Amazon Marketing Techniques: A Practitioner's Breakdown

The five techniques, and what each one is actually built to do

Every tactic on Amazon falls into one of five buckets. Confusing which bucket a tactic sits in is the fastest way to judge it by the wrong number.

• Listing and SEO — titles, backend keywords, images, A+ content. This determines whether Amazon's own search algorithm surfaces you organically. No spend involved, but it caps how efficient every paid tactic below it can be.
• Sponsored Products — cost-per-click bidding on search terms. Bottom-funnel. It converts intent that already exists; it rarely creates intent.
• Sponsored Brands and Sponsored Display — mid-funnel. Builds awareness at the search results page or retargets shoppers who viewed a product but didn't buy.
• Amazon DSP — programmatic display and video, on and off Amazon, bought through a DSP seat rather than the sponsored ads console. Upper-funnel reach plus retargeting at scale, but it needs a different measurement approach than the tactics above it — more on that below.
• Deals, coupons, and promotions — a velocity lever. Moves units and organic rank in the short term, at the cost of margin, and the lift often disappears when the deal ends.

A brand store or enhanced content isn't really a sixth technique — it's what converts the traffic the other five send. It doesn't generate demand on its own.

A worked example: what the numbers look like at scale

Definitions are easy. The part every other guide on this topic skips is what the numbers actually do when a technique runs at volume. Here's a real one: across 30 advertisers we manage, one month's DSP activity ran 78.4 million impressions at a $4.00 CPM, with a blended $1.42 cost-per-click across the whole book — the $0.41 CPC figure that gets quoted everywhere in this category is online-video only, not the whole account.

Do the arithmetic on just the impression volume: 78.4 million impressions ÷ 1,000 × $4.00 CPM comes to roughly $313,600 in display spend before any click-based line items are added. Separately, that same book produced 57,137 attributed purchases at a blended $5.49 cost per acquisition — 57,137 × $5.49 is about $313,700. Those two numbers, arrived at from opposite directions, land within a rounding error of each other. That kind of internal consistency is what you're checking for when you're told a number is real — not whether it sounds impressive.

The book returned 6.04x ROAS overall, and 20.1% of those purchases came from shoppers new to the brand — a figure that matters more for DSP than for Sponsored Products, because upper-funnel spend is supposed to be finding people who haven't bought yet. A number like that only means something measured across the whole portfolio, not pulled from the single best-performing line item.

The common mistakes — including the one almost everyone makes at least once

Judging DSP by last-click sales. Last-click attribution gives credit to whichever ad a shopper touched right before checkout — usually a Sponsored Products ad, because it sits closest to the buy button. DSP frequently gets blamed for 'not converting' when its actual job was to put the shopper in the funnel three days earlier. Last-click can't see that. It was never built to.

Treating ACOS as the whole truth. A campaign with a high ACOS on branded search terms might be fine — you're defending traffic you'd lose to a competitor otherwise. A campaign with a low ACOS on a term nobody else bids on might be doing nothing incremental at all. The number alone doesn't tell you which.

Reporting DSP and sponsored ads separately and adding the results together. This is a mistake we've made ourselves, in the early days of running seats before we had Amazon Marketing Cloud reconciliation in place. If a shopper sees a DSP ad and later clicks a Sponsored Products ad for the same product, both channels can claim the sale in their own dashboard. Add the two dashboards' numbers together and you've double-counted revenue that only happened once.

Cutting a tactic the week it looks bad. Upper-funnel spend often shows soft direct-response numbers by design — it's not supposed to close the sale on the same visit.

When a technique looks like it's not working, check this before you cut it

Bad numbers are usually a measurement problem or a setup problem before they're a strategy problem. Work through these in order:

• Is this the right metric for this funnel stage? Judging DSP on same-day ROAS is like judging a billboard on same-day foot traffic. Use view-through and new-to-brand rate for upper-funnel; use ACOS and CVR for bottom-funnel.
• Is the setting already doing what you think it isn't? Dynamic bidding, negative keyword lists, and placement exclusions get toggled once and forgotten. Check the account before assuming the tactic itself failed.
• Is there channel overlap inflating or hiding the result? If DSP and Sponsored Display are retargeting the same audience, one of them is getting credit that belongs to the other. Reconcile in AMC or a similar clean room before you trust either number.
• Would the sale have happened anyway? This is the question last-click literally cannot answer. A holdout group — some audience segments see the ad, a matched group doesn't — is the only way to see the incremental lift rather than the correlated one.

If you've checked all four and the number is still bad, then it's a strategy problem — wrong creative, wrong audience, wrong bid. But that's the last box to check, not the first.

Where measurement actually fits into this

Everything above is why we build the way we do. reMKTR runs Amazon DSP as a managed service across 109 live advertiser seats, and we hold the seats directly rather than reselling access, because reconciling DSP against sponsored ads in Amazon Marketing Cloud is the only honest way to say whether display added a sale or just claimed one that Sponsored Products already closed. We're part of the Full Circle group, which has managed more than $500M in Amazon spend across 100+ brands. None of that changes the fact that Sponsored Products, listing optimization, and deals are the right starting point for a lot of sellers — DSP earns its place in the mix once there's enough volume and enough budget to measure incrementality properly rather than guess at it.