The ad units that cost publishers the most are the ones buyers have learned not to see. Demand-side platforms price every placement on its viewability history, so where an ad sits and how it loads decide what it earns before any floor or partner setting does. Below: the formats and placements that earn on desktop, mobile web, and in-app in 2026, the viewability mechanics buyers use, and the four technical changes (lazy loading, reserved space, in-view refresh, format discipline) that raise yield without wrecking the page.
Why placement decides yield before anything else
Buyers do not evaluate your site. They evaluate each ad unit, by its identifier, on the metrics their platform has recorded: viewable rate, time in view, click-through, and past win prices. A 300x250 that renders below the fold on mobile is not "a slightly worse unit"; it is a unit with a poor history that bidding algorithms have learned to discount, and that history follows it for weeks after you fix it. The IAB Tech Lab's ad product guidelines exist because standardized, viewable formats are what buyers can price with confidence.
Two failure modes account for most lost yield: too many units competing for the same attention (banner blindness plus latency), and units that technically load but are rarely seen (low viewability, discounted bids).
What is ad viewability and how is it measured?
The Media Rating Council standard, which Google Ad Manager's Active View reports against, counts a display impression as viewable when at least 50% of its pixels are on screen for at least one continuous second (two seconds for video; 30% of pixels for very large units). GAM reports the viewable rate, the measurable rate, and viewable eCPM per ad unit. Buyers see equivalent numbers on their side and bid accordingly.
Practical thresholds: units under 50% viewable are actively discounted by most buyers; 70% and above is where premium demand competes. The difference in eCPM between the two on the same site is routinely more than 30%.
How do formats and placements differ by platform?
Desktop web
- One premium leaderboard at the top (728x90 or 970x250), not repeated at the bottom of the viewport, where it dilutes the value of both.
- Rectangles in the reading path. The medium rectangle (300x250) inside or beside the content, and the half page (300x600) in a sticky sidebar, are what most demand is built to buy. Google's supported ad sizes page lists the full set; use flexible sizes so a slot can accept several.
- Sticky sidebars, done properly. A sidebar unit that stays in view as the reader scrolls is one of the highest-viewability placements on desktop, provided it does not cover content and respects the Better Ads Standards.
- Avoid pop-ups, auto-playing video with sound, and floating units that cover text. They show high CPMs for a week and cost retention for a year, and several are flagged by the Coalition for Better Ads, which Chrome enforces.
Mobile web
- Compact, viewable sizes: the mobile leaderboard (320x50), the large mobile banner (320x100), and the 300x250 placed between paragraphs where the reader's thumb pauses.
- Anchor units with limits. A bottom anchor (320x50) is highly viewable; keep it under 30% of the screen and dismissible, per the Better Ads Standards.
- Spacing. One unit per screen-height of content is a reasonable ceiling. Denser than that and viewability per unit falls faster than the unit count rises.
- Consent and privacy frameworks (GDPR, CCPA) must be wired before the first ad request, or the request goes out without signals and buyers price it as unconsented.
In-app and mobile gaming
- Interstitials at natural breaks only: between levels, after a task completes, on a loading screen. Interrupting active play is the most common yield-destroying mistake in games.
- Rewarded formats where the user opts in for a tangible benefit (an extra life, in-game currency). Google Ad Manager supports rewarded and rewarded interstitial ads for apps; completion rates and eCPMs are well above standard interstitials because the user chose to watch.
- Frequency caps per session, tuned by measuring day-7 retention against ad load rather than by guessing.
Four technical changes that raise yield without more units
1. Lazy load below-the-fold units
Request and render a unit only when the reader is about to reach it. In Google Publisher Tags this is enableLazyLoad with fetch and render margins you can tune per device; a common starting point fetches a few viewports ahead and renders about one viewport ahead, with larger margins on mobile where scrolling is faster. Units that would never have been seen stop being requested (raising viewability and cutting latency), and units that are seen still have time to load.
2. Reserve the space
An ad slot that expands when the creative arrives pushes content down, which hurts Cumulative Layout Shift and pushes other units out of view. Give each slot a fixed minimum height matching its most common size. This one change often lifts viewability on mid-article units by double digits.
3. Refresh in view, and declare it
Refreshing a unit that has been in view for a set time (30 seconds is the common floor) creates new impressions on placements buyers already like. Refresh only units that are actually in view, cap the number of refreshes per page view, and declare refresh to your demand partners; undeclared or too-fast refresh gets a placement priced down or excluded.
4. Match native to where banners are blind
In-feed and in-article native placements, styled to the template and labelled as advertising, earn where display has stopped working, and they survive element-hiding better. They are a complement to display, not a replacement.
Where Optimon fits
Placement problems are slow: a template change lowers one unit's viewability, buyers price it down over a few weeks, and the revenue loss is spread thin enough that nobody opens a ticket. Optimon watches your Google Ad Manager and Prebid data per ad unit and tells you which placements moved and why: a unit whose fill or eCPM drifted, a size that stopped clearing on mobile, a partner that went quiet on one section. To check what a specific page is actually serving, the free Ad Inspector extension shows the live GAM and Prebid picture on any page you open. Your team keeps the layout decisions; it just makes them with the numbers in front of it.
FAQ
What is a good viewability rate for display ads?
Above 70% is where premium demand competes; below 50% most buyers discount the placement. Measure per ad unit and device in Google Ad Manager's Active View reporting rather than as a site average.
Which ad sizes perform best for publishers?
The 300x250 medium rectangle, the 300x600 half page, the 728x90 and 970x250 leaderboards on desktop, and the 320x50 and 320x100 mobile banners carry the broadest demand. Flexible slots that accept several sizes let more buyers compete for the same placement.
Does lazy loading reduce ad revenue?
It reduces impressions that were never going to be seen and raises viewability on the rest, which raises what buyers bid. Revenue per page view usually rises; impression counts fall. Tune the fetch margin so units that are seen still have time to load.
How often can I refresh an ad unit?
Common practice is a refresh after at least 30 seconds in view, capped per page view, and declared to demand partners. Refresh units that are not in view or refresh too fast, and buyers price the placement down.
