The name contains its own refutation, and almost nobody using it notices.

Polaris is chosen as a navigational reference because it is very nearly fixed — under a degree from the celestial pole. It does not respond to being looked at. Sailors have steered by it since the early medieval period — once precession brought it near enough to the pole — precisely because nothing a sailor does can move it.

A metric is the opposite kind of object. It is produced by human behaviour, and the behaviour changes when the metric becomes important. Which is Goodhart's Law, stated in 1975: any observed statistical regularity tends to collapse once pressure is placed on it for control purposes.

So the North Star Metric is, structurally, the practice Goodhart's Law warns about, and it is usually adopted by people who also quote Goodhart's Law approvingly.

Both can be right, briefly, and the sequence is predictable. A well-chosen metric genuinely aligns an organisation — it stops six teams optimising six things that quietly cancel out, and that alignment is worth a great deal. Then, over time, it becomes the thing performance is judged by, at which point people optimise the number rather than the thing the number was standing in for. The metric is now measuring compliance with the metric.

The standard guidance is better than its reputation, and worth stating. Do not use revenue. Pick a metric reflecting core value delivered to users; make it actionable; make it not a ratio — because ratios can be improved by shrinking the denominator, which is to say by losing customers; and make it capable of scaling over time.

One much-cited historical example: the number of users who added seven friends in their first ten days. Oddly specific, and that specificity is the point — it is a measurable proxy for the moment a product becomes hard to leave.

The honest practice is therefore to treat a North Star as a fixed point with an expiry date, and to change it before the gaming does it for you.