North Star Metric: Pick One Number and Build Its Input Tree

Why this matters

The AARRR funnel shows you five stages. The problem is that five stages create five teams with five priorities. Marketing wants signups. Product wants feature usage. Sales wants pipeline. Each team can hit its target while the company stalls.

A North Star Metric (NSM) is the single number that best captures the value customers get from your product. When it grows, the business grows in a durable way. It gives every team the same scoreboard and turns arguments about priorities into a simple question: which of these ideas grows the North Star most for the least effort?

The idea gained wide use in growth teams in the 2010s, and the core logic is simple. Revenue is the result of customers getting value. If you measure the value directly, you see problems weeks or months before revenue shows them.

The framework

Five tests for a good North Star

A good NSM is usually an active count of a core action in a time window. “Weekly active teams that ran at least one report.” “Customers with a repeat order in the last 90 days.” “Nights booked.” Ratios and averages make poor North Stars on their own because they can rise while the business shrinks.

The input tree

The NSM is an outcome. Nobody can move it directly. You need the levers underneath it. Most North Stars decompose into four types of input.

Write the tree out for your business. Each leaf should be a metric that one team can own.

Weekly active teams running reports  (NSM)
├── New teams activated this week         (Breadth: acquisition + activation)
├── Retained active teams from last week  (Breadth: retention)
├── Resurrected teams                     (Breadth: win-back)
├── Reports run per active team           (Frequency)
└── Report success rate                   (Efficiency: data connected, no errors)

Guardrails

Every North Star can be gamed. Pair it with two or three guardrail metrics that must not get worse while the NSM goes up. Common ones: contribution margin per order, 90-day retention, unsubscribe and complaint rates, refund rate, and support tickets per active customer.

What a North Star looks like by business model

Step-by-step playbook

1. List what customers come for. Write one sentence per segment: “Finance teams use us to get invoices paid faster.” The NSM must count that outcome, not your activity.
2. Draft three candidates. For each, write the exact definition: event, unit (user, account, order) and time window.
3. Run the five tests. Score each candidate on the five tests from the table above. Drop any that fail customer value or leads revenue.
4. Check the revenue link with data. Take customers from a year ago. Split them by how much of the candidate metric they had in their first 60 days. If the top group retains or pays much more than the bottom group, the metric leads revenue.
5. Build the input tree. Decompose the winner into breadth, depth, frequency and efficiency. Aim for 3 to 5 leaves, each owned by one team.
6. Model the levers. For each input, ask what happens to the NSM if it improves by the same relative amount, say 10%. The input with the biggest modelled effect and a credible way to move it gets staffed first.
7. Set guardrails and targets. Add two or three guardrails. Set a quarterly target for each input, not just the NSM.
8. Publish one weekly dashboard. NSM at the top, inputs below, guardrails at the side. Review it in the same meeting every week.

Worked example

Say a B2B expense management SaaS picks weekly active companies that submitted at least one expense report as its NSM. Last week the number was 2,000. The team models it as states, the same way Duolingo models learners.

1,700 + 220 + 60 = 1,980. The NSM fell slightly, because 300 companies dropped out and only 280 came in.

Now compare three levers, each improved by the same relative 10%:

• Weekly retention 85% to 93.5%: 2,000 × 0.935 = 1,870 retained, +170.
• Activation 40% to 44%: 550 × 0.44 = 242, +22.
• Resurrection 2% to 2.2%: 3,000 × 0.022 = 66, +6.

Retention is the lever. A 10% relative lift there is worth almost 8 times the same lift in activation this week, and the effect compounds every week after. This is the same pattern Duolingo found with CURR: once a product has a large active base, retaining current users usually dominates.

Checking that the NSM leads revenue

Before committing, the same team tests the revenue link from step 4 of the playbook. It takes 1,200 companies that started a trial 12 months ago and splits them by how many weeks they were active in their first 8 weeks.

40 ÷ 500 = 8%. 120 ÷ 400 = 30%. 195 ÷ 300 = 65%. Companies with high early activity are about 8 times as likely to still be paying a year later. That is strong evidence the metric leads revenue. If the three rows had shown similar retention, the candidate would fail the test, however intuitive it sounded.

B2B SaaS vs D2C: what changes

B2B SaaS. The unit is usually the account or team, not the individual user. One power user in a 50-seat account does not mean the account is healthy, so define “active” at account level with a minimum threshold. Sales cycles are long, so validate the revenue link on renewal and expansion data, not first payment.

D2C. The unit is the customer or the delivered order. First orders are often unprofitable after CAC, so the North Star should lean toward repeat behaviour. In markets with high COD use, count orders only when delivered and kept, otherwise RTO inflates the number. Seasonality is strong, so compare the NSM year over year by week, not just week over week.

How to measure it

• Write a metric spec. Name, event, unit, window, filters (exclude internal and test accounts), data source and owner. Store it where everyone can find it.
• Use a fixed time window. “Weekly” means Monday to Sunday in one time zone. Rolling windows are fine but must be consistent.
• Report counts and rates together. NSM count, then each input as a count and a rate. Rates without counts hide scale. Counts without rates hide efficiency.
• Segment by cohort and plan. Break the NSM into new versus existing and by customer segment. Most surprises live in one segment.
• Validate yearly. Re-run the revenue-link test every year. If the link weakens, the product or market has changed.

Templates

North Star metric spec: