Session

A session is one visit to your website. Analytics does not observe visits directly, so it groups a run of separate events into a session using a timer and a set of rules. The result is a useful approximation and not a count of people arriving at a door.

In short

• A session is a group of events, not an observed visit.
• It ends after a period of inactivity, usually thirty minutes.
• Ordinary behaviour splits one real visit into several sessions.
• It is a comparison over time, not an estimate of how many people came.

How a session is put together

A browser loading a page fires an event. Scrolling fires another. Clicking, watching, submitting and buying fire more. What analytics receives is a stream of these, each stamped with a time and an identifier for the browser.

A session is what you get when that stream is cut into pieces. The rule in Google Analytics 4 is inactivity: if thirty minutes pass with no event from that browser, the session is closed, and the next event opens a new one. The timer resets with every event, so the length of a session has no ceiling.

This is worth stating plainly because the word invites a mental picture that does not match. Nothing in the system knows a person arrived or left. It knows that events stopped for a while.

Sessions, users and page views

Three numbers get confused constantly, and they answer different questions.

• Users. How many distinct browsers analytics has identified. Not people. One person with a phone and a laptop is usually two.
• Sessions. How many visits those browsers made. One user can produce many.
• Page views. How many pages were loaded across all of them. One session can produce many.

The relationship is always the same shape. Page views are the largest, sessions sit below them, and users sit below that. When a report shows sessions and users almost identical, it usually means very few people are coming back, which is a finding rather than a rounding detail.

Where the count goes wrong

Several ordinary things create a session that does not correspond to a visit.

• The timeout. Somebody reads half an article, takes a call, and returns forty minutes later. That is two sessions and one reader.
• Midnight. A visit that spans the end of the day is cut in two by the reporting calendar.
• A campaign parameter arriving mid visit. A visitor already on the site who clicks one of your adverts elsewhere and returns with tracking parameters attached starts a new session, even though they never left.
• Bots and preview tools. Filtered imperfectly, and more visible on small sites where a handful of automated visits moves the total.
• Consent. A visitor who declines tracking may produce no session at all, which means the total is a count of measurable visits and not of visits.

None of this makes the number useless. It makes it a relative measure. A session total repays comparison against last month and against the same month last year. It should not be read as a headcount, and it should never be the number a website is judged on by itself.

What an engaged session changes

Google Analytics 4 added a second definition on top of the first. An engaged session is one that ran longer than ten seconds, or triggered a conversion event, or included more than one page view.

The reason is that a plain session count flattens a real difference. A mistaken click that leaves after two seconds and a careful read lasting four minutes both count as one session. Engagement draws a line somewhere in between so the two can be separated.

It is a convention rather than a measurement. Ten seconds is a setting, not a threshold of human interest, and a visitor who found the phone number in three seconds and left satisfied is counted as unengaged. Treat it as a rough filter that makes trends easier to read, and read it beside bounce rate, which in this system is its exact opposite.

Questions a session can and cannot answer

Because the unit is a grouping rule, it answers some questions well and others not at all. Keeping the two apart saves a great deal of pointless reporting.

It answers these reasonably.

• Is traffic rising or falling? Compared against itself over a long enough period, the trend is trustworthy even though the absolute number is not.
• Which channel is growing? Sessions split by source show where the change came from, which is usually the more useful half of the answer.
• Which pages start visits? The landing page of a session is a fact about how people arrive.

It answers these badly or not at all.

• How many people visited? It cannot. Devices, browsers and declined consent all sit between the count and a headcount.
• How interested were they? A session has no opinion. Engagement time and what people did are the measures for that.
• Did the marketing work? Only inquiries, orders and revenue answer that, and a session total can rise while all three fall.

The habit to form is to put a session count next to something that pays for the site instead of reporting it alone. Traffic and conversion rate read together tell a story. Traffic on its own tells a story that is sometimes true.

Words you will hear

• Session timeout. The period of inactivity that closes a session.
• Engaged session. A session that passed the engagement threshold.
• Session source. Where the visit was attributed to, decided at the point the session opened.
• Returning user. A browser seen in an earlier session.
• Average engagement time. How long the site was actually in the foreground, which is a better length measure than session duration.

Sessions are the unit almost every other metric is built on, so it pays to know what the unit is made of. Once you accept that it is a grouping rule rather than an observation, the reports stop looking precise and start being useful, which is the more honest of the two.