We Tested 6 Ad Layouts: Here's What Actually Moved RPM

Everyone has opinions about ad placement. “Above the fold kills bounce rate.” “In content ads annoy users.” “Sidebar ads do not work on mobile.”

We tested six different layouts across fifty thousand sessions on a finance site pulling a thirty eight dollar RPM baseline. Here's what the data showed.

The test setup

Same traffic source, same content, same time period. We split traffic into six variants and ran each for three weeks. We measured RPM, bounce rate, pages per session, and return visitor rate.

This was not about finding the best layout for every site. It was about seeing which trade offs were worth making for this specific asset and audience.

Layout A: conservative

This was the control variant. Good user experience, but clearly leaving some revenue on the table. It became our baseline for comparison.

Layout B: aggressive above the fold

RPM went up, but engagement dropped sharply. Users felt ambushed at the top of the page. For a property that depends on returning visitors, this trade off was not acceptable. We would rather have slightly lower RPM and keep people coming back.

Layout C: dense in content

This variant produced a strong RPM lift with only a slight hit to engagement. The trade off was acceptable: we lost 0.1 pages per session but gained nearly seven dollars RPM.

Layout D: sticky sidebar

Sticky sidebar on desktop gave a strong RPM lift with almost no impact on engagement. That lines up with the intuition that sidebar units do not interrupt reading flow in the same way as in content or above the fold units.

The takeaway

The best performing setup combined the dense in content approach with a sticky sidebar. The final result was roughly fifty two dollar RPM, an increase of more than thirty percent from baseline, with only minor changes in engagement metrics.

Your numbers will be different. Niche, audience, and traffic source all matter. The important part is the approach: test systematically, measure everything, and optimize for total value instead of chasing RPM in isolation.