Why one number, and what it actually buys you
The case for a single guiding metric is not that other measures are worthless. It is that attention is the scarcest resource on a product team, and a metrics dashboard is a claim on that attention. When a team is told to move ten numbers, it does not move ten numbers a little; it moves whichever number is easiest to move this sprint, and rationalizes the rest. The single number — the one the team is accountable for over a quarter or a year — exists to resolve the dozens of small trade-offs that never reach a planning meeting. Should the onboarding flow add a step that improves data quality but costs activation? Without a top metric, that decision is made by whoever argues hardest. With one, it is made by arithmetic.
The mechanism is forcing function, not measurement. A well-chosen north star compresses strategy into a quantity that a designer can hold in mind while choosing between two button placements. That compression is the entire value, and it is also the entire danger: whatever the number omits, the organization will eventually omit too.
The hierarchy: vanity, proxy, and the thing itself
Most metrics fail because they sit at the wrong altitude. It helps to sort every candidate into three tiers. Vanity metrics — registered users, total downloads, cumulative revenue — only rise, so they confirm motion without proving progress; they are reporting artifacts, not steering instruments. Proxy metrics — weekly active users, feature adoption, session length — move in both directions and correlate with value, but they are stand-ins, and a stand-in can be satisfied without the real thing being delivered. The number that matters sits above both: it measures realized customer value in a way that the business can only win if the customer wins first.
- Ask of any candidate: can it go down? If not, it is vanity.
- Ask: if this number doubled and the customer got nothing, would anyone notice? If the answer is no, it is a proxy you are about to over-trust.
- Ask: does the company capture value only after the customer does? If yes, you have found the thing itself.
The failure mode is the proxy hardening into the goal
The predictable way this goes wrong is Goodhart's law operating on the proxy you chose for convenience. Engagement is the classic trap. A team picks time-in-app because it is easy to instrument and plausibly tied to value, then spends two years optimizing it — adding notifications, autoplay, infinite feeds — and discovers it has built a product that is used compulsively and valued lightly, with churn that the engagement number never saw coming. The proxy did not lie; it was simply asked to carry meaning it could not hold. Every proxy has a gap between what it counts and what you care about, and sustained optimization pressure will find that gap and drive a truck through it.
This is why the top metric should be defined in terms of the customer's outcome, not the product's activity. "Nights booked" survives optimization pressure in a way that "searches performed" does not, because a booked night is value the customer chose to pay for. The closer the number sits to money the customer willingly parts with for a result they wanted, the harder it is to game without actually delivering.
The decision implication: subordinate, instrument, and revisit
Choosing the one number is not the end of measurement; it is the reorganization of it. The discipline is subordination: proxies and vanity metrics still get tracked, but as diagnostics that explain movement in the top number, never as goals that compete with it. When activation rises and the north star does not, that divergence is the most valuable signal the dashboard produces, and a flat hierarchy destroys it by treating both as equal wins.
Two cautions close the argument. First, a single metric is correct for a single strategy; when the strategy changes — land-grab to monetization, breadth to retention — the number must change with it, and clinging to last year's north star is how teams optimize their way into the wrong business. Second, no scalar fully captures a product, so the top metric must be governed by a small set of guardrails — a quality floor, a churn ceiling — that hold the line on what the number ignores. The one number that matters is a steering instrument, not the territory; the job of leadership is to choose it deliberately, defend its primacy against the easier numbers, and retire it the moment it stops describing the value you actually intend to create.