There is a pattern most visitors follow when they land on a webpage. It starts with skimming, reading the headings, bold text, and scanning images, before they read a paragraph properly. All of this takes a split second, but it helps the visitor decide whether to continue or move on.
When a visitor decides to stay, they often finish a paragraph, scroll to the bottom, and pause. Their cursor drifts upward. Maybe they don’t move for a while. The tab is still open, but they’ve already gone to check their phone.
Sometimes your analytics software marks this as a bounce; other times, it’s counted as a single-page session or a lost pageview.
Understanding why a visitor chooses not to engage further helps publishers improve, either by removing friction or by offering what the visitor actually wanted next. Let’s start with the 4 signals that lead to a bounce.
1. Scroll Completion
It’s simple: the visitor reaches the bottom of an article.
Whatever brought them to the page (a search result, a social link, a newsletter) has been delivered. Now they're at a decision point.
Scroll completion is one of the clearest behavioral markers available to publishers. It doesn't mean the visitor loved the content. It means they consumed it. The session is at its natural inflection point (the equivalent of closing a book) and without a clear next step, the default action becomes the close tab button.
For publishers, bounce rates typically range between 50–70%, and content-heavy sites often see rates between 70–90%, since visitors frequently arrive through search, read one article, and leave. For many of those exits, scroll completion was the last detectable moment before the session ended.
The window here is narrow, but it's real. A visitor who has just finished reading is still on the page. That's the last moment where re-engagement is possible without disruption.
2. Idle Time
A visitor stops scrolling. The page is still open. But nothing is happening.
Idle time is a deceptively ambiguous signal. The session is technically active, but the visitor's attention has already moved somewhere else. Maybe another tab, a notification, a distraction in the room. When they return to the browser, they'll often close the tab rather than resume where they left off.
The 2025 benchmark for publisher attention time sits at just 45 seconds, which means a visitor starts showing disengagement signals well before the first minute. The idle state is often where that shows up first.
For ad-ops teams, idle sessions are a particular problem. A visitor who stopped scrolling mid-page still counts in session data, but their actual engagement with ad inventory has already gone.
CPM projections built on raw session counts can obscure how much of that "time on site" is actually idle time, a blind spot that Average Revenue Per Visitor (ARPV) as a metric helps expose more clearly.
The operative question isn't whether someone's tab is open. It's whether their attention is.
3. Cursor Exit
On the desktop, the cursor tells you a lot.
When a visitor's mouse moves toward the top of the browser window, they're heading for the address bar, the close button, or the back button.
The movement is directional and fast, distinct from casual cursor drift during reading. This is the most time-sensitive of the four signals. Once a cursor exits the viewport, the session is usually over within seconds.
Research has shown that cursor movements correlate with eye gaze, making them a reliable indicator of where attention is actually directed.
The tricky part for publishers is that this signal only exists on desktop. Mobile visitors, who account for the majority of publisher traffic, don't leave a cursor trail. For mobile sessions, other signals, particularly idle time and scroll behavior, carry more weight.
4. Attention Decay
This one is harder to pin to a single moment, which is exactly what makes it the most important signal to understand.
Attention decay is cumulative. A visitor arrives, starts reading, and somewhere in the first half of the article, something slips. The pace of their scrolling slows. They stop for longer between paragraphs. They hover without really reading. The session continues, but the quality of engagement is declining.
Chartbeat data across millions of pages puts the average time on page at around 52 seconds, with most users deciding whether to continue within the first 10–15 seconds. But the visitors who make it past that initial filter don't necessarily stay engaged. Attention decay happens throughout a session, not just at the start.
For publishers, this matters because decayed attention still generates pageview data, still loads ads, and still appears in analytics as an engaged session. But the commercial value of that attention has already eroded.
A visitor in a state of attention decay is far less likely to click a recommendation, notice a subscription prompt, or carry that engagement into a second pageview, all of which directly affects your ability to grow monetizable inventory from the traffic you already have.
How These Signals Relate to Each Other
|
Signal |
What it looks like |
What it tells you |
Response Window |
|
Scroll completion |
visitor reaches the bottom of the page |
Content consumed; no clear next step |
Medium; a few seconds to act |
|
Idle time |
No scroll or interaction for 30+ seconds |
Attention has left the session |
Short; attention is already elsewhere |
|
Cursor exit |
Fast upward mouse movement toward browser controls |
Desktop visitor is about to close or navigate away |
Very short; seconds at most |
|
Attention decay |
Slowing scroll speed, long pauses mid-article |
Engagement declining throughout the session |
Wide; but harder to detect in real time |
All of These Signals Can Happen At Once
Each of these signals alone describes a moment. When they appear together, they describe an outcome.
A visitor who has scrolled to the bottom of an article (scroll completion), hasn't moved their cursor in 20 seconds (idle), and then moves quickly toward the browser bar (cursor exit) isn't showing potential disengagement. They're done. The session had three distinct warning signs, and nothing intervened.
This is where most publisher setups fall short. Related article widgets at the bottom of a page address scroll completion to some degree, but they're passive. They don't respond to idle time. They don't react to cursor behavior. And they do nothing for attention to decay that's already happened halfway up the article.
Well-targeted content at the time of exit gets a click-through rate between 8–20% compared to passive related article widgets placed at the bottom of pages. The difference isn't the offer itself. It's the timing and the context in which it appears.
This is also what separates signal detection from journey shaping. It is the practice of actively guiding visitors toward a relevant next step based on what they've just read and where they are in the session, rather than leaving navigation entirely to chance.
What Acting on This Actually Requires
Detecting these signals are more important than adding another widget.
Whenever a visitor is ready to leave, placing a relevant disruption can make them reconsider or check something else on your site. Showing a visitor a story they've likely already seen, or promoting content unrelated to their current topic, doesn't recover the session. It confirms there's nothing worth staying for.
This is the gap that separates detecting an exit from doing something useful with it. It's also why generic exit-intent tools (most of which were built for e-commerce) tend to underperform on publisher sites.
The decision logic is different. The audience is different. And the revenue model is different.
Where Admiral’s Extend Fits In
Admiral's Extend was built specifically around this problem. It monitors the behavioral signals described above, across desktop and mobile; and uses Admiral's Visitor Relationship Management (VRM) layer to determine the right next step at the exact moment a session is winding down.
Not a generic popup. A configurable response tied to content performance data, visitor context, and publisher-defined monetization priorities. Extend can route toward a high-yield article, a subscription offer, a direct-sold inventory page, or a cross-portfolio property; whatever the right move is for that session, at that moment.
Early Extend deployments have projected approximately a 4% lift in total monetized pageviews from existing traffic, without additional acquisition cost. For publishers operating at scale, that kind of gain on the traffic already arriving compounds into meaningful inventory expansion.
The sessions are already on your site. The signals are already there. The question is whether anything is responding to them.
