North Star Decomposition

Picking a north star metric is usually the easy part. It's the single number a company hangs up so everyone is pulling in one direction — weekly active editors, completed orders, nights booked — and most teams land on one without much trouble.

The wall comes right after. The number is on the wall and nobody knows what to do tomorrow that will move it. A headline metric is an outcome that takes a quarter to respond, so it never connects directly to the work anyone did this week.

Pagedeck's quarterly planning meeting looked exactly like that. Grace proposed taking weekly active editors from 11,000 to 15,000. Ravi wanted to spend more on ads and bring in more teams; Mia wanted to rework what new members see right after they're invited. Both sounded reasonable, and there was no way to compare which one got closer to 4,000 additional editors. The meeting went to whoever argued hardest, and when the number stalled at 12,000 three months later, nobody could say what had been missing.

Decomposition is the ladder between the two. Break the headline number into factors that multiply back into it and "let's raise it" splits into specific, separate choices about what to raise. Each factor carries its current value, which also makes it visible which one is cheapest to move.

Write the headline metric on one line and work through five steps.

1. Pin the definition down to something countable. Weekly active editors means people who edited at least one document in the last seven days. Leave this vague and every number underneath it wobbles.
2. Split it one level, using multiplication. At Pagedeck the number reads as active teams × members per team × the share of those members who edit. The factors have to multiply back to the original value. If they don't, what you have is a list of loosely related metrics rather than a decomposition.
3. Fill in today's values. 3,200 teams × 9.4 members × 38 percent lands right around 11,000. The moment real numbers go in, the unusually low factor stands out.
4. Split one more level where it helps. Active teams breaks down again into teams retained from last month plus new teams plus returning teams. Go past two levels and you have twenty boxes nobody reads, so most teams stop here.
5. Write down what a 10 percent gain in each factor costs and how long it takes. Growing team count by 10 percent means ad spend; lifting the edit share by 10 percent means reworking a screen. The two often come out an order of magnitude apart, and that table is what decides the next quarter's plan.

Whichever factor you pick has to be one the team can act on directly. A value the team's own work can move is called an input metric, and pushing the headline number down to input metrics is the whole point of the exercise.

It fits when the headline metric is already chosen and user behavior is being recorded. Quarterly planning across several teams is the typical setting, because each factor can be handed to an owner — acquisition to marketing, edit share to the product team.

There are three limits.

1. It narrows down where to work, not what to do. Establishing that the edit share is low is the end of what decomposition gives you; raising that share is a question for experiments. Work like finding a magic number — how many times someone has to do something before they stick around — comes after this step, not during it.
2. The factors are not independent. Push ad spend hard enough to grow team count 30 percent and poorly matched teams come in with it, dragging the edit share down. The product of the factors can end up exactly where it started, which is why a gain in one box is worth checking against the others before anyone calls it a win.
3. Without tracking you can't begin. In a product that never records who edited a document, the edit-share box simply cannot be filled, and there is no way to reconstruct last quarter's data after the fact.

"Break it down and there's only one place worth touching this quarter: edit share."

Values went into every factor and the cheapest place to move narrowed to one.

"These factors don't multiply back to the original number. What's missing?"

The breakdown doesn't hold as an identity, so a factor has been left out.

"Marketing takes active teams, product takes edit share."

A proposal to assign the factors to owners and make them quarterly goals.

Pagedeck started the next planning meeting with the breakdown instead. Once Mia filled in the three factors, two things were visible: members per team had sat at 9.4 for as long as anyone had measured it, and the 38 percent edit share was noticeably below what teams coming from competing products showed.

Edit share came out as the cheapest of the three, and Ravi's ad proposal moved to the following quarter. The product team changed one screen — the first thing an invited member sees on arrival. By the end of the quarter edit share had gone from 38 to 45 percent, and weekly active editors reached 13,000 with team count barely changed. That fell short of the 15,000 target, but this time the record showed which factor moved and by how much. The quarter ended with an answer rather than an argument, and that answer became the starting point for the next plan.

The most common mistake is turning every factor into a goal. Six boxes came out of the exercise, so raising all six looks like the obvious plan. But the value of a decomposition is in deciding what not to touch. Set six targets and each team nudges its own box a little, no box moves enough to notice, and at the end of the quarter there's no way to tell what worked.

Drawing the tree once and leaving it is the other frequent error. The structure rarely changes, so a document seems like enough. In practice, the bottleneck moves as the values move. Once edit share is up at 45 percent, the cheapest factor shifts over to team count. Teams that never re-enter the numbers repeat last quarter's winning move and can't work out why the same effort produced less.