Amazon Marketing Campaigns: How They Actually Work

What an Amazon marketing campaign actually is

An Amazon marketing campaign is a budgeted unit inside Amazon Ads that pairs a targeting method with a bid, running against a keyword, an ASIN, an audience, or a placement. There are four types, and matching the wrong one to the wrong funnel stage is the most common structural mistake advertisers make before they ever touch a bid.

• Sponsored Products — bottom-funnel, keyword or ASIN targeting, self-service, shows up in search results and on product pages.
• Sponsored Brands — mid-funnel, keyword targeting, self-service, the logo-and-headline block at the top of search.
• Sponsored Display — retargeting and audience-based, self-service, runs on and off Amazon.
• Amazon DSP — programmatic, audience-based, runs across Amazon-owned inventory and off-Amazon placements including streaming, and is typically bought as a managed seat rather than self-service.

Amazon's own advertising page states that working with an Ads account executive on a managed-service basis normally requires a minimum investment of USD 50,000. Self-service Sponsored Products, Sponsored Brands, and Sponsored Display have no such floor — the budget is whatever you set at the campaign level.

How a single campaign is structured, step by step

Inside every campaign type the hierarchy is the same: campaign, then ad group, then targeting, then bid. The campaign holds the budget and the date range. The ad group holds the products you're advertising. Targeting is what triggers the ad — a keyword, an ASIN, a category, or an audience segment. The bid is what you're willing to pay when that trigger fires.

Get the targeting layer wrong and nothing downstream fixes it. A campaign built on broad-match keywords with no negative list will pull in irrelevant search terms and burn budget on clicks that were never going to convert. This is the most common technical mistake in Amazon PPC, and it doesn't look like a failure — it looks like a slow, steady rise in ACOS that's easy to blame on the wrong thing.

A worked example: reading the math on a real campaign book

Here's what that math looks like on a real book, not a hypothetical. Across 30 of reMKTR's DSP advertisers in July 2026, the portfolio ran 78.4 million impressions at a $4.00 CPM, producing a blended $1.42 cost-per-click across the whole book. That matters because the $0.41 CPC quoted everywhere in this category is an online-video-only number — it describes one placement type, not what DSP costs blended across formats.

Those clicks turned into 57,137 attributed purchases at a blended $5.49 cost per acquisition, and the whole portfolio — measured across all 30 advertisers, not cherry-picked from the best line item — returned 6.04x on ad spend. One in five of those purchases, 20.1%, came from a shopper new to the brand. That new-to-brand number is worth watching harder than ROAS: a campaign can post a strong ROAS by re-selling to people who already buy from you, and that's maintenance, not growth.

The relationship to hold in your head, on any account: CPM sets impressions for a given spend, click-through rate converts impressions to clicks, CPC is spend divided by clicks, conversion rate converts clicks to orders, and CPA is spend divided by orders. Move one lever — a tighter audience, a better creative, a narrower keyword list — and every number downstream of it moves too.

When the numbers are bad news — what to check before you pause anything

A high ACOS or a stalled campaign isn't a reason to pause first and ask questions later. Check the search term or placement report before touching bids — a broad-match campaign with a bad ACOS is often being dragged down by a handful of irrelevant terms that are cheap to fix with negative keywords. Check the attribution window next; a short window on a considered purchase will systematically understate what the campaign actually did.

If a DSP campaign looks like it's working but Sponsored Products in the same catalog looks like it's slowing down, don't assume DSP stole the sale. Check Amazon Marketing Cloud first. DSP and sponsored ads report separately by default, and it's easy for one purchase to get counted as a win in both places, which flatters whichever channel's report you happen to be reading.

We've made this mistake ourselves: reading a strong last-click ROAS on a retargeting-heavy DSP campaign and calling it a win, when a holdout test later showed most of those buyers would have converted anyway. Last-click attribution cannot prove a campaign added incremental sales — it can only tell you a sale happened somewhere in the path. Holdout groups and matched controls can tell you what wouldn't have happened without the ad.

The mistakes that show up in almost every account audit

Three problems show up repeatedly. The first is leaving a campaign on Amazon's default bidding strategy indefinitely — a reasonable setting on day one, not a strategy for month six. The second is running Sponsored Products and Sponsored Brands on the same exact-match keywords with no negation plan, so the two campaign types bid against each other for the same search and the brand pays twice to win once. The third is a daily budget capped so low the campaign never clears the learning phase, which shows up as inconsistent week-to-week impressions that look like a targeting problem but are actually a budget problem.

None of these are exotic. They're checked in the first ten minutes of any competent audit, and they're worth checking on your own account before paying anyone to look at it for you.

Where a managed DSP partner fits into this

reMKTR runs Amazon DSP as a managed service — we hold 109 live DSP advertiser seats and reconcile results in Amazon Marketing Cloud so display and sponsored ads stop double-counting each other, which is the only honest way to say whether DSP spend actually added anything on top of search. We're part of the Full Circle group, which has managed more than $500M in Amazon ad spend across 100+ brands. If you're already running Sponsored Products yourself and the open question is whether upper-funnel DSP spend would move new-to-brand sales rather than just re-billing your existing buyers, that's the specific question worth answering — with a holdout test, whether or not you ever hire anyone to run it.