The Science of Ad Layout: Placement, Density and Revenue per Session

Two sites with identical traffic and identical demand stacks can earn wildly different revenue. The difference is usually layout: which slots exist, where they sit, what sizes they serve and how densely they pack the page.

The placement hierarchy

Across verticals, the consistent winners are in-content units placed after the 2nd–4th paragraph (high viewability, natural reading pauses), sticky sidebar units on desktop that stay in view while users scroll, anchor/adhesion units on mobile (60–75% viewability is common), and mid-article video players where content justifies them. The consistent losers: footer leaderboards nobody reaches and any slot that renders off-screen and never enters the viewport.

Density: more isn't more

Each added unit lifts page RPM with diminishing returns while degrading engagement — slower pages, more clutter, fewer pages per session. The revenue-optimal density is almost always lower than the maximum tolerable density, because session length is part of the revenue equation. Optimize revenue per session, not revenue per page.

Test like you mean it

• Change one variable per experiment: a placement, a size set, a density step.
• Split traffic randomly with a persistent assignment, and run to statistical significance — not until the chart "looks better."
• Constrain experiments with UX guardrails: CLS budget, scroll-depth and bounce deltas, page-speed ceilings.
• Re-test seasonally; layouts that win in Q4 demand conditions can lose in Q1.

Per-template, not per-site

Article pages, galleries, homepages and category hubs each have their own engagement physics — one winning layout per site is a myth. Automated multivariate testing per template, with guardrails, is what our AI Layout module industrializes: it proposes variants, runs honest splits and promotes winners continuously, so layout stops being a once-a-year redesign argument and becomes a compounding optimization loop.