Metric trees that product teams actually maintain
A north-star metric with twenty orphans underneath is theater. Build a tree small enough that each leaf has an owner and a weekly ritual.
Metric trees fail when they try to mirror every table in the warehouse. Useful trees start from one outcome the business already debates—activated weekly users, paid conversions, or successful deliveries—and branch only into drivers the team can influence this quarter.
Each node needs a definition sentence, a calculation note, and a named owner. If nobody opens the chart in standup, prune it. We cap most product trees at one north star, three to five input metrics, and a handful of diagnostic guards (latency, error rate, support tickets).
Instrumentation follows the tree, not the other way around. New events must map to a leaf or they wait. That constraint keeps the tracking plan from accumulating vanity taps.
Review the tree when the product strategy changes, not on a fixed annual schedule. A tree that outlives its strategy quietly trains the team to ignore dashboards.