How would you choose a north-star metric for a product, and what makes a metric a good north-star?
A north-star metric must satisfy three properties: it reflects the core value delivered to users, it correlates with long-term business outcomes (retention and revenue), and it is actionable — meaning teams can run experiments that move it. Choosing one requires articulating the product's value exchange and then stress-testing candidate metrics against those three criteria.
How to think about it
Three criteria for a north-star metric
| Criterion | What it means | Red flag |
|---|---|---|
| User-value proxy | The metric rises when users get real value | ”Sessions started” rises when the app crashes and users retry |
| Business correlation | Historically predicts retention or revenue | Raw pageviews uncorrelated with subscription renewal |
| Actionable | Teams can design experiments that directly move it | ”Brand awareness” is too diffuse to test |
Framework
Step 1 — Define the value exchange. Ask: what does the product do for users? Airbnb: helps travellers find accommodation; helps hosts earn income. The value exchange is completed trips.
Step 2 — Enumerate candidate metrics. List 4-6 candidates. For Airbnb: searches, saved listings, booking requests, completed bookings, repeat bookings, host occupancy rate.
Step 3 — Apply the three criteria. Completed bookings passes all three. Searches fail criterion 1 (many searches complete no booking). Repeat bookings are too lagging to run fast experiments against.
Step 4 — Add a time dimension. A good north-star is measured at a cadence that matches the product’s natural frequency. Airbnb: monthly or quarterly (leisure travel is low-frequency). Slack: daily active users sending a message (work tool is high-frequency).
Worked example — Spotify. Value exchange: music discovery and listening. Candidates: streams, listening hours, playlist saves, paid conversions. North-star: monthly listening hours per active user. It reflects value (you only listen if you like the experience), correlates with subscription conversion, and is directly movable by recommendation experiments. Guardrail: ensure it is not gamed by autoplaying ambient/sleep content at low volume.