Explain what each main GA4 metric measures and where it falls short.
Every number GA4 reports has a definition and a blind spot. You don't need to memorise Google's GA4 help centre. You do need to understand the few numbers you report well enough to explain them to someone who has never opened GA4, and to know when not to trust them.
Farah runs an online shop and has reported her GA4 numbers for months. Her partner looked at the GA4 overview and asked about the 3,100 users the shop had last month (an invented figure, like the rest of Farah's numbers), since it sold to about 200 people and had maybe 2,000 on the mailing list. Her partner wanted to know if 3,100 different people had really visited. Farah didn't know, because she had been reporting the number for months without being sure what it counted.
A GA4 user is usually a browser on a device that GA4 has given an identifier to. It is not a person. On someone's first visit, GA4 stores an identifier in a cookie in their browser. Later visits from the same browser carry the same identifier, so GA4 counts them as the same user.
That rule produces some odd counts. A customer who browses Farah's shop on her phone during an MRT ride and buys on her laptop at home counts as two users. One customer using two different browsers on the same phone also counts as two users. Someone who clears their cookies becomes a new user. A family sharing one tablet counts as one user.
So the user count is an estimate of people, and it is usually too high. GA4 can match devices more accurately if the site sends its own user ID when people log in, or through signals from Google accounts. Most small sites don't have that set up, and even with it the match is only partial.
The honest answer to Farah's partner is that about 3,100 browsers visited, so fewer than 3,100 people did, and nobody can say exactly how many fewer. The number is still useful for comparing one month with another, as long as the counting method doesn't change.
A session is one visit: a period of activity from one user that starts when they arrive and ends after a stretch of inactivity. One user can have several sessions in a month, so sessions are usually higher than users.
A session counts even if the visitor leaves almost at once. Someone who taps an Instagram link, sees the page start to load and swipes back has still created a session. GA4 also counts engaged sessions so you can separate visits that mean something from visits that don't.
Google's definition is that a session is engaged if it lasted longer than a set time, or included a key event, or had two or more page views. Any one of the three is enough. Google's GA4 help centre gives the default time as 10 seconds. You can change it in the data stream settings, but most businesses leave it as it is.
Engaged sessions are the better number to watch when you compare channels. If one channel brings 1,000 sessions of people who leave at once and another brings 400 sessions of people who look around (example figures), the 400 are worth more to you, even though the total is smaller.
Engagement rate is engaged sessions divided by all sessions. To work it out, take the number of engaged sessions and divide it by the total number of sessions. Say your site had 2,000 sessions last month and 1,240 of them were engaged, both example figures. GA4 does the same calculation: 1,240 divided by 2,000 is 62 percent.
GA4 made engagement rate its main measure of visit quality. Older Google Analytics led with bounce rate, which then meant the share of visits that saw only one page. That old measure penalised pages where people found what they needed and left, such as a contact page or a blog post that answered the question.
GA4 still shows a bounce rate, but it now means the share of sessions that were not engaged, so it is the opposite of engagement rate. If you compare a GA4 bounce rate with a bounce rate from old Google Analytics, you are comparing two different measures that happen to share a name.
Engagement rate is most useful for comparisons: channel against channel, landing page against landing page, this month against last month. There is no universal good engagement rate. A product page and a blog post will never have the same rate, and neither will two different businesses.
Visitors who decline analytics cookies on the cookie banner may not be recorded at all, or may be recorded only partly. With Google's consent mode, GA4 may fill some of that gap with modelled data, which is an estimate based on the visitors who accepted cookies. Ad blockers and some browser privacy settings also stop the GA4 tag from loading.
How many visitors go missing varies by site, audience and banner design, so there is no correction factor you can apply. GA4 totals are estimates and should not be treated as a register. Count real sales and enquiries in your shop platform, booking system or CRM, not in GA4. Lesson 5.2 comes back to this and covers what to do when numbers from different tools disagree.
Farah now reports the numbers to her partner in that order. She starts with real orders from the shop platform. Then she uses GA4 to explain where the buyers came from, and then how the visits behaved. If you run a booking system or a CRM, start with the numbers in it in the same way.
Writing that kind of plain explanation for someone who has never opened GA4 is a good test of whether you understand the metrics yourself, and it is what you will do next.
Write one paragraph explaining users, sessions and engagement rate to a colleague or client who has never opened GA4.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).