User activity score

A user activity score is a single weighted number that summarises how engaged a user is, built from the actions they take inside a product over a defined period. Instead of looking at logins, feature clicks, and time in app as separate columns, a score combines them into one figure. A user who logs in daily, uses core features, and invites teammates lands a high score. A user who logged in once and never returned lands near zero.

The score works because the actions are weighted, not just counted. Opening the app is a weak signal of value. Completing a core workflow is a strong one. A good activity score gives more weight to the actions that actually predict retention and revenue, so the number reflects meaningful engagement rather than idle clicks. If a user logs in ten times but never does anything, the score stays low on purpose.

Activity scores matter because they turn a wall of behavioural data into something a team can act on. Customer success can sort accounts by score and reach the ones slipping before they churn. Product can see which features move the score and which do not. The score is most useful when it is tied to outcomes, so a falling score reliably warns of a user heading toward churn and a rising score reliably marks a user heading toward expansion.

A user activity score is a weighted sum of the actions a user takes in a period, usually normalised to a fixed range so scores stay comparable. The inputs below show how to build one that means something.

1. 1

   Choose the tracked actions

   Pick the behaviours that signal engagement and value, such as logins, core feature use, content created, teammates invited, and integrations connected. Avoid vanity actions that anyone can trigger without getting value. The shortlist of actions defines what the score actually measures.

2. 2

   Assign a weight to each action

   Give each action a multiplier that reflects how strongly it predicts retention or revenue. A login might score 1, completing a core workflow might score 5, and inviting a teammate might score 8. Weights are where the judgement lives, so anchor them in data on what separates retained users from churned ones.

3. 3

   Count actions over a fixed period

   Decide the window, commonly the last 7, 14, or 30 days, and count how many times each user performed each action inside it. A rolling window keeps the score current and lets a once-active user decay as they go quiet.

4. 4

   Normalise to a comparable scale

   Raw weighted sums vary with how heavy a user is, so map the result onto a fixed range such as 0 to 100. Normalising lets you compare a power user against a new sign-up and set thresholds, for example flagging any account below 30 as at risk.

A metric tree decomposes the user activity score into the action groups that feed it, so a movement in the headline number traces back to a specific behaviour and a specific owner. Because the score is a weighted sum, each branch is one cluster of weighted actions.

The first level splits the score into the categories that make it up, such as core usage, breadth of feature use, collaboration, and setup depth. Each category then breaks into the individual actions and their weights. Core usage might decompose into login frequency and workflow completions. Collaboration might decompose into teammates invited and shared items created. The tree makes it obvious that a falling score caused by collapsing collaboration needs a different fix from one caused by users abandoning core workflows.

KPI Tree lets you attach RACI ownership to each branch. Product owns the activation actions that drive core usage, customer success owns the collaboration and setup signals, and the accountable owner is pushed an alert when their branch drags the score down. A verified impact loop then checks whether the change they shipped actually lifted the score for the affected users.

There is no universal activity score benchmark, because the scale and weights are yours. What is portable is how to read bands of the score against behaviour and outcomes. The table below shows a typical 0 to 100 banding and what each band tends to mean.

The number to watch is not the average score but its distribution and its movement. A product can have a healthy mean while a quarter of users sit in the dormant band heading for churn. Track the share of users in each band over time and watch for users crossing downward, because a user sliding from engaged to light is an earlier and clearer warning than waiting for them to cancel.

Raising the activity score means getting more users to take the high-weight actions more often, not inflating the count of low-weight ones. The aim is real engagement that predicts retention, so improvements should move users up through the bands, not just nudge the raw number.

Start with the branch of the tree that is dragging the score most. If collaboration is near zero across the base, building invite and sharing loops will lift more users than polishing an already-strong core usage branch.

KPI Tree connects the activity score to the actions and owners behind it. Product owns the activation flow that drives core usage, customer success owns the re-engagement of dormant accounts, and the accountable owner is notified when their branch moves the score. The verified impact loop then confirms whether the intervention actually raised the score for the users it targeted, so effort goes where it provably works.

1. 1

   Counting every action equally

   An unweighted score rewards busywork and hides real disengagement. Opening the app is not the same as completing a core workflow. Without weights, the score measures noise instead of value.

2. 2

   Setting weights and never revisiting them

   Weights that were right at launch drift out of date as the product changes. An action that once predicted retention may stop doing so. Review the weights against fresh retention data so the score stays meaningful.

3. 3

   Using a window that is too long

   A score built over a year smooths away the recent drop-off that matters most. A user who has gone quiet this month still looks active on a stale long window. Use a rolling window short enough to let disengagement show.

4. 4

   Watching only the average score

   A healthy mean can hide a large dormant tail heading for churn. Track the distribution and the share of users in each band, not just the single average number.

5. 5

   Treating the score as the goal

   The score is a proxy for engagement, not the outcome itself. Optimising the number without checking it still predicts retention and revenue leads to gaming. Keep validating that a higher score really does mean a healthier user.