Every product has a dashboard. Almost every dashboard has a chart going up. And a lot of teams find out, too late, that the chart going up had nothing to do with the business succeeding.
The problem is rarely a lack of data. It's choosing the wrong metric to look at.
What a vanity metric is
A vanity metric is one that goes up easily and doesn't change any decision. The classic examples:
- Total sign-ups: it only goes up, never down, and doesn't say whether anyone used the product;
- Pageviews: you can inflate them with any campaign, including one that brings the wrong audience;
- App downloads: they measure the ad, not the product.
The test is simple: if this metric doubles tomorrow, what would you do differently? If the answer is "celebrate", it's vanity.
What makes a good North Star
The North Star is the metric that best captures the value the customer receives and that, if it grows sustainably, pulls revenue along with it. It has a few characteristics:
- It reflects value delivered, not effort or reach. "Reports generated and opened by the team" says more than "logins".
- It's a recurring behavior. Value that happens once doesn't sustain a subscription business.
- The team can influence it. If no product decision moves it, it can't guide the product.
- It comes before revenue. When it goes up, revenue goes up afterwards, not the other way around.
The metric tree
The North Star alone is too far from day-to-day work. Nobody wakes up and "moves the North Star". That's why it breaks down into a tree of input metrics, which are the levers the team actually pulls.
A common breakdown:
| Level | Question | Example metric |
|---|---|---|
| North Star | How much value are we delivering? | Accounts with a shared report in the week |
| Breadth | How many reach the value? | Accounts that connected the data source |
| Depth | How well do they use it? | Reports shared per account |
| Frequency | How regularly do they come back? | Active weeks in the month |
| Efficiency | How fast do they get there? | Time from sign-up to first report |
Each team or initiative hangs off a branch. The onboarding team looks after "time to first report". The engagement team, frequency. Everyone knows how their work connects to the top.
How to choose in practice
A plan that works well in one or two sessions with the decision makers:
- Write the value in one sentence. "Our product helps X achieve Y." If it doesn't come out, the problem is positioning, not metrics.
- List the behaviors that prove Y happened. Not what the user clicked, but what they achieved.
- Filter by the criteria above: recurring, influenceable, comes before revenue.
- Validate with historical data. Do retained customers show more of this behavior? If not, the hypothesis is wrong.
- Break it down into the tree and give each branch an owner.
- Review it every quarter. The North Star of a product with five customers isn't the same as one with five hundred.
Common traps
- Choosing revenue as the North Star. Revenue is an outcome, not a lever. It's the metric the North Star should predict.
- Having five North Stars. If everything is a priority, nothing is. One metric at the top, several below.
- Measuring only what's easy. Sometimes the right metric requires instrumenting a new event. It's worth the work.
- Forgetting guardrail metrics. Growing the North Star at the cost of cancellations or exploding support isn't growth. Define limits that can't get worse.
Wrapping up
The North Star is less a metric and more an agreement: about what value is, about what the team pursues and about what stays out. Chosen well, it turns opinion debates ("I think this feature is important") into hypothesis debates ("I think this feature increases frequency, and we'll know in two weeks").
Murilo Anzini
Hands-on writing on engineering, product, data and marketing for people who want to understand, and build, the whole journey of a product.
Modular monolith first: how I structure a SaaS from scratch without regretting it later
Microservices solve problems a new product doesn't have yet. The conventions I use to start with a well-split monolith that can grow and doesn't turn into a tangle.
A measurement plan before the first event: clean GA4, with consent
Installing GA4 takes five minutes; trusting the data takes months if you start wrong. How I design the event taxonomy, consent and measurement flow together with the feature.