learn-metric-decomposition-with-phoebe / Leader session 1 of 6
Learn Metric Decomposition with Phoebe · Leader track · Session 1 of 6

Why a number needs a tree

"Revenue is down 8%." Every leader has heard that sentence and felt the room go quiet. The problem is that it names an outcome, not a cause - so the meeting turns into opinions, not diagnosis. A driver tree breaks a top-line metric into the levers that actually move it, so a drop becomes a routed question with an owner instead of a mystery you argue about. This session teaches the mindset, not the maths: never stare at a number, always break it into the levers underneath.

🔵 Leader track Leaders: CxO · VPs · founders No code - thinking mode 45 min Start here
0-3 · Welcome 3-20 · A number is an outcome 20-42 · Decompose to act 42-45 · Q&A
Part 0

How this track works

Six leader sessions, one habit: read any top-line business metric as a tree of drivers, and use that tree to diagnose a drop in minutes instead of meetings. Today you learn why a number alone can never be acted on and how decomposition assigns a drop to a lever and a lever to an owner. Then a2 separates leading from lagging indicators, a3 walks the diagnosis of a real drop, a4 lines up six domains side by side, and a5 and a6 run the weekly and monthly metric reviews. The builder track (b1-b10) does the same thinking hands-on in a live simulator - you do not need it to lead these conversations well.

Live - presented in session Self-study - read after class Framework sources covered
★ What you walk out with today A durable mental model: a top-line number is a lagging outcome, and behind it sits a small tree of levers, each owned by a team. You leave able to reframe "the number is down" into "which driver moved, and whose is it?" - the single question that turns a blame meeting into a diagnosis.
Part 1 · covers Lean Analytics, DuPont analysis

A top-line metric is a lagging outcome 8 min live

A number like GMV, revenue, or ARR tells you what happened and never why. It is the score at the end of the game, not the play that decided it. Two companies can both report "down 8%" for completely opposite reasons - one bled traffic while conversion held, the other kept traffic but conversion collapsed. Same symptom, different disease, different fix. The driver-tree mindset is simple: never stare at the score, always break it into the levers that produced it.

GMV (the outcome) $208K / mo · lagging Traffic 120,000 visits Conversion 2.8% Avg order value $62 x x owner: Marketing owner: Product owner: Merchandising Decomposition assigns a drop to a lever, and a lever to an owner.
🔍 Click to zoom - the outcome at the root, the levers and their owners at the leaves
LiveWhy "the number is down" is a useless sentence3 min

A top-line metric is a lagging outcome. On its own it starts an argument, because everyone has a pet theory and none of them can be checked. The tree ends the argument by making the drop specific: it was traffic, or conversion, or basket size, and each of those has a name on it. The leader's job is not to know which lever moved - it is to insist the conversation move to a lever before it moves to a fix.

  • The top line is the outcome you report to the board - GMV, revenue, ARR, bookings. Slow, lagging, everyone sees it.
  • The drivers are the levers teams actually pull - traffic, conversion, average order value. Faster, closer to the work.
  • The tree writes the relationship down so the room argues about the same picture instead of competing hunches.
Real world

Two boards, one headline. Lean Analytics (Croll and Yoskovitz) makes the point that the "one metric that matters" depends entirely on your business model and stage - and that the metric you report is never the metric you pull. A subscription board and a marketplace board can both say "down 8%" this quarter and be describing two entirely different problems. Without a tree, the two conversations sound identical and go nowhere.

Self-studyThe one-metric-that-matters idea2 min read

Lean Analytics argues that at any given stage a company should have a single metric it obsesses over - the one that captures whether the current bet is working. That does not contradict the tree; it sits on top of it. The one metric that matters is your root. The tree is how you make it actionable, because you cannot pull the root directly - you pull the levers underneath.

  • Pick the root deliberately. An early startup watches activation; a scaling one watches retention or GMV. The right root changes as you grow.
  • Then decompose it. A root without a tree is a wish. A root with a tree is a plan with owners.
  • One root, not seven. A dashboard with forty numbers and no tree is not visibility - it is noise wearing a suit.
Part 2 · covers DuPont analysis, ownership routing

The tree assigns ownership 7 min live

The quiet superpower of a driver tree is not diagnosis - it is accountability. When a metric splits into levers, each lever gets a team. A drop then becomes a routed question ("conversion fell 6%, product please look") rather than a blame meeting where everyone stares at the floor. The tree is an org chart in disguise: it tells you not only what moved, but whose phone should ring.

LiveFrom blame meeting to routed question3 min

Without a tree, a bad number lands on whoever is most senior or most visible, which is neither fair nor useful. With a tree, the number lands on the lever that moved, and the lever has a standing owner. The emotional temperature of the meeting drops immediately, because it is no longer "who failed" - it is "which lever, and what do we do about it." Good leaders build the tree in calm times precisely so it is on the wall when the number falls.

  • Every leaf has one owner. A driver two teams share is a driver neither team watches.
  • Route, do not blame. "Conversion is the guilty driver, product owns conversion" is a task, not an accusation.
  • Ownership survives reorgs. If a lever loses its owner, the tree tells you before the metric does.
Real world

DuPont, 1910s: the first famous driver tree. An engineer at DuPont broke return on equity into three levers - profit margin x asset turnover x financial leverage - so that a change in ROE could be traced to operations, efficiency, or financing, each owned by a different part of the business. A century later, every serious metric framework is the same move: split the outcome, name the levers, give each an owner. The maths has aged; the idea has not.

Self-studyThe exec who stared at a falling number for a month2 min read

A pattern worth recognising in yourself. A VP watches a weekly number slide for four straight weeks. Each week the update says "we are aware and monitoring." Nobody has decomposed it, so nobody can say what to do - and "monitoring" becomes a month of watching a house burn with a full bucket in hand. The fix was never more data. It was one whiteboard session breaking the number into three levers, at which point the culprit was obvious in ten minutes.

  • Monitoring is not diagnosis. If an update cannot name a lever, it is a status, not an answer.
  • Decompose early, not eventually. The tree is cheapest to draw before the crisis, not during it.
  • Ask the routing question. "Which driver, and whose is it?" ends the month-long stare in one meeting.
Self-studyROE as a tree you already trust2 min read

If the driver-tree idea feels abstract, remember you already trust one: any CFO reads return on equity as margin, turnover, and leverage, and can tell you in seconds whether a change came from pricing, operations, or the balance sheet. That fluency is exactly what this track builds for the metrics your operating teams live in - GMV, ARR, pipeline, DAU. The financial world decomposed its top line a hundred years ago. The operating world is still catching up.

Part 3 · covers Amplitude North Star (preview), decomposition discipline

What a good tree gives leaders 7 min live

A tree on the wall buys a leadership team three things: a shared picture so debate starts from the same map, faster diagnosis so a drop is traced not argued, and aligned accountability so each lever has a name. But there is a discipline to it - decompose to the level that changes a decision, and no deeper. A tree with forty leaves nobody owns is not rigour; it is vanity depth that hides the signal instead of surfacing it.

Top-line outcome board-facing · lagging you report this Traffic lever owner: Marketing · steer this Conversion lever owner: Product · steer this Avg order value lever owner: Merchandising · steer this The root is what you report. The leaves are who you route to.
🔍 Click to zoom - outcome at the root, owned levers at the leaves
LiveThe three gifts, and the one trap3 min

A shared picture means your weekly review starts from a map, not a slide of forty tiles. Faster diagnosis means a drop is a five-minute trace down the branches. Aligned accountability means each lever has a standing owner who is not surprised to be asked. The trap is over-decomposition: past the point where a deeper split would change a decision, extra leaves are cost, not insight. Decompose until the next level would not change what you do, then stop.

  • Shared picture. Everyone debates the same tree, so disagreements are about the world, not about the map.
  • Faster diagnosis. The drop lives on a branch. You trace it, you do not litigate it.
  • Aligned accountability. Each lever has an owner before the crisis, not assigned during it.
  • The trap: vanity depth. A leaf no one owns and no decision needs is noise dressed as rigour.
The stopping rule Split a driver one more level only if that level would change who you call or what you do. If it would not, you have decomposed enough. Depth is a means, never a trophy.
Self-studyA preview of leading vs lagging2 min read

Notice the note on the ownership map: the root is lagging and board-facing, the leaves are the things teams steer. That is the seed of session a2. The Amplitude North Star idea puts a lagging output metric at the top and a small set of leading input metrics beneath it, on the logic that inputs move first and warn early. You will see that the deeper you push a tree, the earlier its leaves warn you - which is why leaders steer by leaves and report by roots.

  • Root: lagging, slow, what the board sees.
  • Leaves: leading, quick, what teams pull.
  • Next session: a2 makes this the whole show - steer by leading, report by lagging.
Take it to your team

Five things to raise this week ◐ discussion prompts

Framework sources

Frameworks this session draws on

Metric decomposition has no single certificate - it is a craft distilled from a handful of durable frameworks. This leader track teaches the thinking core of these, applied to the decisions you make, not the code behind them. This page draws on:

Lean Analytics (Croll & Yoskovitz) - the one metric that matters, per business modelPart 1 · why the root is chosen, and why it still needs a tree
DuPont analysis - the original driver-tree decompositionPart 2 · ROE = margin x turnover x leverage, the century-old ownership split
Amplitude North Star Playbook - leading inputs move the lagging outputPart 3 preview · lagging root vs leading leaves; full treatment in a2
Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · Why can a top-line number, on its own, not be acted on?

A top-line metric is the score, not the play. You cannot move a lagging outcome directly; you move the drivers beneath it. The tree is what turns the number into levers a team can pull.

2 · Beyond faster diagnosis, what does decomposition assign?

The quiet power of a tree is ownership. Each lever gets a team, so a drop becomes a routed question ("conversion fell, product please look") instead of a blame meeting.

3 · What was DuPont's move in the 1910s?

DuPont broke ROE into three levers so a change could be traced to operations, efficiency, or financing, each owned by a different part of the business. Every modern metric framework repeats that move.

Leader session 1 cheat sheet · pin this

Driver treeA top-line metric broken into the levers that multiply or sum to make it. Outcome at the root, owned levers at the leaves.
The number is an outcomeA top line tells you what, never why. Two firms can be "down 8%" for opposite reasons. Never act on the score alone.
Decomposition assignsA drop to a lever, and a lever to an owner. That is the whole point - accountability, not just insight.
Route, do not blame"Conversion fell, product owns conversion" is a task. It drops the emotional temperature of a bad-number meeting.
DuPont, 1910sROE = margin x turnover x leverage. The first famous driver tree. Every framework since is the same move.
One metric that mattersLean Analytics: pick the root deliberately for your stage, then decompose it. A root without a tree is a wish.
The stopping ruleSplit one level deeper only if it changes who you call or what you do. Past that, extra leaves are vanity depth.
Running skillReframe "the number is down" as "which driver, and whose is it?" Next: a2, leading vs lagging.