What Is Display Advertising?

The plain definition

Display advertising means a brand pays to place a visual ad — a static banner, a short video, an audio spot, or an interactive rich-media unit — on someone else's website, app, or streaming service. It's the opposite of search advertising, where the ad only shows because someone typed a query. Display shows up whether or not the viewer was looking for you: on the side of a news article, in the middle of a recipe blog, before a streaming video starts.

Most display is bought programmatically, through a demand-side platform, and priced on a CPM basis — cost per thousand impressions — rather than cost per click. The Interactive Advertising Bureau sets the standard sizes: a 300x250 medium rectangle, a 728x90 leaderboard, a 160x600 wide skyscraper, a 300x600 half page, and mobile banners from 320x50 up. Formats have widened since then to include native ad units that match the look of the page they sit on, and video and audio units that run inside streaming content.

A worked example, on real numbers

Definitions don't tell you whether display works. Numbers do. Across 30 advertisers running Amazon DSP in July 2026, the book delivered 78.4 million impressions at a $4.00 CPM. That's 78,400 units of a thousand impressions, at $4.00 each — roughly $313,600 in media spend. The campaigns returned 6.04x ROAS, meaning that spend generated close to $1.9 million in attributed revenue, across 57,137 purchases at a blended $5.49 cost per acquisition. Multiply 57,137 by $5.49 and you land back near that same spend figure — the math has to reconcile, or the numbers aren't real. Of those purchases, 20.1% came from shoppers new to the brand, which is the number that actually separates display from remarketing: it's the portion of the campaign doing acquisition work, not just reminding existing buyers.

That's the shape of a display campaign: impressions bought at a CPM, a fraction of viewers clicking at a blended $1.42 cost per click, and a smaller fraction converting days or weeks later. Most of the value in display doesn't show up as a click at all — it shows up as someone who saw the ad, didn't click, and bought anyway. That's also why display gets under-credited by last-click reporting, which we cover below.

Examples of display advertising

• Static banner: a product image and a headline on the side of a recipe blog, linking to a product page.
• Retargeting: a shopper looks at a wallet from a brand like Ridge on Amazon, doesn't buy, and later sees that same product image ad on an unrelated news site.
• Video pre-roll: a 15-second clip for a grooming brand like Beardbrand before a streaming show starts.
• Audio display: a spoken ad with a companion image, delivered inside a music or podcast app.
• Rich media / interactive: an expandable unit for a garden brand like Epic Gardening that lets a viewer browse a few products without leaving the page.

What all five have in common: none of them required the viewer to search for anything. That's the defining trait of display versus search or sponsored product ads — the ad interrupts a session that was about something else.

The common mistake — including ours

The single biggest mistake in reading display results is judging them with last-click attribution. Last-click gives all the credit to whichever ad the shopper clicked right before buying — usually a search or sponsored ad — and gives display none, even when the display impression is what made the shopper receptive in the first place. Last-click can't prove incrementality. It never could; it wasn't built to.

We've made this mistake ourselves: reading a display line item as underperforming purely because its last-click ROAS looked thin next to sponsored ads running the same week, when the two channels were influencing the same purchase. The honest fix is a holdout or a matched control — run display against part of the audience and withhold it from the rest, then compare outcomes — reconciled in Amazon Marketing Cloud, where DSP and sponsored ads stop double-counting each other's credit. If the holdout test shows no lift, that's real information. The answer isn't to reweight the attribution model until display looks better. It's to cut the spend or change the creative and test again.

When the number looks wrong

If display CTR looks low — under half a percent is normal, not broken — that's not a signal to panic; display was never meant to perform like search. If ROAS looks bad on a last-click view, check whether the same purchases are also being credited to a sponsored ad campaign before assuming display failed. If a holdout test genuinely shows zero lift over a control group, that's the moment to turn the campaign off, not the moment to find a better attribution window that makes it look fine.

On measurement questions like this, the practical answer is Amazon DSP — Amazon's Demand-Side Platform for programmatic display, video, and audio, not the Delivery Service Partner courier program some searches confuse it with. Dr. DSP runs Amazon DSP as a managed product: every change carries the evidence behind it, a measurement plan, and a rollback trigger before it goes live, and clients choose full human approval, supervised, or fully autonomous execution inside agreed guardrails. Full Circle, the company behind it, has managed more than $500M in revenue across 100+ brands. There's no published price — it's a demo, the first 30 days free, and pricing set on the call against real budget and scope, with the Orbit software suite included at no extra cost.