Ad discovery across devices

An ad discovery scan opens a publisher page on the device you chose, scrolls it the way a reader would, and reports every ad block it finds — each one becomes its own scan with a creative, a landing page and tags. Run the same page on several devices and the counts will differ. Most of that difference is the page's doing; this article explains which part is, and which part the scanner is accountable for.

Why the counts differ

The page lays out different ads for different screens. A publisher sells placements per breakpoint. A desktop or tablet layout has a sidebar with two or three 300x250 rectangles and a 728x90 leaderboard; the phone layout has none of those and instead stacks a full-width banner and a full-width native video unit into the single column. The ads are not fewer or more — they are different placements, and the scan reports what that device's visitor would see.

Ad networks fill differently every time. The same page on the same device serves a different mix on every load: a slot is auctioned each time and sometimes nobody bids. Two scans minutes apart can differ by one or two ads on any device, and a network may skip a device class or a country outright. A single scan per device is a sample, not a count.

Interstitials and pop-unders are gesture-driven. A full-screen overlay or a new tab that only appears after the first tap is reported when the scan triggers it; whether it fires depends on the network's frequency capping and on the device, and it counts as one more ad when it does.

Some publishers serve ads first-party. When a site hosts its own ad system, there is no ad-network request to see. The scanner then relies on the block's shape and on where its link goes, which is weaker evidence than a tagged ad frame, so a first-party slot is found less reliably than a network-served one on every device alike.

What the scanner guarantees

The detector judges a block by the same rules on every device. In particular:

• A full-width mobile unit is recognised at any phone width. A unit sold as 320x50, 320x100 or 300x250 renders as "100% of the column" on a phone, so its pixel width follows the device — 340 px on a 360 px phone, 373 px on a 393 px one. The scanner recognises the unit by its height and by the fact that it spans the column, not by a pixel width, so a 360 px and a 393 px phone find the same block.
• An overlay does not hide the ads under it. A full-screen ad frame layered over the page is reported as its own ad, and the banners it covers are still reported as theirs.
• A wrapper and the ad inside it are one ad, not two. Publishers wrap ad frames in containers; the scanner reports the creative once.

If the same page on two phones of different widths reports a different set of ad blocks — not a different fill, a block that is present on both and reported on one — that is a scanner issue. Send us the two scan IDs.

Comparing devices fairly

• Fix everything but the device. Same geo, same proxy type, same max_discovered_ads, and run the devices in the same hour. A run that fans out across devices does this for you.
• Repeat. Three repeats per device separate fill variance from a real difference; one scan cannot.
• Compare placements, not counts. Open the discovered ads and look at their sizes: a desktop scan reporting three 300x250 rectangles the phone scan lacks is a sidebar, not a gap. Use the scan detail page's screenshot to see where each block sat.
• Expect tablets to look like desktops. An iPad at 1024 px wide, or an Android tablet at 800 px, usually gets the desktop or tablet layout, sidebar included, and its counts follow the desktop's rather than the phone's.

Checking only some formats

A discovery scan can be limited to banners, video, pops, or any mix of them. A pops-only scan never screenshots or clicks a banner, so it is cheaper and gets to the page's pop sooner. With no banner clicks to set it off, the scan deliberately clicks an ad-free point of the page to trigger the pop.

When comparing devices, keep the format selection the same too: a banners-only run on one device and a full run on another differ by every pop the second one found.

See also Campaigns & groups for the max_discovered_ads and ad-format settings and Reading scan results for what each discovered ad's scan contains.