# The north star metric and its KPI tree

> One north star metric for the value users get, and a KPI tree of input metrics that teams can actually move.

- Course: [Product Analytics: From Data to Decisions](https://productovi.com/courses/product-analytics.md)
- Module: Which numbers should we steer by?
- Lesson 5/32, ~28 min
- Source: https://productovi.com/courses/product-analytics/north-star

## Five numbers, one slide

**Alex Tutor:** Investor season at Halves. Maya needs one number, and since you answer from the data on the product team, she's asking you.

**Maya:** The investors want **one** number that shows whether Halves is getting better. My dashboard has five: downloads, MAU, app opens, expenses added and Plus revenue. Which one do I pick?

**Sofia:** MAU? Everyone reports it, so it's easy to compare.

**Alex Tutor:** MAU says how many people touched the app this month. It doesn't say whether splitting a bill worked for them. We want a **north star metric**: the one number that captures the value users get and, because of that, predicts the business.

![A person steers a small boat at night towards one bright star, past a cluster of floating dials pointing in different directions.](https://productovi.com/courses/product-analytics/inline/north-star-illustration-2.webp)

*Many numbers on the dashboard, one to steer by: the value users get.*

**Alex Tutor:** A quick test for any candidate: can it go up while users are worse off? Downloads can double during a paid campaign in which nobody splits a single bill. App opens rise if a bug makes people check their balance three times.

**Theo:** Hold on, downloads are my job. Isn't growth the whole point?

**Alex Tutor:** Growth matters, it just isn't the north star. Your downloads feed into it, and the KPI tree we'll build shows exactly where.

> **Exercise:** Which of these Halves numbers **can't** rise while users get less value from the app?

## Where Halves creates value

**Alex Tutor:** Start from the value, not from the dashboard. Halves does one job: a group knows who owes whom without awkward conversations.

**Sofia:** So one person typing in expenses alone isn't getting that value?

**Alex Tutor:** Alone, it's a spreadsheet. The value appears when a second member sees the balance and trusts it. So the unit is the **group**, not the user.

![One person sits alone with a receipt and a phone, looking unsure; beside them three friends lean over one shared phone and nod.](https://productovi.com/courses/product-analytics/inline/north-star-illustration-3.webp)

*Halves pays off when a second member sees the balance and trusts it.*

**Maya:** Then something like “weekly active groups”: groups where at least two members did something real that week.

**Alex Tutor:** A strong candidate. Weekly, because most groups share costs every week. And written as a full definition, like any metric: server-created recurring bills and staff accounts don't count.

> **North star candidate: weekly active groups**
>
> - **Qualifying action:** `expense_added` by a person (not `entry_method = recurring`) or `settle_up_recorded`
> - **Unit counted:** distinct `group_id` where **at least two different members** did a qualifying action
> - **Exclusions:** staff and test accounts
> - **Time window:** Monday 00:00 to Sunday 23:59
> - **Reporting time zone:** UTC

> **Exercise:** Which properties should a good north star metric have? Select all that apply.

## How to build a KPI tree from a north star

**Sofia:** But no team can walk in on Monday and “move weekly active groups”. It's too big.

**Alex Tutor:** Right, so we break it into a **metric tree**, also called a **KPI tree** or driver tree: the top metric split into the **input metrics** that drive it, until each branch is something one team can move.

**Alex Tutor:** This week's active groups come from three places: groups active last week that stay active, groups active for the first time, and groups coming back after a break.

> **A metric tree for Halves**
>
> **Weekly active groups** =
>
> - **Retained groups:** active last week and this week
>   - share of groups with a recurring bill set up
>   - share of balances settled within 14 days
> - **Newly active groups** = new groups × share that get a second member × share that add a shared expense in week 1
>   - new groups: store search, paid social, invites
> - **Returning groups:** active before, quiet last week, active again
>   - trip groups reused for a new trip
>   - reminder opens by quiet groups

**Theo:** So my campaigns sit down there, under “new groups”.

**Alex Tutor:** Yes, and the tree shows the catch: new groups only count once they get a second member. A campaign that brings people who never invite anyone raises your number and leaves the top where it was.

> **Exercise:** Match each input metric in the Halves tree to the team best placed to move it.

## Reading the tree: a flat week that isn't calm

**Maya:** Let's try it. Last week we had **1,600** weekly active groups. This week **1,360** of them were active again, **190** groups were active for the first time and **70** came back after a break.

**Sofia:** So I add up the three branches and compare the total with 1,600?

**Alex Tutor:** Exactly. A group counts once, in exactly one branch, so the branches add up to the top.

**Maya:** Go on then. Did we grow this week?

> **Exercise:** Last week Halves had 1,600 weekly active groups. This week 1,360 of them were active again, 190 groups were active for the first time and 70 came back after a break. How many weekly active groups does Halves have this week?

**Alex Tutor:** Before the story behind the +20, one input on its own: how much of last week's base did Halves keep?

> **Exercise:** Of last week's 1,600 weekly active groups, 1,360 were active again this week. What share of last week's groups was retained, in %?

**Alex Tutor:** Plus 20 looks like a quiet week. Inside it, 240 groups went silent and 260 new or returning groups replaced them.

![Weekly active groups split into retained, newly active and returning branches: +20 overall hides 240 groups going silent and 260 replacing them.](https://productovi.com/courses/product-analytics/diagrams/m2-l1-kpi-tree-week.png)

**Sofia:** So a flat top can be a busy revolving door. If new groups slow down, say Theo pauses a campaign, the top falls fast.

![Groups of friends walk out of a café's revolving door while as many new groups walk in; the tables inside stay equally full.](https://productovi.com/courses/product-analytics/inline/north-star-illustration-4.webp)

*A flat top can hide many groups leaving and as many arriving.*

**Theo:** Great, so if I pause for a week, I get blamed for the whole chart.

**Alex Tutor:** Not with a tree. It shows whether the drop came from fewer new groups or from more groups going quiet. That's what it's for: telling you which branch to open next.

**Alex Tutor:** Let's test that on a made-up week after Theo pauses paid social. Which branch would you open first?

> **Exercise:** The week after a 1,620-group week, Halves has 1,540 weekly active groups: 1,375 retained, 95 newly active and 70 returning. The week before, the branches were 1,360, 190 and 70. Which branch explains the drop of 80?

> **Exercise:** Theo doubles sign-ups with a new campaign, but none of the new users ever gets a second member into their group. **True or false:** weekly active groups go up.

## Why revenue isn't the north star metric

**Maya:** Honest question: why not Plus revenue? It pays the salaries.

**Alex Tutor:** Revenue is a result of value, and it lags: a group that stops sharing costs today cancels Plus next month. Steer by revenue and the easy moves are paywalls in front of the features groups need.

**Alex Tutor:** Keep revenue next to the north star, checked every month. If weekly active groups grow and revenue doesn't, the problem is in pricing, not in the product's value.

**Alex Tutor:** Maya's board will push for revenue anyway. Here are two ways to set up the dashboard; pick one.

> **Exercise:** Which setup would you recommend to Maya?

**Maya:** Then test me. Someone on the team suggested “expenses added per week” as the north star. What's wrong with it?

> **Exercise:** Someone proposes **“expenses added per week”** as the Halves north star. In 3–5 sentences, give one strength of this metric, at least two weaknesses, and say what you'd use instead.

> **Key ideas**
>
> - A **north star metric** captures the value users get. Test: can it rise while users are worse off? Downloads, opens and MAU can.
> - For Halves the value lives in a group, so the unit is the group: **weekly active groups** with at least two members acting.
> - Write the north star as a full metric definition: action, unit, exclusions, window, time zone.
> - A **metric tree** splits the top into **input metrics** teams can move: retained + newly active + returning.
> - A flat top can hide big flows in and out; the branches show which one to fix.
> - Revenue sits next to the north star, not in its place.

**Alex Tutor:** So Maya's slide gets one number, weekly active groups, with its definition underneath and three branches behind it.

- Previous lesson: [A tracking plan for a new feature](https://productovi.com/courses/product-analytics/tracking-plan.md)
- Next lesson: [Volume, quality and Simpson's paradox](https://productovi.com/courses/product-analytics/volume-vs-quality-metrics.md)

---
Productovi Learning. Free, no sign-up needed to read.