Stop 04 · Data & Metrics for PMsTopics · 3/12

North Star Metric Design

In short

Picking a North Star is easy. Picking a good one is hard. This lesson gives you a simple test to tell a strong North Star from a misleading one.

Going beyond the intro from Phase 1 — designing a North Star Metric that's a leading indicator of long-term value, moves with real usage, and can't be gamed by short-term tricks.

The purple blob lifts dumbbells in a busy gym, surrounded by other members working out.

Think of it like this

What should a gym count?

A gym could count memberships sold, but many people pay and never show up. Counting workouts per member each week shows whether people are actually getting fitter, and those people renew.

Watch it happen

Testing candidates for a recipe app

Downloads

Revenue

Recipes cooked/wk

Time in app

1/5Four candidates for the North Star. Test each against three questions.

The big ideas

What you need to know

Captures user value

It should rise only when users get what they came for.

Leads to revenue

When it grows, revenue should follow over time. Otherwise it's measuring the wrong thing.

Breaks into inputs

Split it into a few input metrics so each team knows which lever it pulls.

Has a time window

"Per week" or "per month" keeps it current. All-time totals hide recent drops.

Save yourself the pain

Common mistakes

Oops

Using time spent in app as the North Star for every product.

Do this instead

Time spent fits entertainment. For a payments app, less time can mean a better experience.

Oops

Choosing a metric no team can directly influence.

Do this instead

Make sure it breaks down into inputs that specific teams can work on this quarter.

Your mission

Design a North Star

~25 min

Quick check

Which is the strongest North Star for a job portal?

Go deeper

1Critique 2-3 real, public North Star Metrics for weaknesses (can they be gamed?).
2Design a North Star Metric plus 3 supporting input metrics for a product idea.
3Learn why revenue is usually a bad North Star Metric (it's a lagging, not leading, signal).