# Product analytics: activation, retention cohorts and the North Star metric

Source: https://sarang.pro/product-analytics-activation-retention-north-star/
Author: Sarang | Published: 25 October 2026

Most small product teams have plenty of data and very little insight. The analytics tool shows page views, sign-ups and daily active users, the numbers go up and down, and nobody is sure what to do differently. What helps is a short list of measures that track whether users get value, and the habit of looking at them by cohort.

This guide covers three ideas that do most of the work: activation, retention cohorts and the North Star metric. It explains each one, how to set it up in common tools, and how they fit together.

## Start with what value looks like
Before choosing any metric, write one sentence describing the moment a user gets real value from your product. For a booking tool it might be "a client books an appointment online without phoning". For an invoicing app, "an invoice gets paid". Everything else in this guide hangs off that sentence.

Amplitude calls this the [aha moment](https://amplitude.com/blog/aha-moment): the point when a user grasps the product's core value. Its guide links the aha moment closely to activation and quotes Sean Ellis describing it as "the moment that the utility of the product really clicks for the users."

## Activation: did new users reach value?
An activation metric measures the share of new users who reach that value moment within a set time. It is written as a specific action and a deadline, for example "created and sent a first invoice within seven days of signing up".

To find a good activation event, compare the early behaviour of users who stayed with those who left. Amplitude suggests using product analytics to compare retention across groups, such as users who adopted a certain feature in their first days against those who did not. It also gives a warning that matters: these patterns show correlation, not cause. If you only study users who finished a particular onboarding flow, you may credit the flow for retention that came from something else.

So treat a candidate activation event as a hypothesis. Then test it by helping more users reach it and checking whether their retention improves.

## Retention cohorts: are they still here?
A cohort is a group of users who share a starting point, usually the week or month they signed up. [Cohort analysis](https://en.wikipedia.org/wiki/Cohort_analysis) tracks each group separately over time, so you can see whether newer users behave differently from older ones. A single blended active-user figure cannot show this, because growth in new sign-ups hides losses among existing users.

Both major tools build these reports in a similar way. In [Google Analytics 4 cohort exploration](https://support.google.com/analytics/answer/9670133), you set an inclusion criterion (such as first visit), a return criterion (such as a purchase) and a granularity (daily, weekly or monthly). GA4 offers three calculations: standard counts users who returned in that specific period, rolling counts users who returned in that period and every earlier one, and cumulative counts users who returned in any period up to that point. [Mixpanel retention reports](https://docs.mixpanel.com/docs/reports/retention) use a birth event and a return event, and default to "on or after" retention, which counts users who return in a period or any later one.

Choose the return event carefully. "Any event" makes retention look better than it is, because opening an email link counts. Use the action that represents value, ideally the same one behind your activation metric.

## The North Star metric: one number for the team
Amplitude's [North Star guidance](https://amplitude.com/blog/product-north-star-metric.md) describes a single metric that best captures the value customers get from the product and links that value to revenue. It sets three tests for a good one:

- **It reflects customer value,** usually through actions that deliver realised value. Registered users and daily active users fail this test.
- **It represents product strategy,** so someone reading it can see where the product is heading.
- **It is a leading indicator** that predicts future success, unlike revenue, which reports the past.
Beneath it sit three to five input metrics that the team can influence directly. Amplitude suggests thinking about breadth (how many users), depth (how much they do), frequency (how often) and efficiency (how quickly). Its examples include "number of mobile orders delivered" for a retailer and "trial accounts with more than three users active in week one" for self-serve software.

## Putting it together
An example (illustration only). A UK company offers online booking software for independent hair salons.

- **Value sentence:** a salon's client books an appointment online without calling.
- **North Star:** online bookings completed per week across all salons.
- **Inputs:** salons with at least one online booking this week (breadth), online bookings per active salon (depth), share of salons taking bookings every week of the month (frequency), and median days from sign-up to first online booking (efficiency).
- **Activation:** share of new salons that receive their first online booking within 14 days.
- **Retention:** monthly sign-up cohorts, with "received an online booking" as the return event.
When the team looks at cohorts, it sees that salons which add their booking link to Instagram in the first week retain far better. That is a correlation, so the team tests it: half of new salons get a prompt and a short guide to adding the link, the other half do not. If the prompted group activates and retains better, the prompt stays.

## Common mistakes
- **Vanity metrics.** Total sign-ups and page views rise even when the product is failing.
- **Revenue as the North Star.** Amplitude notes that revenue lags, so it gives no early signal of whether product changes work.
- **Weak return events.** Counting any activity as retention flatters the numbers.
- **Too many North Stars.** Most single-product businesses need one.
- **Mistaking correlation for cause.** A behaviour linked to retention is a lead to test, not a proven lever.
- **Tracking everything.** Events nobody looks at cost time to maintain. Track what supports your value sentence first.

## When this is worth the effort
This set-up pays off once you have enough users for cohorts to mean something, usually a few hundred sign-ups a month. Below that, individual conversations with users teach you more than charts. Start with the value sentence and an activation metric, add cohort retention when you have a few months of data, and agree a North Star when more than one team needs a shared target.