Metric Trees

Revenue fell 8%. A metric tree turns that sentence into four candidate causes in ninety seconds.

Metrics & KPIs · Intermediate · ~5 min

In plain English

Revenue is customers × how often they buy × how much they spend. When revenue drops, you check three dials instead of guessing.

Why it's worth your time

It turns 'why did the number move' from an afternoon of ad-hoc queries into a ninety-second walk down a branch.

If you remember three things

Each level must arithmetically reconstruct its parent
Branches should be as independent as possible to be diagnostic
Log decomposition attributes a multiplicative change cleanly

Overview

A metric tree decomposes a headline number into factors that multiply or add to it: revenue equals users times conversion times average order value; users equals new plus retained plus resurrected minus churned. The value is diagnostic — when the top moves, you walk down the branches and find which factor actually changed, instead of guessing or running twenty ad-hoc queries. It is also organisational: each branch can be owned by a different team, and everyone can see how their local metric contributes to the number the company cares about.

In an interview

A metric tree decomposes a top-line metric into multiplicative or additive drivers — revenue = users × conversion × AOV — so a movement at the top can be traced to the branch that moved. It makes diagnosis systematic instead of exploratory, and it gives each team a metric that provably rolls up to the company one.

Production defaults

Verification: prove the children reproduce the parent on historical data
Ownership: one owner per branch, each able to actually move their metric
After diagnosis: segment the identified branch — flat factors with shifting mixes are common

What breaks

Factors don't explain the parent's move — The decomposition isn't exact, or a mix shift is hiding inside a flat factor.
Every branch moves together — They're not independent. Re-cut the tree so each branch is separately actionable.

Watch it explained

What are KPIs and Metrics? | Data Fundamental for Beginners — Alex The Analyst, 5:46