsApp tracks messages sent. Facebook, at its high-growth phase, obsessed over daily active users. Each of these numbers is a North Star Metric — and each one was chosen with unusual care.

The companies that chose poorly — tracking raw sign-ups, total downloads, or gross revenue as their primary focus — discovered a familiar trap: those numbers go up even when the product is failing. Downloads grow through paid acquisition while retention crumbles. Revenue grows through price increases while customers quietly churn. A metric that can rise while the business is dying is not a North Star. It is a vanity metric wearing a suit.

What a North Star Metric is (and is not)

A North Star Metric (NSM) is the single leading indicator that best captures the value your product delivers to customers, and that reliably predicts long-run revenue. The word “leading” is the crux: the NSM should move before revenue moves, not after.

This disqualifies a large class of candidates immediately. Revenue, GMV, and profit are lagging indicators — they are outputs of value delivered, not evidence of it. When revenue drops, something upstream already broke weeks or months ago. Using revenue as your NSM means you find out about problems after they have compounded.

It also disqualifies ratios and rates you cannot directly act on. NPS is frequently proposed as a North Star. It has real value as a diagnostic and a guardrail. But no team wakes up and runs an experiment whose goal is “move NPS by two points.” NPS is an outcome of dozens of upstream product and service decisions. It is too distal to drive prioritisation.

A genuine NSM has four properties:

1. It reflects customer value directly — growth in the metric means more customers are getting more of what they came for.
2. It leads revenue — historically, when the NSM rises, revenue follows within a predictable lag.
3. Teams can move it — product, growth, and engineering can run experiments whose primary outcome is the NSM.
4. It is a counter or volume (not a ratio) — you can measure absolute progress, not just efficiency.

Input vs output, leading vs lagging

Every metric in a business sits somewhere on two axes: input or output, and leading or lagging.

Output metrics — revenue, profit, churn rate — measure what has already happened. They are essential for understanding the health of the business, but they react to decisions made months earlier. You cannot course-correct in real time using a lagging output.

Input metrics — features shipped, experiments run, support tickets resolved in under four hours — measure activities that are within a team’s direct control. They are highly actionable but only useful if you know which inputs actually cause the output you care about.

The NSM sits in a specific position: it is an output from the customer’s perspective (they received value) but a leading indicator for the business (that value will convert to revenue). Nights booked is an output of Airbnb’s matching, pricing, and trust systems — and a reliable predictor of the host and guest fees that follow.

The metric tree: decomposing the NSM

Choosing a North Star is the strategic act. Making it actionable requires decomposing it into a small set of input metrics that product teams can actually influence. The structure looks like a tree: the NSM at the root, four to six input metrics as branches.

A useful decomposition framework breaks the NSM into four multiplicative drivers:

NSM = Breadth × Depth × Frequency × Efficiency

• Breadth: how many customers are using the product at all (reach)
• Depth: how much of the product’s core value each customer extracts per session
• Frequency: how often they return
• Efficiency: how little friction they encounter getting there

Not every NSM decomposes neatly along all four axes, but the structure forces teams to think about which lever is actually constraining growth. Most stalled products have a frequency problem masquerading as a breadth problem.

The tree is not just an org-chart exercise. It makes the strategy legible: if nights booked is stagnating, is it a supply problem (breadth), an experience problem (depth), or a loyalty problem (frequency)? Each answer points to a different team and a different experiment backlog.

Anti-patterns: metrics that lie to you

Vanity metrics grow independent of value delivered. Raw sign-ups, total page views, and cumulative downloads are the classics. A mobile game with ten million downloads and 0.1% day-30 retention is not a product with traction — it is a leaky bucket with a big top opening. Optimising sign-ups without checking if those users ever activate is the most common growth mistake in early-stage product work.

Gameable metrics collapse under incentive pressure. If a support team’s NSM is “tickets closed per hour,” they will close tickets quickly regardless of resolution quality. If a sales team’s NSM is calls made, they will make short calls. Any metric that a team can move without doing the actual work of delivering value is a gameable metric, and it will be gamed.

Revenue as NSM sounds sensible until you try to use it. Revenue is correlated with everything — a price increase moves it, a one-time promotional burst moves it, a mix shift toward higher-margin customers moves it — none of which necessarily signals that customers are getting more value. Revenue is also reported on a lag, often with quarterly smoothing. By the time a revenue problem is visible, the product problem behind it is months old. Revenue belongs on your executive dashboard and your guardrail set. It does not belong at the top of the metric tree.

Ratios without volume are another trap. Conversion rate is a useful diagnostic but a dangerous NSM. You can raise conversion rate by removing low-intent traffic — which might reduce overall bookings. Efficiency ratios always need a volume companion, which is why the NSM is usually a volume number.

How to choose yours

Run this filter in order:

Does it measure value delivered to the customer, not activity by the company? Sessions started, emails sent, features shipped — these are internal activity metrics. The NSM must capture something the customer actually received: a file shared, a task completed, a night stayed, a song heard to completion.

Does it lead revenue historically? Pull two or three years of data and check whether the candidate metric correlates with revenue with a meaningful time lag. If it does not lead, it is a coincident or lagging indicator and will not give you early warning.

Can a cross-functional team run an experiment against it? If the answer requires waiting for a quarterly cohort to mature before you can read a result, the feedback loop is too long to drive product decisions. Good NSMs have result cycles of days to a few weeks.

Is it a volume or count, not a ratio? Prefer “messages sent this week” to “messages per active user.” The latter is an important diagnostic but can be raised by churning low-engagement users — a perverse incentive.

Guardrail metrics: the NSM’s immune system

Optimising a single number creates a gradient. Engineers and PMs will follow that gradient, and without guardrails they will follow it into places the business does not want to go.

Guardrail metrics are the metrics the NSM must not harm as it grows. Common guardrails include:

• Retention (day-7, day-30 active rate) — the NSM should not grow by pulling in low-quality activations that churn immediately.
• Support ticket volume — engagement growth driven by confusion is not value growth.
• Revenue per engaged user — you should not grow the NSM by subsidising activity that destroys unit economics.
• Content or host quality scores — a marketplace NSM can grow by degrading supply quality; guardrails catch this.

The formal rule: any experiment that moves the NSM in the right direction but moves a guardrail metric in the wrong direction by a statistically significant amount should be rejected. The guardrail is a hard constraint, not a soft preference. This is closely related to the question of what it means for a test to “win” — a topic examined in detail in the post on A/B testing for business decisions.

For a complementary view on how metric choices interact with cohort health over time, the post on cohort retention analysis shows how to read the long-term signal that your NSM is pointing at.

Putting it together

The North Star framework is not a metrics layer on top of your roadmap. Done properly, it is the roadmap. When the decomposition is right, every team’s quarterly objective maps onto one leaf of the metric tree, and every experiment has a primary metric that connects upward to the NSM.

The practical sequence: agree on the NSM at leadership level (this is a strategic conversation, not a data conversation), decompose it into four to six input metrics that different teams own, define three to five guardrails, then let each team build their own sub-tree of diagnostic metrics below their input metric.

The result is not a dashboard of forty KPIs that nobody reads. It is a single number that every person in the company can explain, connected to a small set of actionable levers that teams can pull — with guardrails that prevent optimisation from eating the business it was supposed to grow.

For a broader orientation to the methods behind this kind of business measurement, the Business Analytics section covers the full analytical stack. If you are building the decomposition from scratch and want to understand why averages in your diagnostic metrics might be hiding the real story, the post on averages that lie is directly relevant. Definitions for all the metric types discussed here are in the glossary.

---

Frequently asked questions

What is a North Star Metric in simple terms? A North Star Metric is the single number that best captures the value your product delivers to customers and reliably predicts long-run business growth. Unlike revenue or profit, which lag behind reality, a good NSM moves first — it gives teams an early signal that the product is working (or failing) before it shows up in the financials.

Can a company have more than one North Star Metric? Most practitioners recommend one primary NSM, precisely because the discipline of choosing one forces clarity about what the product’s core value proposition actually is. Some companies define a primary NSM and one or two secondary “constellation” metrics for distinct business lines, but if every business unit has its own North Star, the term has lost its meaning. The constraint is the point.

How is a North Star Metric different from a KPI? KPIs are general management metrics — there can be dozens of them measuring different parts of the business. The NSM is a specific concept: the one leading, customer-value-reflecting metric that the whole company organises around. All input metrics in the NSM tree are KPIs, but not all KPIs are part of the NSM tree. The NSM is the apex of a hierarchy; KPIs are the hierarchy.

What happens when your North Star Metric stops working? Markets and products evolve, and an NSM can decay. Facebook’s long obsession with DAU became less meaningful as the platform matured and engagement patterns fragmented. When the NSM decelerates despite input metrics improving, or when it starts to decorrelate from revenue, it is a signal to revisit the decomposition — not to add more metrics to the dashboard, but to honestly re-examine whether the product’s core value proposition has shifted.