What Does DSP Stand For on Amazon?

Two different "Amazon DSPs" — pick the right one first

DSP stands for Demand-Side Platform. If you're reading about advertising, that's your answer: it's the software Amazon Ads uses to buy digital ad inventory automatically, in real time, across exchanges, rather than an advertiser negotiating placements by hand.

There is a second, entirely unrelated program that also goes by "Amazon DSP": Delivery Service Partner, the franchise-style program that lets someone start a local package-delivery business using Amazon vans and routes. Same three letters, same parent company, nothing else in common. If you searched this because you're looking into starting a delivery business, this page won't help you — you want Delivery Service Partner, not the advertising platform. If you're trying to understand how Amazon sells ads, or why your account manager keeps saying "DSP," keep reading.

Inside advertising, Demand-Side Platform is a category, not an Amazon invention. Google, The Trade Desk and Amazon all run one. What Amazon's version does that the others can't: it can target based on Amazon shopping and streaming signals, and it can buy placements on Amazon-owned properties like Fire TV, Twitch and IMDb, as well as on third-party sites and apps through the open exchange.

How Amazon DSP actually works — a worked example

Amazon DSP runs on real-time bidding. A shopper loads a page, the publisher's ad server announces the available ad slot through a supply-side platform, and DSPs on the buy side bid for it in milliseconds based on who that shopper is and what the advertiser is willing to pay. Whoever wins fills the slot before the page finishes loading. That part matches every other DSP in the market — it's plumbing, not magic.

What the plumbing produces, in real terms: across 30 advertisers running Amazon DSP through Full Circle and reMKTR in July 2026, the book delivered 6.04x return on ad spend, 78.4 million impressions at a $4.00 CPM and a blended $1.42 cost-per-click. Blended cost per acquisition was $5.49 across 57,137 attributed purchases, 20.1% of them from shoppers new to the brand.

Read those numbers as a chain, not a scoreboard. CPM buys the impressions. CPC tells you what it cost when someone acted on one. CPA tells you what it cost per purchase once those clicks converted. The new-to-brand share is the number that actually separates DSP from a retargeting exercise — it's evidence the campaign reached people who hadn't already decided to buy, which is the entire point of upper-funnel spend.

DSP vs Sponsored Ads: they're not competing for the same job

Sponsored Products, Sponsored Brands and Sponsored Display are auction-based and mostly keyword- or product-targeted, running inside Amazon's own search and detail pages. Amazon DSP is programmatic, audience-targeted, and runs everywhere — on Amazon and off it. Sponsored ads catch demand that already exists. DSP is more often trying to create it.

That difference is also why comparing their ROAS side by side is misleading. A shopper who saw a DSP display ad three days ago, then searched the brand name and clicked a Sponsored Brands ad, will usually get credited to the sponsored ad on a last-click basis — because it touched the sale last. The DSP ad did real work and got zero credit for it. Last-click attribution can't see that chain; it can only see the final click.

The honest fix is a holdout: run DSP against one audience segment and withhold it from a matched control, then compare purchase behavior between the two groups. That isolates what DSP actually added, instead of what it happened to touch last. Reconciling DSP and sponsored ads together in Amazon Marketing Cloud is the only way to see whether they're double-counting the same shopper — which they very often are.

The mistake that wastes DSP budget

The single most common error, including ones we've made ourselves: cutting DSP spend because its last-click ROAS looks weak next to sponsored ads, without checking whether DSP was influencing sales that got credited elsewhere. That's not a DSP problem. It's an attribution problem, and it points the axe at the wrong budget line.

• Bad signal: DSP ROAS looks low on a last-click report, so spend gets cut.
• What actually happened: DSP reached new-to-brand shoppers who converted later via a branded search click, which absorbed the credit.
• How to check it: run a holdout or matched-control test, or reconcile both channels in Amazon Marketing Cloud, before touching the budget.

The second common mistake is the opposite: assuming DSP is working just because impressions and reach numbers look big. Reach isn't proof of anything by itself. A CPM of $4.00 buying 78.4 million impressions only means something once you can trace a share of those impressions to purchases that wouldn't have happened otherwise — which is what a holdout is for, not a dashboard.

What to do when the DSP number is wrong

If your CPA suddenly jumps, check audience overlap first — bidding against your own retargeting pool and a broad prospecting segment at the same time inflates cost per acquisition without you noticing. If ROAS looks too good, check whether the reporting window overlaps with a branded search campaign that would have converted those same shoppers anyway; DSP shouldn't get sole credit for demand it didn't create. If a fix doesn't move the number after a full attribution cycle, the fix didn't work — revert it and re-test one variable at a time, not three.

Dr. DSP, from Full Circle, runs Amazon DSP as a managed product and treats this as the baseline, not an add-on: every change carries the evidence behind it, a measurement plan, and a rollback trigger before it goes live. That's the practitioner discipline this page is arguing for, whether or not you ever run DSP through us. Full Circle has managed more than $500M in revenue across 100+ brands, and Dr. DSP has no published price — it's a demo, a free first 30 days, and pricing set on a call against real budget and scope, with the Orbit software suite included either way.