Amazon DSP Marketing, Explained With Real Campaign Numbers

What Amazon DSP Marketing Actually Means

Amazon DSP marketing means buying display, video, audio and streaming TV ad space through Amazon's demand-side platform — a programmatic buying tool, not a keyword auction. Sponsored Products, Sponsored Brands and Sponsored Display run on Amazon's search-ads side and are bought by keyword or product targeting. DSP runs on audience signals: what someone has browsed, bought, or streamed, on Amazon and off it.

That distinction matters because DSP is the only one of the two that can run entirely off Amazon.com — on Prime Video, Twitch, Fire TV, IMDb, and third-party sites and apps through Amazon Publisher Direct. You don't need to sell a product on Amazon to buy DSP; travel brands and services companies use it purely for the audience data.

Access comes two ways: self-service, where you or your team run the seat directly, or managed service, which Amazon itself gates behind a minimum spend commonly cited around $50,000, because the tooling — bid strategy, audience building, creative rotation, AMC queries — takes real operating time. Agencies and in-house teams that hold DSP seats are managing that operating time on a client's behalf, not adding a markup for access.

How The Buy Actually Happens

Every DSP impression is bought through real-time bidding. A shopper loads a page, the publisher's ad server flags available inventory to a supply-side platform, and DSPs — including Amazon's — bid on it in the milliseconds before the page renders. Amazon's advantage is what it's bidding with: purchase history, browse behavior, and streaming activity that no outside DSP can see.

What that buys, at scale, is worth pinning to real numbers rather than a category average. Across 30 of reMKTR's live advertisers in July 2026, the book ran 78.4 million impressions at a $4.00 CPM, blending to a $1.42 cost-per-click across the whole book — not just the online-video line item, which is where the often-quoted $0.41 CPC figure in this category actually comes from. Blended CPC across every format is a different, more honest number than the cheapest channel's number.

The gap between those two figures is the first thing to check when someone hands you a DSP quote or a case study: ask which formats are included in the CPC they're citing.

What The Funnel Looks Like With Real Numbers

A DSP campaign has four things worth tracking as it moves from reach to revenue: impressions, clicks, purchases, and — critically — how many of those purchases were from shoppers who hadn't bought the brand before. The table below is that same July 2026 book, 30 advertisers, one portfolio, not a best-performing line item pulled out to look good.

The number worth sitting with is the last one. 57,137 attributed purchases at a blended $5.49 cost per acquisition, with one in five of those purchases — 20.1% — coming from a shopper new to the brand. That new-to-brand share is the actual argument for DSP over more Sponsored Ads spend: Sponsored Ads mostly catches demand that already exists. DSP is one of the few Amazon ad products built to create demand that doesn't exist yet.

The Attribution Mistake That Makes DSP Look Better Or Worse Than It Is

Last-click attribution hands 100% of the credit for a sale to whichever ad the shopper clicked last. If a shopper sees a DSP display ad on Monday and clicks a Sponsored Products ad on Friday before buying, last-click gives the sale entirely to Sponsored Products — even though the display ad may have been what brought them back. Run it the other way and DSP takes credit for a sale a Sponsored ad already had covered. Either way, the two channels are double-counting each other, and a last-click report can't tell you which.

The honest fix is to stop asking which ad got the click and start asking whether the sale would have happened without the DSP spend at all. That's an incrementality question, and it needs a holdout group or a matched control — shoppers who weren't shown the ad, compared against shoppers who were. Amazon Marketing Cloud is where that comparison happens, because it's the only place DSP and Sponsored Ads events sit in the same log without either channel's dashboard claiming the sale first.

This is a mistake worth naming plainly because it's an easy one to make even when you know better: it's tempting to report the bigger, cleaner-looking last-click ROAS number to a client or a boss, especially when the AMC number comes in lower. The AMC number is the one that's actually true.

When The DSP Numbers Are Bad News

Sometimes the honest measurement is the unwelcome one. If ROAS in AMC comes in well below the last-click number you were reporting before, that's not a broken query — that's the double-counting finally showing up. Don't rerun the report hoping for a better answer; rerun it with a longer holdout window and check whether the audience segments overlap with Sponsored Ads targeting already in flight.

If CPA is high and new-to-brand share is low, the campaign is probably reaching people who were going to buy anyway — check whether audience definitions are too narrow (repeat purchasers, existing cart abandoners) rather than genuinely prospecting. If CPM is high relative to a $4.00 blended benchmark, check format mix before assuming the auction is expensive; online video and streaming TV cost more per thousand than standard display, and a bad benchmark comparison will make a normal campaign look broken.

And if the fix doesn't work the first time — the ROAS doesn't move after adjusting the audience or the creative — that's usually a sign the test wasn't isolated. A holdout that ran during a sitewide promotion, or against a product that was out of stock for two of the four weeks, will give you a number that has nothing to do with the ad.

Where reMKTR Fits

reMKTR runs Amazon DSP as a managed service and holds real DSP seats — 109 of them live at the moment — rather than reselling access through someone else's. We reconcile every campaign in Amazon Marketing Cloud specifically because last-click can't answer the incrementality question above, and we'd rather show a client a true, lower number than a flattering, wrong one. That's the same discipline behind the numbers used throughout this page. Whether or not that's the right setup for your team, the questions above — which CPC is being quoted, whether new-to-brand share is being tracked, whether the ROAS is a last-click number or an AMC one — are the ones worth asking of any DSP partner, including us.