Quick answer
User engagement is how much people take part in a product in ways that matter. Measure it as breadth (share of active users who did a key action), depth (key events per active user) and, for daily products, frequency. One number such as time spent hides what changed.
What is user engagement?
User engagement is how far people take part in a product in ways that matter to them and to you. The academic starting point is O'Brien and Toms (2008), who described engagement as a quality of the user's experience and as a process with four stages: a point of engagement, a period of sustained engagement, disengagement, and re-engagement. Product teams cannot observe the experience, so they measure behaviour that stands in for it: what active people actually do, how many of them do it, and how often.
Two things it is not. It is not retention: retention asks whether people come back, engagement asks what they do while they are here. And on social platforms "engagement" means likes, comments and shares per post, which is a different measure; this page is about people using a website, an app or a product. For a catalogue of engagement metrics with formulas, see engagement metrics. For the session-level version, see engaged session.
Why one number misleads
Time spent is the usual headline, and it can move the wrong way. Sample data: a product has 10,000 active users in each of two months. In month 2 a redesign ships.
| Month 1 | Month 2 | Change | |
|---|---|---|---|
| Active users | 10,000 | 10,000 | 0 |
| Average session length | 4.0 min | 5.5 min | +37.5% |
| Users who did the key action | 3,200 (32.0%) | 2,700 (27.0%) | −5.0 points |
| Key events in total | 9,600 | 8,370 | −12.8% |
| Key events per active user | 0.96 | 0.837 | −12.8% |
| Key events per user who did it | 3.0 | 3.1 | +3.3% |
How to recompute: 3,200 ÷ 10,000 = 32.0%; 9,600 ÷ 10,000 = 0.96; 9,600 ÷ 3,200 = 3.0; 8,370 ÷ 2,700 = 3.1; 0.837 ÷ 0.96 − 1 = −12.8%. A dashboard that tracks session length alone calls month 2 a win. The two ratios tell a different story: fewer people reached the key action, and those who did used it about as much as before. Longer sessions can mean more interest or more confusion; the falling share of people reaching the key action says which to suspect first.
The three lenses: breadth, depth and frequency
Engagement needs at least two numbers, taken over the same active users and the same window, and a third if the product is meant for daily use.
| Lens | Question | Formula | Month 1 sample | Blind spot |
|---|---|---|---|---|
| Breadth | How many active users did the key action? | Users who did it ÷ active users | 3,200 ÷ 10,000 = 32.0% | One click counts as much as fifty |
| Depth | How much do they do it? | Key events ÷ active users; also ÷ users who did it | 9,600 ÷ 10,000 = 0.96; 9,600 ÷ 3,200 = 3.0 | A few heavy users can lift the average |
| Frequency | How often do they show up? | Average daily active users ÷ monthly active users | 1,800 ÷ 10,000 = 18%, about 5.4 of 30 days | Depends on how often the product is naturally used |
Frequency is the DAU/MAU ratio; a tax tool used a few times a year will score low and be healthy. Rules of use:
- Fix the definition of "active" before you compare anything. Any event, a page view after sign-in, and a core action give very different counts.
- Report breadth and depth side by side. If breadth falls while depth among doers holds, the change is about reach, not enjoyment.
- Add frequency only when the product has an expected rhythm, and compare with your own history or between user groups.
When the average hides the shape
Depth as an average can hide how unevenly use is spread. Sample data: two products with 1,000 active users each and the same mean of 3.0 key events per user.
| Product X | Product Y | |
|---|---|---|
| Users with 3 key events | 1,000 | 0 |
| Users with 30 key events | 0 | 100 |
| Users with none | 0 | 900 |
| Mean events per user | 3.0 | 3.0 |
| Breadth | All 1,000 users | 10% |
| Median | 3 | 0 |
How to recompute: product Y has 100 × 30 = 3,000 events over 1,000 users = 3.0, and 100 ÷ 1,000 = 10% breadth. Product X is used a little by everyone; product Y is used heavily by a tenth of its users and not at all by the rest. The means are identical; the right responses are not. If the median is far below the mean, or the median is zero, look at the distribution before you trust the average.
How to choose the key action
The measures above are only as good as the key action. A usable one passes four tests:
- It carries the value. Created, shared, sent or completed something, not just logged in or viewed a page.
- It is not a visit. Logins and page views measure presence; they belong in the definition of "active", not in the key action.
- It happens often enough. In a 30-day window enough people should do it for the groups to be big enough to compare.
- It is hard to trigger by accident. An accidental tap should not count.
| Product | Key action that passes | Action that fails |
|---|---|---|
| Messaging app | Sent a message | Opened the app |
| Online store | Placed an order | Viewed a product page |
| Document tool | Edited a document | Signed in |
| Learning app | Completed a lesson | Started the app |
Keep the list short. A composite of several core events is covered in product engagement score, which suggests 3–10.
Common mistakes
- Time spent as the headline. It rises with confusion as easily as with interest.
- Changing the definition of active mid-trend. Counting any event one month and a core action the next resets every comparison.
- Counting background activity. Syncs, push receipts and system events make everyone look active.
- Mixing audiences. A month with many new users will show lower engagement than a month of long-time users; compare like with like.
- Treating engagement as the goal. Pushing the number directly, for example with more notifications, can raise it while value falls. Use it to find where to look.
User engagement in TapCub
The month-1 numbers above can be read in TapCub like this.
- Define the key action and the active rule. Send the key action as an event. A server-side event counts as activity only when set to Counts as active in All-platform analytics › Data › Events & properties; otherwise a server-only project shows no active users.
- Read breadth and depth in one chart. In All-platform analytics › Behavior › Behavior analysis, the distribution model has a participation matrix: each row is a custom event, breadth is the percent of active people who fired it, and the second axis is the average number of times per person. Up to 30 events are shown; events whose names start with a double underscore are skipped.
- Check the shape. The same model can show how many times each person fired one event, or how many days they were active, as a histogram with the mean and the median.
- Add frequency and session-level views. For apps, App analytics › Users › Users shows daily, weekly and monthly active users and App analytics › Users › Engagement shows distributions of duration, frequency, interval, screens and starts. For websites, the engaged rate in Web analytics › Reports counts a session as engaged at 10 seconds, 2 page views, or a key event.
- Compare groups. Build a segment in Segments & tags, such as users who signed up this month, and open it with Analyze to see the same measures for that group only.
TapCub measures behaviour. It does not run surveys, so satisfaction and intent have to come from somewhere else.
See also: Monthly Active Users (MAU): Definition, Formula and Pitfalls.
See also: User Engagement Platform: Capabilities by Use Case.
Sources
- O'Brien, H. L., and Toms, E. G. (2008). What is user engagement? A conceptual framework for defining user engagement with technology. Journal of the American Society for Information Science and Technology, 59(6), 938–955.
- Google, Analytics Help: Engagement rate and bounce rate: definition of an engaged session (longer than 10 seconds, a key event, or at least two page or screen views).
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Data Analyst
Explores traffic sources, conversion funnels, and retention, using clear metric definitions and concrete examples to explain what the numbers mean.