What Counts as an Ecommerce Advertising Solution—and How to Choose One

What 'ecommerce advertising solutions' actually covers

The phrase bundles three different things: the channels you can buy (search, social, display, video, CTV, Amazon), the software you buy them through (self-serve ad accounts, DSPs, ad networks), and the service model wrapped around it (self-managed, agency, or outcomes-based managed service). Most comparison shopping goes wrong because nobody separates these before evaluating options.

Channels answer where a shopper sees the ad. Software answers who controls the bid and owns the data. Service model answers who is accountable when a number comes back wrong. A $2,000/month Meta account and a six-figure Amazon DSP seat are both technically "ecommerce advertising solutions," and almost nothing else about them is comparable — different minimums, different attribution problems, different failure modes.

reMKTR operates the DSP layer specifically — 109 live Amazon DSP advertiser seats at last count — so the worked example later in this page leans on that data. Where it isn't the right lens for your situation, we'll say so.

Where each channel actually sits in the funnel

Every channel gets pitched as full-funnel by whoever sells it. In practice each one does one or two jobs well and gets judged on the wrong metric constantly — display measured on last-click sales it didn't cause, video measured on click-through when its job was awareness. The table below is the honest version: what a channel is built to do, and where reporting on it usually goes wrong.

The pattern underneath all six rows is the same: a channel built for discovery or awareness gets measured on a bottom-funnel metric, looks like it's underperforming, and gets cut — while the channel that actually closed the sale gets credit it didn't earn on its own.

A worked example: what a real Amazon DSP book looks like

Category labels don't tell you much. Numbers do. Across 30 Amazon DSP advertisers reMKTR managed in July 2026, the portfolio returned 6.04x ROAS — measured across the whole book, not the single best line item pulled out and presented as typical.

That return ran on 78.4 million impressions at a $4.00 CPM, with a blended $1.42 cost-per-click. If you've seen a $0.41 CPC quoted for Amazon DSP elsewhere, that number is real but it's online-video only — it disappears once you blend in display and every other format, which is most of a real book. Blended cost per acquisition was $5.49 across 57,137 attributed purchases, and 20.1% of those purchases came from a shopper who was new to the brand.

Put those five numbers together and you get the question a solution actually has to answer: not "what's the CPM" but "at that CPM and that CPC, how many purchases came in, what did each cost, and how many were customers you didn't already have." Any vendor can quote the first number. Fewer will hand you all five in the same sentence.

The mistake that inflates almost every ecommerce ad report

Last-click attribution hands full credit to whichever ad happened to run last, regardless of whether it caused the purchase or just happened to be nearby. Run display and sponsored search at the same time and last-click will credit both — a shopper who saw a display ad, ignored it, then clicked a search ad three days later, shows up as a win for the display campaign too. That's double-counting, not incrementality.

We made this mistake ourselves in earlier reporting cycles: crediting DSP performance from last-click pulls that overlapped with sponsored ads, which overstated what display was actually adding on top. Reconciling both in Amazon Marketing Cloud, so a purchase gets counted once instead of twice, is what fixed it — and it means our own current ROAS numbers are lower, and more defensible, than the ones we used to publish.

The real fix isn't a smarter attribution model. It's a holdout: run the channel for one audience and withhold it from a matched control, then compare outcomes. That's the only way to answer whether an ad caused a sale that wouldn't have happened anyway, and it's a harder standard than most dashboards are built to report against.

What to check before you cut a channel that looks like it's failing

A bad ROAS number is a symptom, not a verdict. Before pulling budget, check these in order:

• Spend threshold — most algorithms need a minimum volume of conversions before they optimize well; a channel starved of budget will look broken when it's actually just unfed.
• Measurement window — a seven-day attribution window will undercount a channel that influences a purchase two weeks later.
• Double-counting — check whether the same sale is being credited to two channels; this is the single most common reason display or CTV looks like it's failing.
• Creative fatigue — the same three ad units running for months will decay regardless of platform quality.
• Wrong metric for the funnel stage — judging an awareness channel on last-click ROAS will always look bad, because that was never its job.

If a channel still looks weak after all five checks, the honest answer is: it might genuinely not be working for your product. That happens. Not every channel suits every catalog, and no managed service — including ours — should tell you otherwise.

Where reMKTR fits in this

reMKTR runs Amazon DSP as a managed service, sold on outcomes rather than dashboards. We hold real DSP seats and reconcile in Amazon Marketing Cloud so DSP and sponsored ads stop double-counting each other, and we lean on holdouts and matched controls where last-click attribution can't answer the incrementality question. We're part of the Full Circle group — $500M+ in managed Amazon spend across 100+ brands, with 70+ live across the group right now. If your question is specifically about Amazon DSP at real spend, that's our lane. If it's about Meta, Google, or TikTok, the platforms covered elsewhere on this page are better places to look.