North Star Metric

Definition

A north star metric is the single measure a company treats as the best summary of customer value over time. It is the number the product, marketing, and experiment program are trying to move in the long run. It is not the primary metric of every A/B test.

What North Star Metric means in A/B testing

The star has three jobs. It describes value from the customer's side, not only the company's. It is a rate or a count you can track on a cadence, usually weekly or monthly. And it is one metric, so a trade-off has a referee. Inputs sit underneath it. For a store, revenue per visitor can be the star, with conversion rate and average order value as inputs. For a product, it might be the count of accounts that complete the core action in a week. A guardrail metric is narrower: it is what one test must not harm. The north star is what the series of tests is for.

Why North Star Metric matters

Without it, each team optimizes the metric it can move this month. Support deflects tickets by hiding the contact link. Growth adds steps that raise signups and cut activation. Both dashboards look good. The customer is worse off, and nobody's chart says so. Experiments need the star so a winning test on a local metric can still be rejected when it pulls the star the wrong way. They also need patience: most tests are too short to call the star directly.

Example of North Star Metric

A project tool's north star is weekly active projects that receive a completed task. A test shortens onboarding and lifts account creation by 12%. Activation, the share of new accounts that finish a task in seven days, falls from 40% to 31%. Account creation was the test's primary metric. The star, measured on the same cohorts a week later, does not rise, because the extra accounts do not do the core job. You do not ship on the signup lift. You rewrite onboarding so the people who arrive can complete a task.

How to use North Star Metric

Pick a metric a customer would recognize as value, and that you cannot inflate with a trick that makes the product worse. Name the two or three inputs that multiply into it. For each experiment, choose a leading primary metric and say how it is supposed to reach the star. Review the star on a fixed cadence, not at the end of every test. If a shipped winner and the star diverge for a month, treat that as a finding, not as a delay in the dashboard.

Common mistake

A common mistake is promoting whichever KPI improved last quarter into the north star. The star should stay stable enough that this quarter's tests are comparable with last quarter's. Another is using revenue alone when the product is paid by a sponsor and the user is someone else. The star has to match the person whose value you claim to grow.

Related A/B testing terms

• Metrics
• Guardrail Metric
• Conversion Rate
• Revenue per Visitor (RPV)

FAQ

Can a company have three north stars?

It can have three important metrics. Calling all three the north star removes the point, which is a shared priority when two teams want opposite moves. Keep one north star and let the others be inputs or guardrails. If two businesses share a brand, they may honestly need two stars. Write that down rather than averaging them into a number nobody owns.

Should every experiment use the north star as its primary metric?

Usually no. A homepage headline will not move monthly retained revenue inside a two-week test, and powering that test on the north star will make almost every result inconclusive. Use a leading metric the change can affect, and watch the north star on a holdout or a longer readout so a local win cannot quietly shrink it.

What makes a bad north star?

A count you can raise while customers get less value. Raw signups, raw page views, and time-on-site are typical failures: they go up when the product gets harder or noisier. A better star includes the valuable action, such as a retained account that completes the core job, or revenue per visitor when revenue is the value the business is paid for.