North Star Metrics: What Actually Drives Growth

Most organisations have a North Star metric.

Few actually use it.

The problem

North Star metrics sound great in theory. One clear number that:

• guides decisions
• aligns the organisation
• drives growth

In practice, it often becomes:

• a KPI in a presentation
• something people reference
• but don't work with

What a North Star actually is

A North Star metric isn't a number. It's a representation of value.

For the customer. For the business.

If it doesn't reflect both → it doesn't matter.

Common mistakes

1. Too abstract

"Engagement." "Activation." → Says nothing about revenue or impact.

2. Too far from reality

If it takes months to influence → you can't work with it.

3. Not connected to experiments

The North Star exists. But tests are tied to CTR, conversion, bounce rate. → Disconnect.

4. No decomposition

The North Star lacks drivers, components, structure.

From North Star to system

Example:

North Star: Revenue per user

Decomposes to: Traffic × Conversion rate × Average order value

And further:

• Traffic → channel, quality
• Conversion → UX, messaging, friction
• AOV → bundling, pricing, offers

Now you can work. That's growth systems thinking in practice.

Formulas aren't maths — they're prioritisation

The formula above isn't just a calculation. It tells you: where to focus.

Examples:

• Low conversion → CRO
• Low AOV → offers
• Wrong traffic → performance marketing

Connection to experimentation

Every test should answer: Which part of the North Star does this affect?

Examples:

• New product copy → conversion
• New pricing → AOV
• New channel → traffic

If the answer is unclear → don't test.

That requires a clear measurement plan connecting each experiment to the right part of the formula.

What happens when it works

• fewer debates
• clearer priorities
• faster iteration

The organisation moves in the same direction.

What happens when it doesn't

• optimising the wrong things
• short-term wins
• stagnation

Conclusion

North Star metrics work. But only when they:

• are decomposed
• are connected to work
• are used in experiments

Otherwise, they're just a number.