Amazon Advertising as a Business: Where It Fits in Internet Marketing and How the Numbers Work

Where Amazon advertising sits inside internet marketing

Internet marketing is the whole toolbox: SEO, email, social, search marketing, display, affiliate. Amazon advertising is the piece of that toolbox that runs inside Amazon's own data and inventory — you're buying attention from people already shopping, using signals like search terms, purchase history and browsing behavior that Google or Meta simply don't have.

Two engines do the buying. Sponsored ads (Sponsored Products, Sponsored Brands, Sponsored Display) are self-serve, cost-per-click, and targeted by keyword or product — the Amazon equivalent of search marketing. Amazon DSP is programmatic: it buys display, video and audio across Amazon-owned properties and third-party sites, priced on CPM, and it needs a seat to run it. Most businesses start with sponsored ads because there's no seat requirement and no minimum spend commitment attached to them.

What each layer actually costs — and what the headline number leaves out

Amazon's own advertising page states plainly that a managed-service option — working with an Amazon Ads account executive — typically requires a minimum investment of $50,000. Sponsored ads, run yourself through the self-serve console, carry no upfront fee; you pay per click, set by auction, with no published floor.

Neither number tells you the rate you'll actually pay once real budget is moving. CPC and CPM depend on category competition, placement mix, and whether sponsored and DSP campaigns are set up so they aren't bidding against each other in the same auction. That's the figure worth asking any partner for directly — a minimum spend requirement and an actual blended rate are two different questions, and vendors that answer only the first one are giving you half the picture.

Running Amazon advertising as a business: break-even first, then who runs it

Before any budget goes in, work out what each order can afford to pay for advertising. Take an example product that sells for $40 and has $14 left after product cost and Amazon fees. That is 35% of the price ($14 ÷ $40), so a 35% ACoS is break-even: spend more than that per sale and the order loses money. With a 10% click-to-order conversion rate, the most you can bid per click is $14 × 10% = $1.40. This is an illustration with made-up inputs, not a quote or a client result, so swap in your own price, margin and conversion rate.

Then decide who runs it. There are three routes, and each suits a different business:

• In-house, self-serve. Best when you have one person who can check search term reports weekly and a catalog small enough to review by hand.
• An Amazon ads management company or agency. Best when ad spend is large enough that a missed week is expensive. Ask how they are paid (flat fee, percentage of ad spend, or both), what the notice period is, and who owns the ad account.
• A managed-service provider for Amazon DSP. Best when sponsored ads already work and you want reach beyond search with proper measurement. This is what reMKTR does.

For sponsored ads alone, Dr. PPC (drppc.ai) is our autonomous ad-management product: $300 a month plus 2% of ad spend, the first 30 days are $1, Orbit included, no contract.

A worked example: what 'good' actually looks like at scale

Numbers without a book behind them are just copy. Here's one from an actual DSP portfolio: across 30 advertisers running through reMKTR in July 2026, the book delivered 6.04x return on ad spend — measured across the whole portfolio, not pulled from the best-performing line item. That same book ran 78.4 million impressions at a $4.00 CPM and a blended $1.42 cost-per-click, with a blended cost-per-acquisition of $5.49 across 57,137 attributed purchases. One in five of those purchases — 20.1% — came from a shopper new to the brand, not a repeat buyer being re-served an ad they'd have converted on anyway.

That $1.42 blended CPC is worth sitting with, because a much lower number circulates constantly in this category: $0.41. That figure is real, but it's online-video only — one placement inside DSP, not the whole book. Quoting it as 'the' Amazon DSP CPC is the single most common way this category gets misread, and it happens because online-video is the cheapest placement type, not the most typical one.

The mistake that breaks the measurement, not just the media plan

Last-click attribution tells you which ad touched a sale last. It cannot tell you whether that sale would have happened anyway — and treating it as proof of incrementality is the most common analytical mistake in this business, including ones we've had to walk back in our own client reviews. If a shopper saw a sponsored ad, then a DSP retargeting ad, then bought, last-click hands all the credit to whichever touchpoint fired closest to the purchase. It has no mechanism to prove the DSP spend added anything at all.

Amazon Marketing Cloud addresses the double-counting piece specifically: it reconciles DSP and sponsored ads against the same purchase events so one sale doesn't get claimed as a win by both channels. But reconciliation isn't incrementality. Proving that the ad caused the sale, rather than just witnessed it, needs a holdout test or a matched control group — a bigger attribution window doesn't get you there.

What to do when the numbers are bad news

• ROAS is falling. Check creative fatigue first. The same three DSP creatives running for eight straight weeks is a far more common cause than audience decay, and it's cheaper to fix.
• New-to-brand share is low. That usually means DSP is retargeting people sponsored ads would have converted anyway, not finding anyone new. Separate the audiences before cutting the budget.
• CPA is climbing. Check whether sponsored ads and DSP are bidding against each other inside the same auction — it happens more often than either team notices, and it's an easy structural fix.
• The holdout test shows no lift. That's a real result, not a broken test. It means the spend didn't add sales beyond what would have happened without it. The honest move is to reallocate the budget, not to re-run the test until it says something else.

Where reMKTR fits in this

reMKTR runs Amazon DSP as a managed service. We hold our own Amazon DSP accounts and reconcile every campaign through Amazon Marketing Cloud so DSP and sponsored ads stop double-counting each other before we tell a client what actually worked. We're part of the Full Circle group (100+ brands managed, $500M+ in Amazon revenue managed). None of that matters if you're running sponsored ads yourself on a five-figure budget — for that, Amazon's own self-serve console and its free Ads Academy courses are the right tool, and this page owes you that answer as much as any other on it.