Course contentsShow
← Data Science
Lesson 1623 of 2,14539. Business Metrics and KPI DesignFree lesson

Depth vs Breadth in Metric Trees

Deciding how many levels deep and how many branches wide your metric tree should extend.

Depth vs Breadth in Metric Trees

You'll learn how to decide whether your metric tree should go deep with many levels or stay wide with many branches at each level.

The Core Tradeoff

When building a metric tree from your North Star, you face a structural decision: should you drill down deeply into fewer dimensions (depth), or should you break your metrics into many parallel categories at the same level (breadth)?

Depth means creating many hierarchical levels. For example: Revenue → Product Revenue → Premium Product Revenue → Premium Product Revenue from Enterprise Customers → Premium Product Revenue from New Enterprise Customers. This gives you fine-grained specificity but can become overwhelming to track.

Breadth means splitting into many parallel branches at fewer levels. For example: Revenue → [Product Revenue, Services Revenue, Advertising Revenue, Partner Revenue]. This captures more dimensions simultaneously but may lack the granular insight needed for action.

Finding the Right Balance

Think of your metric tree like a family tree: too deep and you're tracking distant cousins nobody remembers; too wide and you can't see meaningful relationships.

Guidelines:

  • Go deeper when you need to diagnose problems in a specific area (e.g., why did mobile conversions drop?)
  • Go wider when you need comprehensive coverage (e.g., tracking all revenue streams for a CFO)
  • Practical limit: Most effective trees are 3-5 levels deep with 3-7 branches per node
  • Match depth to your team's bandwidth for metric ownership—every leaf needs an owner

The best trees balance both: wide enough to capture all critical drivers, deep enough in specific areas where actionability demands detail.

Key Takeaway: Design your metric tree's depth and breadth to match your organization's decision-making needs and monitoring capacity—not every branch needs equal depth, and not every level needs equal breadth.