Google Analytics 4 for Publishers: The Metrics That Actually Matter
GA4 counts a ten-second skim as engagement, so publishers need sharper measures of reading, loyalty and revenue than the defaults provide.
A reader who opens a story, skims the headline and leaves after eleven seconds counts as an engaged visitor in Google Analytics 4.
In Universal Analytics, the same visit was recorded as a bounce. A publisher can watch its engagement rate climb after migrating without a single reader behaving differently, and the dashboard will still look like good news.
Trending Now!!:
The GA4 metrics worth a publisher’s attention are engaged sessions and engagement rate, average engagement time, views per active user, the share of returning users, and key events tied to a real business outcome such as a newsletter signup or a subscription.
Raw sessions, total users and the default scroll event describe traffic, not audience value.
Why Publisher Analytics Needs Its Own Metric List
GA4 ships with a reporting model built around events and a generic idea of engagement. It does not know whether a site sells ad impressions, subscriptions or both, and it does not know that a 1,500-word feature and a 90-word news brief deserve different yardsticks. The metric list has to be assembled by the publisher, and it helps to sort every candidate into one of three jobs.
Attention measures whether anyone actually read the page. Loyalty measures whether readers come back. Outcome measures whether reading turned into revenue or a relationship: a sign-up, a registration, a paid subscription. A metric that cannot be placed under one of those headings is probably decoration.
Most of the confusion starts with attention, because the engagement metrics look familiar but are computed differently from their Universal Analytics ancestors.
Engaged Sessions and Engagement Rate
An engaged session is one that lasts longer than 10 seconds, includes a key event, or contains at least two page or screen views. Engagement rate is engaged sessions divided by total sessions. Bounce rate is its mirror image: the percentage of sessions that were not engaged.
The definition carries a built-in generosity for publishers. Ten seconds is enough to read a headline and a standfirst, and a second pageview can be an accidental tap on a related link.
Engagement rate therefore works better as a comparison tool than as a grade. Set against traffic sources, landing pages and templates, it exposes real differences.
A story that holds search readers but loses social readers is probably being promoted with a headline that promises more than the article delivers.
Raising the Timer Without Breaking Comparisons
The engaged-session timer sits in the Google tag settings and can be moved from the default 10 seconds up to 60. Publishers of long-form work can raise it to 30 seconds or more so that a skimmed headline no longer counts.
The change applies from the moment it is saved and does not restate history, so the date should be recorded, and engagement rate before and after should never be plotted as one continuous trend.
A hypothetical example shows the effect. Suppose a site records 100,000 sessions, of which 58,000 are engaged, a 58 percent engagement rate.
Of those, 9,000 ended between 10 and 30 seconds with a single pageview and no key event. Raising the timer to 30 seconds would reclassify them, dropping the rate to 49 percent without any change in readership.
Average Engagement Time: Reading Time, Not Visit Length
Average engagement time is the total time the site was in focus in the browser, or the app in the foreground, divided by the number of active users. A tab left open in the background adds nothing.
Universal Analytics, by contrast, could not measure time on the final page of a session, so a reader who finished one long article and left registered as zero seconds. The newer metric repairs that blind spot, which is why publishers see longer times after migrating.
Per-page engagement time sits in the Pages and Screens report and is the nearest thing GA4 offers to a reading-time measurement. Two cautions apply.
The figure is an average across every visitor to the page, so a small group of devoted readers and a large group of glancers blend into a number that describes neither. The headline figure is also divided by active users rather than sessions, so it will not match per-session figures elsewhere.
A more useful version is a read-through ratio: average engagement time on a page divided by the time a reader would need to finish it. Adult silent reading speed is commonly put at roughly 200 to 250 words a minute, so a hypothetical 1,200-word article needs about five minutes, or 300 seconds, at 240 words a minute.
An average engagement time of 60 seconds on that page gives a ratio of 20 percent. Compared across articles of similar length, the ratio reveals which stories are read and which are merely opened. It is a screening tool, not a verdict, because galleries, video pages and live blogs break the arithmetic.
Active Users, Returning Users and Stickiness
Where GA4 reports show Users, they are usually showing active users, meaning people with at least one engaged session. Total users remains available as a separate metric and will be higher.
Publishers moving from Universal Analytics often find lower user counts and suspect broken tracking. The definitions explain much of the gap.
User counts also depend on cookies, devices and consent choices. One person reading on a phone and a laptop appears as two users unless the site passes a User-ID for logged-in readers. Treat the number as a good estimate, not a headcount.
Loyalty shows up in the new versus returning split and in user stickiness, which GA4 calculates automatically as the ratios of daily to monthly active users, weekly to monthly, and daily to weekly.
Higher ratios suggest stronger engagement and retention. No universal benchmark exists, since a daily news brand and a monthly long-read magazine have different natural rhythms. The useful comparison is the site against its own history.
The sharper cut is returning-user share by acquisition channel. A site whose readers arrive almost entirely from search and rarely come back is renting its audience. One whose newsletter and direct readers return weekly owns a meaningful part of it.
Views and Visit Depth
Views, formerly pageviews, feed two metrics worth keeping: views per active user and views per session. Ad-funded publishers should watch views per active user in particular, as the closest GA4 proxy for inventory per reader. Subscription publishers can read depth as a sign that readers find a second and third story worth opening.
The number is easy to corrupt. Auto-refreshing pages and infinite-scroll setups that rewrite the URL as each new article loads can record a view the reader never chose to open.
Duplicate installation is worse. A WordPress plugin and a hard-coded snippet can both fire a page view, giving every session two views and therefore an engaged status. The telltale sign is a bounce rate near zero, which no real publisher achieves.
Key Events: Defining Success Narrowly
In March 2024, Google renamed GA4 conversions as key events, reserving the word conversions for outcomes measured in Google Ads.
Marking an event as a key event takes effect from that moment and leaves historical data untouched. A publisher who later decides that a newsletter signup deserves key event status will not see the past retroactively recolored.
A short list serves publishers best:
- Newsletter signup
- Account registration
- Subscription or membership start
- Click on an affiliate or partner link, if that is a revenue line
- A completed read, defined by a custom event
Restraint matters for a reason tied to the engagement metrics above. A key event makes its session engaged by definition, so marking a high-frequency event such as scroll as a key event pushes engagement rate toward 100 percent and strips it of meaning.
Key events work best as rates, such as signups per 1,000 engaged sessions by landing page, which allow fair comparison between a high-traffic story and a quiet one.
Scroll Depth is Not What It Looks Like
The scroll event in enhanced measurement fires once per page, the first time a visitor reaches 90 percent vertical depth. It records nothing at 25, 50 or 75 percent, so it shows who reached the bottom and says nothing about where everyone else left.
Page height makes the signal less reliable still. Related stories, comments and footers sit below the article text, so a reader who finishes the last paragraph may be at 60 percent of the page and never trigger the event. On very short pages, the 90 percent line can be visible almost at once.
The fix is custom tracking in Google Tag Manager. Fire events at 25, 50, 75 and 100 percent, or better, use a visibility trigger on an element placed at the end of the article body. Disable the enhanced measurement scroll toggle to avoid duplicate events, and register the event parameter as a custom dimension so it appears in explorations and Looker Studio.
The result shows how many readers reached each quarter of the story, which is the data editors need when deciding where to put a newsletter prompt or cut a flabby middle.
Acquisition Reports That Mislead
GA4 offers two acquisition views that answer different questions. User acquisition credits the channel that first brought a reader, while traffic acquisition credits the channel of each session.
Daily channel analysis belongs in traffic acquisition. User acquisition answers a better strategic question: where do loyal readers originate?
The direct channel deserves suspicion. Visits arriving without referrer data land there, including many from messaging apps, email clients and some social apps, and any newsletter link without UTM parameters. Tagging every newsletter and partner link shrinks the bucket and reveals how much audience the newsletter really drives.
Google Search Console remains the source for queries, impressions and Discover performance. GA4 begins at the click, so the two tools answer different halves of the same question and should be read together.
Limits That Quietly Distort Publisher Data
Retention
Event-level data retention defaults to two months and can be raised to 14 on standard properties. The setting affects explorations and funnel reports, not standard aggregated reports, so the headline dashboards keep their history while custom analysis silently loses everything older than the window.
Raising it to 14 months is a quick change in Data Settings, and it applies to data already collected and not yet deleted. Year-over-year analysis in explorations remains impossible beyond 14 months, which is the strongest argument for exporting data elsewhere.
Thresholding and Sampling
GA4 withholds rows with small user counts when demographic or Google signals data is involved, and the thresholds cannot be adjusted.
Switching to device-based reporting identity or dropping demographic dimensions reduces the effect. Page-level analysis of niche articles combined with age or interest data is the most likely place to hit it.
Explorations can also sample data when a query exceeds the property’s event limit. Looker Studio does not display the warning indicators that the GA4 interface does, so a clean-looking dashboard can hide withheld or sampled rows.
BigQuery Export
The free BigQuery export is the standard answer to retention and sampling limits, but standard properties are capped at 1 million events per day for the daily export. Every pageview generates more than one event, so a publisher serving a few hundred thousand pageviews a day can cross that line.
A property that consistently exceeds it has its daily export paused, and earlier days are not reprocessed. Filtering out low-value events keeps the volume down, and the streaming export carries no event limit, though BigQuery storage and streaming charges apply.
A property that grows into Large status also has its event-level retention cut to two months, so volume deserves monitoring.
A Three-Row Scorecard
The metrics above fit on one page. The first row covers attention: engagement rate by channel, average engagement time on the top pages, and the read-through ratio for text-led stories.
The second covers loyalty: the returning-user share by channel and user stickiness. The third covers outcome: key events normalized per 1,000 engaged sessions.
A hypothetical 40 newsletter signups across 8,000 engaged sessions is 5 per 1,000, a number that can be compared between landing pages, channels and months.
Sessions, total users, average session duration and demographic breakdowns can stay available for diagnosis but do not belong on the page people read each week. Bounce rate adds nothing that engagement rate does not already say.
The test for any figure on the scorecard is whether a different number would change what an editor, a product owner or an ad sales lead does on Monday. A metric that fails the test can stay in GA4, ready for diagnosis, and off the page that people actually read.
What People Ask


