How to define a north star metric

A north star metric is the single number a company or product team treats as the best proxy for whether it is delivering value to customers in a way that will eventually show up as durable revenue. It exists to solve a coordination problem: without one, every team optimizes the metric closest to it — support optimizes ticket close time, marketing optimizes lead volume, engineering optimizes uptime — and those local optima can all improve while the business doesn't. A good north star gives disparate teams a single, shared thing to point their work at.

What makes a candidate metric actually work

Most north-star candidates fail one of these tests, which is why so many companies pick one, then quietly stop using it within a year:

• It reflects customer value, not company convenience. Revenue is tempting because it's unambiguous, but it's a lagging indicator that goes up even when the underlying product experience is getting worse, right up until it doesn't. A better north star sits upstream of revenue — the behavior that, when it happens, reliably predicts revenue will follow.
• It's measurable frequently enough to act on. A metric you can only compute quarterly can't guide weekly decisions. Weekly or daily cadence is the usual bar.
• It's a single number a whole company can understand, not a composite index only the analytics team can explain. If explaining the metric takes a slide deck, it won't survive contact with a hallway conversation.
• Teams can actually move it. A metric driven mostly by macroeconomic factors or by a handful of enterprise accounts isn't something day-to-day product and marketing work can influence, which makes it useless as a coordinating target.

Input metrics and guardrails around the north star

A north star metric on its own invites gaming — optimize the one number and something else breaks. Two supporting layers fix this:

Input metrics (sometimes called driver metrics) are the smaller, more controllable numbers that roll up into the north star. If "weekly active users completing a core action" is the north star, its inputs might be new-user activation rate, feature adoption rate, and reactivation rate — each one a lever a specific team can actually pull, with a clear line back to the metric that matters.

Guardrail metrics are the things that must not get worse while the north star improves: page load time, support ticket volume, unsubscribe rate, revenue per user. Without guardrails, a team can hit the north star by doing something that technically counts but damages the business — sending more notifications to inflate a weekly-active number, for instance, at the cost of long-term engagement.

Examples by business model

• Subscription product (B2B SaaS). A common choice is weekly or monthly active accounts completing a defined core workflow, guardrailed by retention and net revenue retention, so growth in activity doesn't mask accounts quietly churning.
• Marketplace. Often a two-sided metric like completed transactions, because it requires both supply and demand to be healthy simultaneously — optimizing either side alone can hide the other decaying.
• Content or media. Time spent or return visit frequency, guardrailed against content quality signals, since raw engagement can be juiced with low-quality, high-friction content that damages trust.
• B2B sales-led business. Often a product-qualified lead count or a product-usage signal that correlates with expansion, since revenue itself moves too slowly and too indirectly to guide weekly product decisions.
• Consumer app with a freemium model. Weekly active users completing the action that most strongly predicts eventual paid conversion, rather than total downloads or signups, which say nothing about whether the product delivers value.

Finding the right one for your business

The reliable method is retrospective, not aspirational: look at your existing retained, paying or engaged customers and ask what behavior, in their first weeks, most strongly predicted they'd still be around months later. That behavior — not a metric that sounds impressive in a board deck — is the north-star candidate worth testing. A cohort retention analysis segmented by early behavior is usually how this gets found: cohorts that hit a specific milestone early retain meaningfully better than cohorts that don't, and that milestone is your signal.

Tools that track behavioral cohorts and retention at the event level, such as Amplitude and Mixpanel, are where this analysis usually happens, since finding the predictive early behavior requires cutting retention by many candidate actions before one stands out. Where a north star has been chosen, it's common to track progress against it through the same OKR process the rest of the company uses — tools like Betterworks handle the goal-cascading and check-in side of that, though they report on progress people log, not instrumented product data, so the underlying metric still needs to come from analytics, not from a goals platform alone.

Questions to ask before you commit to one

• If this metric doubled tomorrow, would we be confident the business is healthier — or could it double for a bad reason?
• Can we compute it at least weekly, and does the team see it that often?
• Which specific teams can move this number, and do they know it's theirs to move?
• What are the two or three guardrails that must hold steady while we push this metric up?

Common mistakes

• Picking revenue itself as the north star, which measures the outcome rather than the upstream behavior teams can influence day to day.
• Choosing a metric with no guardrails, inviting optimization that technically hits the target while damaging the business.
• Changing the north star every few quarters as leadership changes, which destroys the multi-year comparability that makes it useful.
• Picking a composite or weighted-index metric that only the analytics team can explain or recompute.
• Treating the north star as fixed forever — a metric right for an early-stage product can become wrong once the product matures and a different behavior predicts value better.

For the retention work that usually surfaces the right candidate metric, see how to run a cohort retention analysis. Browse product analytics tools and performance management and OKR tools for the full field.