GA4 Website Metrics: What Actually Deserves Tracking

The bounce rate sitting in your Google Analytics today does not measure what it measured in Universal Analytics. The label stayed, the math changed completely. That is why so many reports show a drop in bounce rate that nobody could ever explain: the site was not fixed, the ruler was swapped.

Anyone running a website lives on comparisons. This month against last, one landing page against another, paid against organic. Comparison only means something when the ruler holds still, and GA4 quietly retired several old rulers while rewriting the rules for the ones that survived.

So it is worth sorting out what changed, what deserves a permanent slot in the report, and what you can skip without guilt.

Website metrics in GA4 revolve around engaged sessions. A session counts as engaged when it lasts longer than 10 seconds, fires at least one key event, or racks up two or more page or screen views. Engagement rate is the share of sessions that qualify; bounce rate is simply the leftovers, meaning 100% minus the engagement rate.

Why GA4 bounce rate cannot be compared to Universal Analytics

In Universal Analytics, a bounce was a session with a single hit. Someone opened your article, read it for eight minutes, closed the tab perfectly satisfied and left: bounce. That number punished exactly the content that held attention, which is why every serious analyst worked with adjusted bounce rate, scroll events, time on page, some workaround.

In GA4, those same eight minutes clear the 10-second bar and become an engaged session. The new ruler fixed the old injustice, then overshot in the other direction. Ten seconds is almost nothing. Someone who lands on a page, realises they are in the wrong place and takes twelve seconds to find the back button enters your statistics as an engaged visit.

Two practical consequences follow. First, never compare bounce rate before and after the migration, because they are different metrics wearing the same label. Second, engagement rate on its own says far less than it appears to.

How does GA4 decide a session was engaged?

There are three criteria and only one has to be met, per the Google Analytics documentation: the session runs longer than 10 seconds, or it triggers at least one key event, or it logs at least two page or screen views.

Two details usually slip by. Automatic events such as session_start, first_visit and first_open are excluded from the calculation, even if you mark them as key events. And the 10-second threshold is not carved in stone: in the Google tag settings, the engagement timer can be stretched up to 60 seconds.

That is probably the cheapest fix available in GA4 right now. Raising the cutoff from 10 to 30 seconds will not improve your site, but it makes engagement rate go back to separating people who read from people who passed through. The cost is losing comparability with history before the change, so do it once, write down the date and leave it alone.

Which website metrics to track in GA4

A short list of six covers most situations. All of them sit in the standard reports, with no exploration and no extra code.

Metric	What it answers	Where it misleads
Active users	How many distinct people showed up	Cookie blocking and device switching inflate the count
Engaged sessions per user	Whether people come back with intent	It disappears into the average when traffic is very seasonal
Engagement rate	What share of visits cleared the minimum bar	It rises on its own when the site slows down and people wait
Average engagement time	How long the tab was actually in the foreground	Long pages pull the average up without producing results
Key events	How many valuable actions happened	Marking too many events as key drains the metric of meaning
Session conversion rate	What share of visits turned into a valuable action	Its meaning shifts every time you change the tagging

The order matters. Volume first, quality second, outcome last. Start the report at the outcome and you end up explaining a conversion swing that was really a traffic swing. If your team is still assembling its indicator set, it helps to see how these numbers slot into digital marketing KPIs by funnel stage.

Two identical engagement rates, two opposite sites

What follows is a worked example built to illustrate the point, not a study. Picture two service websites, both with 10,000 sessions in a month and 62% engagement rate. A tie, apparently.

Indicator	Site A	Site B
Sessions	10,000	10,000
Engaged sessions	6,200	6,200
Average engagement time	14 seconds	1m48s
Views per session	1.2	3.1
Key events	40	310
Session conversion rate	0.4%	3.1%

Site A is scraping past the 10-second criterion. Most of its sessions probably cross the line by a hair, which explains 1.2 pages per session and almost no events. Site B has people genuinely browsing. B converts nearly eight times better than A, and engagement rate never showed it.

The lesson is simple and slightly dull: engagement rate is a triage metric, not a diagnostic one. It tells you something moved. What explains the movement is average time and key events, cut by channel and by landing page.

The limits that get in the way

Three constraints surface early in any operation that takes analysis seriously.

Data retention. On standard properties, event data stays available for 2 or 14 months, depending on the property's retention setting. Longer windows exist only in Analytics 360. If your property sits at 2 months, year-over-year comparison in explorations simply will not run. Worse, changing the setting is not retroactive, so data that already expired is gone. If you have never checked this, check it today.

Cardinality. Dimensions with too many distinct values, page URL on a large site being the classic case, get bundled into a row labelled "(other)". The totals stay correct, the detail vanishes. Anyone with a large catalogue lives with this every month.

Inconsistent tagging. No website metric survives sloppy UTMs. One channel spelled three ways becomes three channels, and source comparison falls apart before it begins. When that is the case, the problem is not GA4, it is UTM standardisation.

The first two limits push every growing company toward the same conclusion: GA4 is excellent at showing what happened on the site and poor at holding history and joining it to sales. Exporting to a database or consolidating with other sources stops being a luxury around the moment someone asks for the 24-month series.

Building the weekly read

Four questions, in order, cover almost everything a marketing meeting needs to know about the site.

1. Did volume change? Active users and sessions, always against equivalent periods.
2. Did quality change? Engagement rate and average time, broken out by channel.
3. Where did they come in? Landing pages ranked by sessions, with engagement alongside.
4. What came of it? Key events and conversion rate, crossed with the landing page.

When one of those four answers looks out of line, you know where to dig. It also pays to log configuration changes in a plain file, with dates: engagement timer adjusted, new event marked as key, consent banner replaced. Without that log, six months from now someone will spend an afternoon explaining a step in the chart that you created yourself. The same logic applies to conversion rate by channel, which also shifts its base every time the tagging moves.

Frequently asked questions

What is a good engagement rate in GA4?

There is no universal number, and be sceptical of anyone publishing one. The 10-second criterion naturally puts content sites above transactional ones. What works is comparing your rate to your own average per channel, in the same period of the year.

Is GA4 bounce rate the same as Universal Analytics bounce rate?

No. In Universal Analytics, a bounce was a single-hit session. In GA4, it is the complement of engagement rate, meaning the share of sessions that met none of the three criteria. Placing the two series side by side produces a wrong conclusion.

How do you increase website engagement rate?

Load speed, message match between the ad and the page, and a visible next step above the fold solve most cases. Before any of that, check whether irrelevant traffic is coming in through broad keywords or automatic network placements.

Does GA4 still show average session duration?

The day-to-day equivalent is average engagement time, which counts only the period when the tab was in the foreground. It is more honest than the old session duration, which counted forgotten open tabs.

Can the 10-second threshold be changed?

Yes. In the Google tag settings, the engagement timer can be adjusted up to 60 seconds. The change applies going forward and does not alter history already collected.

The report is not the work

Most of the effort in site analysis still goes into assembling the report rather than deciding what to do with it. Export, paste into a spreadsheet, format, compare with last month, write the summary. By the end of the afternoon what remains is a tidy table and no decision made.

That is where asking your data in plain language changes the pace. With the sources connected in one place, Sherlok answers things like "did organic engagement rate drop this week?" and brings the landing page breakdown with it, without you opening a single exploration. The analysis is still yours. The manual work of getting to it is what disappears.