Metric trees

A north star you cannot decompose is a cheerleading slogan. A metric tree breaks it into 3–5 input metrics that teams own. When the star is flat, you diagnose which factor stalled instead of shipping random features.

The big idea

A useful pattern (breadth / depth / frequency / efficiency):

• Breadth — how many customers get value?
• Depth — how much value per customer?
• Frequency — how often do they return?
• Efficiency — how fast or friction-free is the path?

Algebraically, many north stars factor. For a marketplace:

nights_booked ≈ bookers × bookings_per_booker × nights_per_booking

Each factor can be owned by a different squad. Acquisition owns new bookers; host-supply owns listed nights; search ranking owns conversion from search to book.

A kitchen ticket board is a useful picture: the night's covers (north star) split into covers per section, courses per cover and turn time. The analogy breaks when inputs are substitutes — acquisition flooding low-intent signups that crush activation — so the weekly review must look at the whole tree, not one leaf.

Link the tree to how you prioritize and set OKRs: inputs become team objectives; the north star is the company scoreboard.

Walking through it

Worked tree for a B2B notes product. North star: documents collaboratively edited per week (WCED). Last week: WCED = 48,000.

WCED = active_workspaces × docs_edited_per_workspace
     = 6,000 × 8

active_workspaces = new_activated + retained
                  = 900 + 5,100

docs_edited_per_workspace = editors_per_workspace × docs_per_editor
                          = 2.5 × 3.2

Check: 6,000 × 2.5 × 3.2 = 48,000

A product change improves activation: new activated workspaces rise from 900 → 1,200 (+33%), retained stays 5,100. Active workspaces become 6,300. If depth is unchanged:

WCED' = 6,300 × 8 = 50,400
lift  = (50,400 − 48,000) / 48,000 = 5.0%

Same 5% lift via depth: need docs_edited_per_workspace to rise from 8 → 8.4 while breadth stays 6,000. That might mean shipping templates that add 0.4 docs per workspace — a different roadmap than pouring money into ads for more signups. The tree tells you which lever is cheaper.

Common mistakes

• A star with no owners on the inputs. Nobody knows who moves breadth vs depth.
• Local optima. Each squad optimizes its leaf while the star stalls because inputs are substitutes.
• Under-instrumented ambition. Declaring "weekly active teams that share a board" without events for share or team membership.

Check yourself

1. WCED is 48,000 from 6,000 workspaces × 8 docs. Activation adds 300 workspaces. What is the new WCED if depth is unchanged?

6,300 × 8 = 50,400, a 5% lift.

2. You need the same 5% via depth instead. What must docs per workspace become?

8.4, because 6,000 × 8.4 = 50,400.

Key takeaways

• Decompose the north star into breadth, depth, frequency and efficiency inputs with owners.
• Use the identity to size which lever is cheaper before picking a roadmap.
• Review the tree, not one leaf — inputs can fight each other.

Further learning

The North Star Playbook (Amplitude)Put the North Star Framework to work for your product. Get expert tips on defining and integrating your North Star Metric into your development processes.Amplitude

Next: protect the star with guardrail metrics.