Metric Tree: How to Build One and Use It to Explain Changes

A metric tree, also called a KPI tree or driver tree, breaks one top-level metric into the smaller metrics that produce it. Revenue becomes customers times revenue per customer; customers become new plus retained; and so on. The tree shows which lever moved when the top number changes. This guide shows how to build one, a sample with numbers and how to use it.

What a metric tree is

A metric tree is a diagram that breaks one top metric, the root, into the metrics that drive it. Each level is linked to the one above by a formula, so the numbers add or multiply up.

It is often built under a north star metric. The north star says what matters; the tree says what produces it and who can move each part.

How to build one

The table walks through five steps. Two rules matter most:

• Use equations, not associations. "Active users" and "page speed" are both related to revenue, but only an equation like revenue = customers × revenue per customer tells you how much each part contributes.
• Stop when a leaf is actionable. A leaf should have an owner and something they can change, such as trial-to-paid rate or onboarding completion.

If the branches do not add or multiply up to the root, the tree is a list, not a tree.

A sample tree

For a subscription product:

• MRR = paying accounts × average revenue per account
  • Paying accounts = last month's accounts − churned accounts + new paid accounts
  • New paid accounts = trials × trial-to-paid rate
  • Churned accounts = last month's accounts × churn rate
  • Average revenue per account = base plan price + expansion revenue per account

Using the tree to explain a change

MRR fell by $2,800 this month. Walking down the tree:

1. Revenue per account is flat at $70, so the change sits in paying accounts, down 40.
2. Churned accounts were 190, in line with recent months.
3. New paid accounts were 150, down from about 190. Trials were steady, but the trial-to-paid rate fell from 15% to 12%.

The question is no longer "why did revenue fall" but "why did fewer trials convert this month", which one team owns and can investigate.

Keeping the tree in your analytics

A tree on a slide goes stale. Keep it where the data lives:

• Save each leaf as a metric with its own definition and filters.
• Build higher levels as formulas over saved metrics. In TapCub a custom metric can be an event aggregation or a formula, and a formula can reference saved metrics that are not formulas themselves. So keep leaves as plain metrics and combine them one level up.
• Put the tree on a dashboard, one card per level, with a text block that writes out the equation.

An analysis holds up to 12 measurements, which is usually enough for the top two levels of a tree side by side.

Common mistakes

• Branches that overlap, so the parts add up to more than the whole.
• A tree so deep that leaves have no owner.
• Changing a definition at one level without updating the levels above.

FAQ

What is the difference between a metric tree and a KPI dashboard?

A dashboard shows metrics side by side. A tree shows how they are connected by formulas, so you can trace a change in one to the others.

How deep should a metric tree be?

Usually three or four levels. Stop when each leaf has an owner and a clear action.

Can a metric tree have non-mathematical links?

It is better if every link is an equation. Associations without an equation belong in notes beside the tree, not in it.

Related: north star metric and what is a metric.

See also: Product analytics framework.