Ad Layout Rebuild: Fewer Units, Higher Session RPM

What happens when units come out: Three article template rebuilds in the last two years took slots per page from an average of nine to five. Session revenue rose in all three, by 6 to 14 percent, because viewable rate on the surviving slots moved from the high fifties to above eighty and readers went one page deeper before leaving. That range is our history, not your forecast; it depends on your traffic mix and on how bad the starting layout is. But the direction is not an accident, and it is why we do not add units. How we decide which slots die: Every slot is valued on viewable time, not on viewability. The MRC threshold, half the pixels for one continuous second and two for video, is a pass or fail that hides the thing buyers pay a premium for, which is dwell. We pull Active View by unit, position and viewport bucket from the ad server, join it to scroll depth, and rank slots by revenue per unit of reader attention consumed. Slots that earn their place are the first in-content unit and the one sitting where readers actually pause. Slots that never survive are the ones below the point where 80 percent of readers have already left, and any two units that share a viewport with each other. Lazy loading and refresh get tuned, not switched on. Fetch and render margins are set per template against measured scroll velocity, because a recipe page and a liveblog do not scroll alike, and the defaults were written for neither. Refresh runs on continuous in-view time past thirty seconds, requires an interaction signal inside the window, caps per session, and never fires on a slot that has not been viewable once. How we test it: Buckets are assigned server side and passed to the ad server as a key-value, so revenue and behaviour can be read from one report. Two weeks minimum per arm. The decision metric is revenue per session and pages per session, together. Unit CPM rises when you remove units and on its own it proves nothing. Health and pharma pages: Native and recommendation-style formats stay out of symptom, dosage and treatment articles. A unit borrowing editorial styling beside a dosage table reads as clinical advice, and no yield gain covers that. In-content units stay out of the space between a claim and its citation. If therapeutic content and commercial content share one template, we separate them before we tune anything. What we will not ship: Interstitials and vignettes on mobile web. Anchor units that cover a subscribe or consent control. Refresh under thirty seconds. Ad density above the mobile Better Ads threshold, which we treat as a ceiling and not a target. And we will not add a unit to reach a number this quarter; if that is the brief, the honest answer is a rate card change, not a layout change. Not for you if: Your revenue plan depends on units you cannot remove because they are contracted. Your CMS will not let anyone change template markup. Or ad operations and editorial cannot sit in one meeting, which is the failure we see most: this work moves the shape of the page, and it dies when it has to be negotiated twice.