Metric Tree

What is a metric tree?

A metric tree is a hierarchical decomposition of a top-level outcome metric into the sub-metrics that drive it, recursively down to the operational levers the team can directly act on. The root might be quarterly revenue; one level down, ARR and NRR; another level down, new ARR and expansion ARR; another, pipeline volume and win rate; another, MQLs and conversion rates.

The tree makes causation visible. When revenue moves, the operator can trace the cause through the structure — was it new business or expansion? Was new business pipeline or win rate? Was pipeline volume or quality? Without a tree, the question is answered through investigation; with one, it's answered by reading the tree.

Why metric trees matter

Companies without a metric tree report numbers without structure: "Revenue is up 8% this quarter" with no decomposition. Companies with a metric tree report: "Revenue is up 8%, driven by 12% growth in new ARR (pipeline up 6%, win rate up 5%) offset by 3-point compression in expansion (driven by 4% NRR drop in the SMB segment)". The second report is actionable; the first is a fact.

For weekly business reviews and operating reviews, the metric tree is the agenda backbone. Each branch maps to an owner; each leaf maps to a measurable lever. Variance discussions follow the tree rather than wandering.

Sample metric tree: SaaS revenue

Revenue (root)
├── ARR
│   ├── New ARR
│   │   ├── Pipeline ($)
│   │   │   ├── MQLs (×)
│   │   │   ├── MQL-to-SAL %
│   │   │   ├── SAL-to-Opp %
│   │   │   └── Avg deal size
│   │   └── Win rate %
│   ├── Expansion ARR
│   │   ├── Seat expansion %
│   │   └── Module attach rate
│   └── Churn ARR (subtract)
│       ├── Logo churn %
│       └── Avg lost ACV
└── Services revenue
    ├── Onboarding fees
    └── Professional services

Each leaf node has an owner.
Each owner reports variance in WBR.

How to build a metric tree

• Choose the root. The single top-level outcome that aligns the company. Usually revenue or ARR, sometimes contribution margin or net income.
• Decompose mathematically. Each level must be a mathematical identity — child nodes sum to parent (or multiply, or are weighted components).
• Stop at operational levers. Leaf nodes should be things a team can directly affect — not "customer happiness" but "support response time" or "feature shipped this quarter".
• Assign ownership. Every node has one owner. Every leaf has one team accountable for moving it.
• Use in operating cadence. WBR and operating reviews use the tree as agenda. Variance discussions follow tree structure.

Common mistakes

• Non-mathematical decomposition. If child nodes don't sum/multiply to parent, the tree doesn't explain variance — it's just a hierarchy of related concepts.
• Too many leaves. A tree with 200 leaves is unusable. Aim for 30-60 leaf nodes in the top three levels.
• Unowned nodes. Every node, especially leaves, needs an owner. Unowned nodes mean unaccountable variance.
• Static tree. The tree should evolve as the business does — new product lines, new channels, new segments. Review structure annually.

Related concepts

Metric tree is the structural artifact; north star tree is a specific variant placing the north star metric at the root. Used in WBR and operating review cadence. Supports single source of truth discipline by making metric definitions explicit.

Frequently asked questions

Sources

1. Sean Ellis, Morgan Brown. Hacking Growth, Crown Business, 2017.
2. Andy Grove. High Output Management, Vintage, 1983.
3. Bain & Company. Outcome-Driven Metric Trees in B2B SaaS, 2024. bain.com