PM Gym

KPI Trees

Break a big goal metric into the levers you can actually pull

A single top-line number like revenue is too blunt to act on. A KPI tree breaks it into the drivers that feed it — turning a vague goal into a map of specific levers, and showing exactly where a metric moved when it does.

Step-by-step lessons

Build & Use a KPI Tree

Four short lessons: what a KPI tree is, how to build one, how to find the best lever, and the pitfalls to dodge.

1

What a KPI Tree Is

A KPI tree (or metric tree) takes one top-level goal metric and breaks it down into the smaller metrics that mathematically combine to produce it. ("KPI" just means Key Performance Indicator — a number you track to see how you're doing.) Suddenly a huge, untouchable number becomes a set of pullable levers.

Everyday example — the household budget

"We're spending too much money" is impossible to act on — there's no single knob for "spending." But break it apart: spending = rent + groceries + subscriptions + transport + eating out. Now you can see the levers. Maybe rent is fixed, but three forgotten subscriptions and daily takeout are the real drivers. A KPI tree does the same thing to a business metric: it turns a vague worry into a short list of things you can actually change.

Revenue, decomposed

"Grow revenue" is hard to act on. But:

Revenue = Active Users × Conversion Rate × Average Revenue Per User

Now you can see three distinct levers. A team can go after conversion while another grows active users — and you can tell which one actually moved the top line.

Quick check

What is the main thing a KPI tree gives you that a single top-line number doesn't?

2

Building the Tree

You build a KPI tree top-down: start with the goal metric, then keep asking "what does this break into?" — where each level's children combine (by × or +) to produce the parent. The key discipline is keeping each split MECE — a fancy term for a simple idea, explained just below.

Everyday example — slicing a pizza (MECE)

MECE stands for "Mutually Exclusive, Collectively Exhaustive." In plain words: cut the pizza so every bite belongs to exactly one slice (no overlaps), and the slices together make up the whole pizza (no missing bits). "New users + returning users" is a clean MECE cut of all active users — everyone is one or the other, and together they're everybody. "Happy users + users on iPhone" is not: someone can be both (overlap) and plenty are neither (gap). Good tree branches are always clean slices.

1

Start at the top

The one metric that matters most right now — often the North Star.

2

Decompose one level

Break it into the 2-4 inputs that mathematically produce it.

3

Repeat downward

Break each input into its own drivers until you reach something a team can act on.

4

Keep branches clean

Inputs at each level should be MECE — no overlaps, no gaps.

Quick check

You're decomposing "monthly active users." Which is a valid next level of the tree?

3

Finding the Best Lever

Once the tree is drawn, not every branch is worth chasing. The best lever scores high on two things at once: how much it could move the top metric (impact), and how realistically you can move it (feasibility).

Everyday example — fixing up a house

Say you want to raise your home's value. Repainting a doorknob is easy but changes almost nothing (low impact). Adding a second floor would change a lot but is wildly expensive and slow (low feasibility). Refreshing the kitchen is both meaningful and doable — that's the sweet spot. In a KPI tree you're hunting for the "kitchen refresh" branch: a real driver of the top number that you can actually move with the time and resources you have.

Chase this

High impact + high feasibility

A big driver of the top metric that you can realistically influence. Your prime target.

Maybe later

High impact, low feasibility

Matters a lot but hard to move (e.g., market-wide pricing). Note it; don't start here.

Skip

Low impact

Even if easy, moving it barely nudges the top line. Easy wins that don't matter are still a waste.

Quick check

Your tree shows two levers: (A) checkout conversion — a huge driver of revenue and clearly fixable, and (B) footer link clicks — easy to boost but a tiny driver. Where do you start?

4

Vanity & Broken Branches

KPI trees fail in predictable ways. Two traps do most of the damage: measuring things that look good but mean nothing (vanity metrics), and branches that don't actually add up.

Everyday example — the odometer vs. the speedometer

A car's odometer (total miles ever driven) only climbs — it never tells you how fast you're going right now or whether you're about to crash. The speedometer moves both ways and reflects your current state. "Total sign-ups ever" is an odometer: it always goes up, even while the product is falling apart. "Weekly active users" is a speedometer — it can drop and warn you. Build your tree from speedometers, not odometers.

The vanity trap

"Total registered users, all-time" only ever goes up and never goes down — it feels great and tells you nothing about whether the product is healthy today. That's a vanity metric.

Prefer actionable metrics: ones that a specific decision can move, and that can go down as well as up. And check every branch actually composes its parent.

Quick check

Which of these is a vanity metric you should be wary of putting at the heart of a KPI tree?

Review the concepts

KPI Trees Flashcards

6 cards covering the essentials. Click a card to flip it.

Core Card 1 of 6

KPI Tree

Click to flip

Tip: say the answer out loud before flipping.

Explanation

In practice

1 / 6
Apply what you learned

Practice Scenarios

15 situations that test whether you can build a clean tree, diagnose a metric move, pick the right lever, and spot vanity numbers. Choose the strongest move, then read why the others miss.

Scenario 1

Revenue is down 10% this quarter and leadership wants to know why. You have a KPI tree: Revenue = Active Users × Conversion × ARPU.

How does the tree help you diagnose it?

Scenario 2

A teammate proposes making "total lifetime registered accounts" the team's headline metric because "it always goes up and looks great to investors."

What's your response?

Scenario 3

You want to grow your North Star. Two levers sit in the tree: (A) improving activation — a major driver you can influence — and (B) adding more social-share icons, easy but a minor driver.

Which do you prioritize, and why?

Scenario 4

A colleague draws a tree: "Revenue = Marketing spend + Number of features + Team happiness."

What's fundamentally wrong with this tree?

Scenario 5

You split "active users" into "users who log in via email" and "users who are happy with the product."

Why does this split break MECE?

Scenario 6

Revenue is flat, but your tree shows active users up 20% while conversion rate dropped 18%. Everyone is confused about whether things are good or bad.

What does the tree tell you?

Scenario 7

Two levers have equal impact potential. Lever A requires a six-month platform rebuild; lever B is a copy-and-layout change your team can test next week.

Which do you start with, all else equal?

Scenario 8

Your team obsesses over "cumulative app installs," which is now 5 million and rising. Meanwhile monthly active users has quietly fallen for three straight months.

What's the diagnosis?

Scenario 9

A stakeholder wants the KPI tree taken down to twelve levels of depth "so we capture every possible sub-metric."

What's the practical caution?

Scenario 10

Your North Star is "weekly active teams." You decompose it as: New teams activated + Existing teams retained + Churned teams reactivated.

How good is this decomposition?

Scenario 11

The exec team wants ONE North Star metric, but sales wants revenue, growth wants signups, and the CS team wants retention.

How can a KPI tree resolve this?

Scenario 12

Conversion rate is your chosen lever. A designer suggests boosting it by auto-checking an "add premium add-on" box by default, which users often miss.

What's the systems-aware caution before celebrating a conversion bump?

Scenario 13

You need to raise "orders per month." The tree splits it into Traffic × Conversion × Repeat-purchase rate. Data shows traffic is huge, conversion is average, and repeat-purchase is far below industry benchmark.

Where's the most promising lever?

Scenario 14

A teammate insists "average revenue per user" is meaningless because your users are wildly different — some pay nothing, a few pay thousands.

What's the sharpest response?

Scenario 15

Leadership asks you to prove that last quarter's onboarding project "worked." You have a full KPI tree in place.

How does the tree let you answer cleanly?

Lock it in

Guess the Term

Read the clues and name the concept. The fewer clues you need, the more points you score.

Round 1 Score 0
Keep it handy

KPI Trees Quick Reference

The whole topic on one screen.

Building the Tree

Top-down, each level composing the one above.

1

Start at the top

The one metric that matters most — often the North Star.

2

Decompose

Break it into the 2-4 inputs that produce it (× or +).

3

Repeat down

Until you reach a metric a team can act on.

4

Keep it MECE

No overlaps, no gaps at each level.

Choosing & Checking

Best lever

High impact on the top metric and feasible to move.

Diagnose moves

When the top metric shifts, find which branch moved.

Avoid vanity metrics

Prefer actionable metrics that can fall as well as rise.

Vanity vs. Actionable

The odometer-vs-speedometer test.

Vanity (odometer)

Only ever rises: total sign-ups, cumulative installs, all-time revenue. Hides decline.

Actionable (speedometer)

Moves both ways: weekly actives, conversion rate, retention. Reflects health now.

Averages hide spread

If an average blurs wildly different users, split it by segment.

Watch offsetting moves

A flat top line can hide two big branches cancelling out.

Common Decompositions

Starting formulas you can adapt.

Revenue

Active users × Conversion × Avg. revenue per user.

Active users

New activated + Retained + Reactivated (a clean MECE split).

Orders

Traffic × Conversion × Repeat-purchase rate.

Stop rule

Decompose only until a branch is something a team can act on.

Notification