Sessions and Users at a glance
In GA4Google's event-based analytics platform., a session starts when the tag records activity after a period of inactivity and ends when the inactivity timeout passes. Each session carries the source, medium and campaign it began with, which is why acquisition reports group by session. The session glossary entry covers how a session is defined and why counts differ between GA4, its BigQuery export and ad platforms.
A user is the identifier GA4 uses to recognise the same visitor across sessions. Which visits count as one user depends on the property's reporting identity setting, and the Users column in most standard reports is active users rather than everyone who sent an event. The GA4 reports and dimensions guide explains the user and session scopes, the different user metrics and the engagement definitions in detail; this comparison does not repeat them. For sites where people sign in, the guide to User-ID in GA4 explains how a site's own identifier changes which visits count as one user.
The two counts describe the same traffic at different grains. One person who visits three times is three sessions and one user. Every rate built on them inherits that difference: engagement rate is engaged sessions divided by sessions, the session key event rate divides by sessions, and the user key event rate divides by active users. The choice of denominator is the real decision, and it is the one teams most often leave unstated.
Compare table
Sessions vs Users
| Row label | Sessions | Users |
|---|---|---|
| What is counted | Visits: groups of activity separated by inactivity | Identifiers: a browser, device or signed-in person, as GA4 can tell |
| Scope | Session scope, carrying the source and campaign the visit began with | User scope, carrying attributes such as first-user source |
| How a new one starts | Activity after the inactivity timeout | A new identifier appears, blended according to the reporting identity setting |
| Effect of consent and cookies | Declined consent removes visits from the observed count, and reports may add modelled data | Declined consent removes people too, and cleared cookies turn one returning person into several users |
| Effect of devices | Each device produces its own sessions | One person on two devices is two users unless signed-in or modelled data links them |
| Typical denominator | Engagement rate, bounce rate and session key event rate | User key event rate, reach of a campaign, outcomes per person |
| Where totals differ | Usually higher than users for any period longer than a day | Not additive across days or dimensions, because each report deduplicates |
| Best use | Judging traffic sources, landing pages and visit quality | Judging audience size and how many people eventually acted |
The rows that matter most are the typical denominator and where totals differ. Most confusion in GA4 reporting is a rate built on one count being compared with a rate built on the other, or a user total being summed when it cannot be.
When to choose Sessions
Use sessions when the question is about visits. A campaign, a source or a landing page brings visits, and the quality of those visits is best judged per visit: how many were engaged, how many included a key event, how many ended on the first page. Engagement RateShare of sessions that were engaged; on social, interactions relative to reach. is defined on sessions for this reason, and the session key event rate is what most teams mean when they say conversion rate.
Sessions are also the more stable unit for comparing traffic over time. They depend less on identity than users do: when a browser shortens how long cookies last, a returning person starts appearing as a new user, while each of their visits is still one session. For a weekly report on whether paid traffic is landing well, sessions and the rates built on them are the right basis.
Where sessions mislead is in describing people. A customer who visits five times before buying produces one conversion and five sessions, so a session-based rate understates how many of the audience bought, and a channel that brings people back repeatedly can look weak on sessions while it is doing useful work.
When to choose Users
Use users when the question is about people. How many distinct people did a campaign bring, how many came back, and what share of them eventually converted are user questions, and dividing by sessions answers a different one. The user key event rate suits products with long consideration, where several visits before a purchase are normal and a per-visit rate would make a healthy journey look weak.
Users are also the unit for cohort and retention questions: how many of this month's new users returned, and how their behaviour developed. GA4 explorations handle those questions better than standard reports, because they let the analyst fix the user population and follow it.
The cost is that users are the less certain count. An identifier is not a person, and cleared cookies, multiple devices and consent choices all split one person into several users. User totals also do not add up across days or dimensions, because each report deduplicates within its own range. For the same reason, a new user is not a new customer: the guide to new and returning customers explains how to take customer status from the order system instead.
A garden supplier settles on one conversion rate
Corrievale Garden Supply, an online garden retailer, reports a conversion rate of 1.5% in its weekly marketing pack and 3.6% in the ecommerce team's dashboard, for the same week and the same 540 orders. The marketing pack divides orders by 36,000 sessions; the dashboard divides by 15,000 active users. Both are correct, and the board had been told both. The analytics lead agrees one rule: campaign and landing page reports use the session rate, labelled "per session", and the quarterly customer report uses the user rate, labelled "per user". A footnote on every rate names its denominator, and the two figures stop being compared with each other.
Decision table
| Situation | Use | Why |
|---|---|---|
| Comparing traffic sources or campaigns on quality | Sessions | Each source brings visits, and visit quality is judged per visit |
| Testing a landing page | Sessions, or the unit the test randomises on | The page is experienced once per visit |
| Reporting how many people a campaign reached on the site | Users | The question is about people, not visits |
| Products with long consideration and repeat visits | Users for the conversion rate | A per-session rate understates how many people eventually buy |
| Cohort and retention analysis | Users | The population has to be fixed and followed over time |
| Summing figures across days or channels | Sessions | User counts are deduplicated per report and cannot be added |
Choosing the denominator for a rate
Every rate in a GA4 report is a count divided by sessions or by users, and the choice changes the answer. Decide it once per question, write it into the measurement plan and print it beside the figure. Visit-level questions, such as which campaign or landing page performs, take sessions. People-level questions, such as how many of the audience converted or returned, take users. A single key performance indicator should use one denominator and keep it, because switching between the two looks exactly like a change in performance.
Explaining the mismatch to stakeholders is mostly a matter of naming. "Conversion rate" with no qualifier invites two people to read the same number differently; "orders per session" and "orders per active user" do not. When someone asks why the figures do not add up, the honest answer is that sessions add up across days and users do not, and that both are estimates rather than exact counts. The GA4 common mistakes guide covers the misreadings that follow when this is forgotten, and server-side collection, compared in the client-side versus server-side tagging comparison, changes how reliably either count is captured rather than what each one means.
Common mistakes
- Reporting "conversion rate" without its denominator, so a session rate and a user rate are compared as if they were the same figure.
- Summing daily users into a weekly or monthly total, which counts returning people more than once.
- Judging a long-consideration product on the session rate alone, which makes a normal multi-visit journey look weak.
- Treating users as people, when cleared cookies, devices and consent choices split one person into several.
- Comparing the old Google Analytics session counts with GA4's, which define a session differently.
- Switching the key performance indicator between the two rates from one report to the next.
- Expecting reports and the export to match exactly, when report counts are estimated and the export is not.
