Ecommerce Advertising Strategies: How to Build One That Survives Contact With Real Data

Why most 'ecommerce advertising strategy' guides aren't strategies

Most pages that rank for this term give you a list: search, social, display, video, email, CTV, maybe affiliate. That's a media plan, not a strategy. A strategy is the decision about which channel does which job, and the rule for how you know if it's working.

The list matters, but the ordering matters more. Running social ads and paid search with the same target CPA is a common error, because they're not competing for the same demand. One captures a shopper who already decided what they want. The other creates the want in the first place. Grading them on the same scorecard guarantees you'll defund the channel that's actually doing the harder job.

Match the channel to the job, not the other way around

Before picking budgets, decide what each channel is being asked to prove. Search and Sponsored Products should be judged on conversion efficiency, because that's the job — intent already exists, you're just capturing it. Social, display, and video should be judged on whether they're building demand that shows up later as new-to-brand purchases, not on last-click ROAS, because last-click will always undercount them.

The table below is the version of this we actually use to sort a media plan. It's not a channel list — it's a job description for each line item, plus the pitfall we see most often on that specific channel.

A worked example: what 'strategy working' looks like in real numbers

Numbers make this concrete faster than another framework. Across 30 of reMKTR's Amazon DSP advertisers in July 2026, the portfolio-wide return was 6.04x ROAS — measured across the whole book, not pulled from the single best-performing line item, which is how most case studies get built.

That book ran 78.4 million impressions at a $4.00 CPM and a blended $1.42 CPC. The $0.41 CPC figure that circulates in this category is real, but it's an online-video-only number, not the whole book — quoting it as a blended average is the single most common way this category misleads itself. Blended cost per acquisition was $5.49 across 57,137 attributed purchases, and 20.1% of those purchases came from a shopper new to the brand.

That last figure is the one worth sitting with. A strategy that produces a great blended CPA but a shrinking new-to-brand share isn't growing the brand — it's harvesting the same warm audience more efficiently. Both numbers have to move together, or you're not measuring what you think you're measuring.

When the numbers say the strategy isn't working

If ROAS looks strong but growth has stalled, check new-to-brand share before you touch creative. A high blended ROAS propped up by repeat buyers means your top-of-funnel spend isn't doing its job, no matter what the dashboard says.

If CPA is climbing and you can't tell why, don't assume it's audience fatigue — check for overlap first. Amazon DSP and Sponsored Ads frequently get credit for the same purchase in separate reports, which inflates both numbers and hides which one actually caused the sale. This is a reconciliation problem, not a bidding problem, and no amount of bid adjustment fixes it.

If a channel's reported performance won't hold up when you pull it out entirely for a control group, it wasn't performing — it was riding on demand another channel created. A holdout test (running the campaign for one matched group and withholding it from another) is the only way to find this out. Last-click attribution cannot prove incrementality; it was never built to.

The mistakes that wreck a strategy, including ones we've made

• Grading upper-funnel channels on last-click. Display and video get starved of budget because last-click makes them look worse than they are — they're doing a different job than search.
• Quoting a category-average CPC as if it applies to your mix. A $0.41 CPC headline number is usually one format, not the blended reality of a full media plan.
• Chasing the cheapest CPC channel this month. Cheap clicks without a new-to-brand check just means you're paying to reach the same people repeatedly.
• Treating DSP and sponsored ads as one number. We've reported early DSP results before we had clean deduplication against sponsored ads in place, and the ROAS looked better than it was. It wasn't dishonest reporting — it was reporting before the measurement caught up to the media plan. Reconciling the two in Amazon Marketing Cloud is what fixed it, and it's why we don't report a DSP number now without checking it against sponsored ads first.

Does the strategy change if you sell on Amazon?

The channel logic is the same — capture existing intent, build new demand, retain. What changes on Amazon is that two of your channels (Sponsored Ads and DSP) report through the same platform and can double-count the same shopper without either report telling you so. Off-Amazon, a display impression and a Google search click are obviously separate systems. On Amazon, they can look like one continuous funnel that's actually reporting the same conversion twice.

The fix isn't a different strategy — it's reconciling the two data sets in Amazon Marketing Cloud before you draw conclusions from either one, and using a holdout or matched-control group to check whether the DSP spend caused a sale that Sponsored Ads would have gotten anyway.