North Star Metric (definition)

A north star metric is the one number that best represents the value your product delivers to users, not the value the business extracts from them. It aligns a team on a shared definition of success, leads revenue over time, and cannot be gamed without genuinely helping users. Singular by definition: if you are naming three metrics, you do not have a north star, you have a dashboard.

What makes a north star good

Four criteria, all required:

• Reflects real user value. Users got something done, not just arrived. “Sessions started” fails; “problems solved” passes.
• Leads revenue. It moves before the business outcome does, which is what makes it useful for steering. Revenue itself is a lagging outcome, not a north star.
• Can’t be gamed without helping users. If your team can move the number without improving user outcomes, the metric will be optimized in the wrong direction.
• The team can move it. A metric that only shifts with macroeconomic conditions or competitor launches is a weather report, not a north star.

The three game types (a fast interview heuristic)

Amplitude’s research across hundreds of product teams identified three business-model archetypes. Each implies a different NSM shape, and naming the right one in an interview is what lets you justify your metric instead of just asserting it.

• Attention game (Facebook, Netflix, TikTok): depth and frequency of engagement. Facebook’s original NSM was “users who add 7 friends in the first 10 days,” not DAU. DAU came later and is a proxy.
• Transaction game (Airbnb, Amazon, DoorDash): completed exchanges. Airbnb’s NSM is “nights booked,” which captures host income, guest satisfaction, and company revenue in a single number.
• Productivity game (Salesforce, Notion, Figma): tasks completed or outcomes reached. “Projects published” or “workflows automated” over counts of logins.

Identifying the game type first gives you a principled reason to propose the NSM you do. That reasoning is what earns credit, not the metric name alone.

Good vs. vanity

Page views measure arrival, not value. DAU and MAU are proxy metrics: they lag real value and can move in the wrong direction (a bug that traps users in a loop raises DAU). A strong answer names the distinction explicitly and proposes a metric that survives the “can we game this without helping users?” test.

Guardrail metrics: the follow-up you should pre-empt

A north star optimized in isolation will be gamed. Every NSM needs at least one guardrail: a metric that caps acceptable tradeoffs. Airbnb’s “nights booked” needs average guest rating or refund rate alongside it; otherwise the team optimizes for volume at the cost of satisfaction. Naming the guardrail before the interviewer asks is one of the clearest senior signals in an analytical round.

NSM vs. OMTM vs. OKR key result

These terms are frequently conflated. OMTM (one metric that matters) is a startup framing from Lean Analytics, typically used before product-market fit when the team is still searching for the value to sustain. NSM implies you have found that value and are scaling it. An OKR key result is a time-bound target set against a metric (“reach 10M nights booked by Q4”), not the metric itself. Conflating any of these in an interview signals shallow preparation.

AI products in 2026: why the classic criteria break

Feasibility is effectively free in 2026: an AI feature can be shipped in days. That makes “what does good look like?” the hardest PM question, which is precisely what the north star metric answers. But for AI products, the classic criteria fail in a specific way: ambient AI features inflate session count and DAU without delivering value. A chat feature that sits open in a sidebar all day looks like high engagement even when users are ignoring the output.

The NSM for an AI product must be outcome-based, not usage-based. Candidates who earn senior signal name this trap and propose alternatives: “tasks completed without human correction,” “problems solved per active user per week,” or “outcomes achieved vs. a control baseline.” All of these require experimental comparison, not point-in-time readings. Data leader Eric Weber has written directly on this: the metric you pick today will probably be wrong in six months because the AI capability will have shifted. That means AI product NSMs need a defined review cadence built in from the start, not treated as a one-time decision.

PMs who can articulate the ambient-adoption trap, propose an outcome-based metric, and name the guardrail against gaming are the ones clearing the bar at Anthropic, OpenAI, and Notion in 2026.

When the north star frame is the wrong tool

Two cases where a single NSM misleads more than it guides. First: early-stage products before PMF. You are still validating what “value delivered” means, so OMTM is the right frame. Second: platform products with two-sided value (marketplaces, developer platforms). Optimizing one side’s north star frequently damages the other. Marketplaces track a composite like GMV or nights booked precisely because it captures both sides simultaneously.

Interview: weak vs. strong answers

For how to define a north star on the spot in an interview, see the North Star framework. For related metrics questions, see DAU/MAU, retention, and OKRs.