What is engagement?

Engagement is the average number of events per session: how much happens in a typical visit.

If your product recorded 4,000 events across 500 visits in the last 30 days, engagement is 8. The average visit involved eight tracked actions.

Unlike every other metric on the overview, this one counts visits rather than people. Anonymous visits are included, because it measures activity volume rather than who did it.

What it tells you that the other metrics don't

Repeat rate tells you people came back. Engagement tells you whether anything happened when they did.

The two split apart more often than founders expect. A product with strong return numbers and an engagement of 2 is being opened and abandoned. People have a reason to check in and no reason to stay.

Read it as depth per visit. High and steady means visits have a purpose. Falling while sessions rise means you're getting more visits that do less.

What good looks like

There is no benchmark, and any number you see quoted elsewhere is meaningless for your product. Engagement depends entirely on how many events you chose to track.

A team that tracks every click will show 30. A team that tracks five meaningful actions will show 3. Neither is better. Comparing your number to another company's is comparing instrumentation, not products.

So the number only means something against itself:

• Compare it week to week. The trend is the signal.
• Compare it before and after a change. A redesign that raises engagement did something.
• Ignore it entirely for a fortnight after you add tracking. New events will move the average for reasons that have nothing to do with users.

How to use it well

Track outcomes, not clicks. If your events are mostly page views, the average measures navigation and rises when people get lost. Events named after things users accomplished make the number readable.

Watch it alongside sessions per user. Both climbing means people come more often and do more each time. Sessions up with engagement down usually means something is broken or people are searching for a feature they cannot find.

Use it as a before-and-after for one change at a time. Ship, wait a week, compare against the week before you shipped.