Metric Tree

A metric tree decomposes a headline number into the inputs that arithmetically produce it, and then decomposes those, until you reach quantities a team can actually influence. Revenue breaks into customers times average price; customers break into new plus retained; new breaks into traffic times signup rate times paid conversion. The point is not the diagram. It is that the relationships are arithmetic rather than rhetorical, so a movement in the top number has to be attributable to a movement somewhere below it, and finding where is a mechanical exercise instead of a debate. That property is what makes a tree useful in the two situations where teams usually argue. When a number moves unexpectedly, walking down the branches finds the cause in a few steps and rules out the branches that did not move, which is often more valuable than the branch that did. And when planning, a tree makes the size of a proposal legible: a change that improves a step near the bottom of a long multiplication may be a great deal of work for very little at the top, and the tree shows that before the quarter is spent rather than after. Building one has two failure modes worth avoiding. The first is a tree that does not actually multiply out — if the children do not reconcile to the parent, it is an organisational chart of concepts and it will not survive its first argument. The second is decomposing until every leaf is something no team owns, which produces a beautiful structure nobody uses. Stop where a named team can move the number, and accept a coarser tree that is used over a complete one that is admired. A tree is also a useful way to see which inputs are shared: two teams optimising different branches that multiply into the same parent can each report a win while the parent stays where it was, and the structure makes that visible instead of leaving it to be discovered at a review.