What Makes a Good North Star Metric
Not every metric qualifies as a north star. Several characteristics separate a genuine north star from a vanity metric or a proxy. A good north star metric captures value delivery to users, not just activity. It reflects that users are getting the benefit your product promises, not just that they clicked around. A north star metric is a leading indicator of revenue, not lagging: it predicts future commercial success rather than measuring past revenue. It is influenceable by the product team: the team's decisions should causally affect the metric, not merely correlate with it. It is a single number, not a composite. And it is long-term stable: a metric that changes every quarter is not a north star, it is a priority of the month. Classic examples: Airbnb's north star was nights booked, because it captured both host and guest value simultaneously. Spotify's was time spent listening. Slack's was number of messages sent within an organisation. Each of these metrics is a direct proxy for the value the product delivers.
North Star Metrics for AI Products
AI products require north star metrics that capture AI-delivered value, not just product activity. An AI document analysis product might use 'number of documents with at least one insight acted on' rather than 'documents processed', because processing without value delivery is not the goal. An AI customer support tool might use 'issues resolved without human escalation' rather than 'conversations started', because deflection is the value, not engagement. An AI writing assistant might use 'documents completed and exported' rather than 'suggestions generated', because suggestions only matter if they are accepted and used. Choosing a metric that captures actual AI value delivery forces honest reckoning with whether the AI is working. A product where users regularly start AI interactions and abandon them midway is not delivering the north star metric even if raw session counts look healthy.
The Metric Tree: North Star and Supporting Metrics
The north star metric sits at the top of a metric tree. Below it are the input metrics that drive it: acquisition (how many new users reach the product), activation (how many reach the key value moment), engagement (how often users perform the core value action), retention (how many users return over time), and monetisation (how many convert to paid and at what average revenue). These input metrics are the levers the team can pull to improve the north star. A weekly north star review identifies which input metric is lagging and focuses the next sprint on improving it. A north star that is declining despite increasing acquisition tells you the activation or retention problem is the bottleneck. This tree structure gives the north star metric its operational value: it is not just a number to watch, it is the root of a diagnostic system.
Common Mistakes in North Star Metric Selection
Several selection errors are common. Choosing revenue as the north star conflates the north star with the business outcome: revenue is what the north star predicts, not the north star itself. Early-stage products with few users have low revenue by definition, so revenue as a north star produces a meaningless signal in early development. Choosing a metric that captures supply rather than demand is another error: 'features shipped' or 'AI responses generated' tells you about activity, not value. Choosing an overly granular metric that the team can manipulate without genuinely improving user value defeats the purpose. The north star must be chosen with the question: if this metric is consistently high, do we believe the business will succeed? If yes, it is probably the right metric.
Using the North Star Metric in Product Development
The north star metric should appear in every sprint review and every product planning session. Feature prioritisation should be argued in terms of predicted impact on the north star: 'this feature will increase activation rate by X percent, which we predict will move the north star by Y percent'. A/B test results should be evaluated against the north star, not just the local metric the test was optimising. Roadmap sequencing should prioritise the layer of the metric tree with the highest leverage: if retention is the bottleneck, retention-improving features rank above new acquisition features even if acquisition features are more exciting. Not every sprint can improve the north star directly, and some north stars have long feedback loops. The discipline is in the periodic review: quarterly, at minimum, ask whether the trajectory of the north star matches the level of team activity.
North Star Metrics at SpeedMVPs
During discovery for every AI MVP engagement, SpeedMVPs works with clients to define a north star metric and the measurement infrastructure to track it. Defining the success metric before building ensures the MVP is scoped to generate signal on that metric, not just to demonstrate features. Analytics instrumentation included in every delivery captures the input metrics that compose the metric tree. For clients in their second or third MVP engagement, north star metric analysis from the previous cycle informs which part of the tree to invest in next. GDPR-compliant analytics design ensures that metric tracking is privacy-preserving by design, collecting only the data necessary for the defined measurement purpose. Projects from GBP 8,000 with 2-3 week delivery.