Why this matters
A team that reports downloads is rewarded for filling the top of a leaky FunnelModel of the customer journey stages., whether or not anyone learns a word. A north star metric replaces that with a count of value delivered, chosen to lead revenue, so that every team can see whether its work helps learners.
This exercise applies the tests in the north star metric and objectives framework and the mapping method in the growth loops in practice guide. The newsletter growth loop case shows the same move on a small scale: its team stopped watching open rate and chose a measure tied to the readers the business actually needed.
The app and its current measure
Parlaro is a fictional app that teaches conversational Spanish, Italian and Portuguese through short daily lessons. It is free to use with a paid tier that unlocks speaking practice. Its founders present downloads at every investor update. The product lead suspects that number hides a problem and has pulled the figures below.
Parlaro: one quarter's figures
| Measure | Value |
|---|---|
| Downloads in the quarter | 180,000 |
| Registered accounts created | 96,000 |
| Monthly users who opened the app at least once | 71,000 |
| Learners who completed three or more lessons in a week (average week) | 14,500 |
| Lessons completed in the quarter | 2,340,000 |
| Total minutes in the app | 19,800,000 |
| Paying subscribers at quarter end | 6,200 |
| New learners who completed three lessons in their first week | 23,000 |
| "Practise with a friend" invites sent | 11,400 |
Retention cut: of new learners who completed three or more lessons in their first week, 41% were still completing lessons in week eight. Of those who did not, 9% were. Paying subscribers come almost entirely from learners who reached three lessons a week for a month.
Comparing candidate metrics
Treat each candidate as a hypothesis and put it through the five tests: does it rise only when a learner got what they came for; does it move before revenue; can it be counted today, by cohort; can product, marketing and support each name an input they control; and can it be inflated without creating value?
Then decompose the winner. The usual shape is breadth (how many learners), depth (how much each learns), frequency (how often they return) and efficiency (how fast a new learner reaches value). Each input becomes a KPIA metric tied directly to an objective. that a team can own.
Finally, look for the loop. Ask what an engaged learner produces that could bring the next one, put a rate on each arrow and find the arrow where a feasible improvement moves the loop most.
Task
- Score each candidate metric in the report against the five north star tests and record which it fails.
- Choose the north star and write its definition: the event, the threshold, the population and the period.
- Break it into input metrics, naming which team owns each.
- Sketch one growth loop that feeds the north star, put a rate on each arrow, and mark the weakest arrow.
- Name one guardrail metric the north star cannot see.
Worked answer
Scoring. Downloads and accounts fail on value and are easy to inflate with spend. Monthly openers fail on value: an open is not a lesson. Total minutes fails as hard to game, since a slower lesson raises it. Lessons completed is close but is driven by a few heavy users. Paying subscribers is revenue in disguise and lags. Weekly learners completing three or more lessons passes every test, and the retention cut shows the threshold separates learners who stay from those who leave.
Definition. Weekly active learners: distinct learners who complete at least three lessons in a calendar week, counted every week and readable by sign-up cohort.
Parlaro: the north star and its inputs
| Level | Metric | Current reading | Main owner |
|---|---|---|---|
| North star | Weekly active learners (three or more lessons in the week) | 14,500 | Whole company |
| Breadth | New learners reaching three lessons in their first week, per week | 23,000 ÷ 13 weeks ≈ 1,770 | Marketing and onboarding |
| Efficiency | Share of new accounts reaching that threshold in week one | 23,000 ÷ 96,000 ≈ 24% | Product |
| Frequency | Share of weekly active learners active again the next week | to be measured by cohort | Product and lifecycle email |
| Depth | Lessons per weekly active learner | to be measured | Content team |
| Guardrail | Speaking-practice completion among paying subscribers | to be measured | Support and product |
Loop. The "practise with a friend" feature is an invitation loop: an engaged learner invites a friend, the friend installs, completes lessons and becomes an engaged learner who can invite again.
Parlaro: mapping the invitation loop for one quarter
Engaged learners in the quarter: 23,000. Invites sent: 11,400, about 0.5 per engaged learner. Invites accepted with an install: 2,850, a 25% acceptance rate. Invited installs reaching three lessons in week one: 1,140, a 40% rate, well above the 24% for all accounts. Loop factor: 1,140 ÷ 23,000 ≈ 0.05 new engaged learners per engaged learner per quarter. Doubling invites per engaged learner to about 1.0, with the other rates unchanged, would roughly double the loop's output to around 2,280 engaged learners a quarter.
The weakest arrow is the invitation itself: few engaged learners invite anyone, while invited friends activate unusually well. A prompt at the moment a learner finishes a streak week is the feasible change. The loop sits far below one, so it subsidises acquisition rather than replacing it, which is normal.
Guardrail. Lifetime value, LTVTotal value a customer brings over time., is invisible to the north star; a push that adds weekly learners through free-tier gimmicks could leave paid conversion flat. Watching speaking-practice use among subscribers and paid conversion by cohort catches that.
Common mistakes
- Choosing revenue or subscribers. It lags value and tells no team what to change this week.
- Picking a threshold without evidence. The three-lesson line is justified by the retention cut, not by taste.
- Inputs nobody owns. An input with no named team becomes a number everyone watches and nobody moves.
- Working the share prompt before activation. Invitations come from learners who got value; the activation rate still decides how many can invite.
- Calling a funnel a loop. If no measured arrow returns from an existing learner to a new one, say so.
Test yourself next with the north star and loops checkpoint, or see the module in order on the beginner growth marketing level page.
