Retail Media Creative Optimization: The Best Practices That Actually Move ROAS (2026)

Last updated: June 2026. Reviewed by Kunal Damgude, Growth and Product Marketing Manager.

You optimize retail media creative by matching the asset to the placement, testing one variable at a time, and measuring the full path from impression to attributed order rather than stopping at the click. Creative is the last controllable lever between an auction win and a sales outcome: the targeting decides who sees the ad, but the creative decides whether they act. On a retail media network, the shopper is already mid-purchase, so the creative that wins is the one that reduces friction to the cart, not the one that tells the prettiest brand story. One scope note first: this is about the formats where you actually upload and control an asset, namely display, sponsored brand, video, and off-site. Sponsored product ads work differently, because the platform builds them automatically from your product listing, so there is no ad creative to upload and the lever there is the quality of your catalog image and detail page. For the formats you do control, programs that treat creative as a fire-and-forget upload plateau at low ROAS, while programs that run placement-native assets, structured A/B tests, clean specs, and systematic scale operations are the ones that pull ahead.

Why Does Creative Quality Move ROAS on a Retail Media Network?

Creative quality moves ROAS because it is the single largest controllable driver of advertising sales lift. Foundational Nielsen research (2017), cited by Recast, found that 47% of a campaign's contribution to sales was attributable to creative, the largest share of any element, ahead of reach, targeting, and recency. That finding predates the retail media boom, but the mechanism is the same: once the auction places your ad in front of the right shopper, the asset itself carries most of the remaining outcome. For the broader context on how this channel developed, see our guide to retail media evolution; this article is the deep dive on the creative layer specifically.

The math is simple to follow. Effective ROAS on an RMN is the product of click-through rate, conversion rate, average order value, and how reliably your ad fills its placement. Creative touches the first three of those directly. A thumbnail that reads cleanly at small sizes lifts CTR; an image that matches the shopper's intent lifts conversion; a clear offer lifts basket size. Small inputs compound, which is why a marginally better asset can produce a disproportionately better return.

Retail media creative also behaves differently from creative on Google or Meta. On a search or social feed, you are interrupting someone who is browsing. On an RMN, you are reaching someone who is already shopping, often with a specific category in mind. That shift in mindset rewards product clarity over persuasion. The market is rewarding this discipline: retail media click-through rates rose 9% in Q4 2025, with clicks up 34% while cost-per-click fell 1% (Skai Q4 2025 Trends Report), evidence that better-performing creative is scaling more efficiently, not just more expensively.

What Creative Works Best by Placement Type?

The best creative is the creative built for its specific placement, because each format rewards a different shopper behavior. Before optimizing within a format, it helps to know the full range of retail media ad formats and where each one sits in the funnel. Here is what performs in each.

Sponsored product. There is no separate creative to upload for this format. The ad is assembled automatically from your product listing, pulling the main catalog image, title, price, and rating, so the lever is the quality of the listing itself rather than an ad asset you design. Because the platform matches that listing to the shopper's query, a clean, high-resolution main catalog image that genuinely represents the product is what protects the intent match: a query for "running shoes" should surface the shoe, not a lifestyle scene. This intent match is why sponsored products convert so well. Equativ reports that Amazon Sponsored Products achieve average conversion rates of roughly 10%, precisely because the placement meets shoppers at the moment of decision.

On-site display and banner. Pair a first-party audience signal with a geo-targeted message, and place the unit above the fold where it earns the most attention. The creative needs a visible brand lockup and one clear offer; a banner that tries to say three things says nothing. If you want background on how these units developed, our piece on the evolution of display ad formats in retail media covers that history.

Off-site. Keep the creative consistent with the on-site experience so the shopper recognizes the brand across the walled garden and the open web. First-party retargeting audiences are what make off-site efficient: you are extending reach to known shoppers rather than buying cold impressions.

Video. Design for autoplay on mute, hook the viewer in the first few seconds, and add captions because most retail video plays silently. Video investment is climbing fast, with Prime Video ad spend up 127% year over year in Q4 2025 (Tinuiti, March 2026) and upper-funnel DSP investment up 72% year over year while costs fell 24% (Skai Q4 2025). On our platform, Adscape Video Ads run an automated video quality check before launch and report play-quartile data, so you can see exactly where viewers drop off and fix the asset rather than guess.

In-store and DOOH. In-store screens give creative a longer dwell window than a quick-glance digital banner, which lets the asset carry richer product detail and comparative content. Placement is strategic: the creative has to account for the shopper's mindset in that part of the store and the traffic flow past the screen. With Adscape in-store ads, section-level targeting puts the right creative in the right aisle, and QR-enabled tracking bridges that in-store exposure to online conversion so the format stops being a measurement black box.

How Does Dynamic Creative Optimization (DCO) Work in Retail Media?

DCO assembles an ad from modular components in real time, tailoring parts of the creative to each impression instead of serving one fixed version to everyone. Amazon Ads defines it directly: "DCO stands for dynamic creative optimization. In advertising, DCO technology rapidly builds multiple iterations of an ad using the same base creative, while tailoring parts of the ad based on audiences, context, and past performance." RMIQ frames the retail media application the same way: DCO "dynamically assembles ad elements in real time using data and algorithms to deliver personalized, contextually relevant messages to each prospect."

In a retail media context, the most valuable thing DCO personalizes is the product data itself. A dynamic product ad pulls the live price, the current image, and in-stock status straight from the catalog, so a shopper never sees an ad for a sold-out item or a stale price. On our platform, Adscape PDA (Product Display Ads) delivers this as a built-in capability: inventory-synced dynamic product listings with real-time price updates and keyword targeting, which is native DCO for product creative without bolting on a separate DCO vendor.

DCO earns its keep when complexity is high. If you run thousands of SKUs, sell across regions with different pricing, or operate in price-sensitive categories where a stale number kills trust, dynamic assembly beats static creative because no human can hand-build and maintain that many accurate variants. Skai notes that brands incorporating shoppable video, dynamic creative optimization, and personalized ad variations see significantly higher engagement. Vendors that run DCO platforms report ROAS gains above 50%, which is directional rather than a guaranteed benchmark, but the direction is consistent: serving each shopper the most relevant, accurate variant lifts performance. For a smaller catalog with stable pricing, a few well-made static assets can match it, so DCO is a fit decision, not a default.

How Do You Run Creative A/B Tests on a Retail Media Network?

You run a clean A/B test by deciding what counts as a win before launch, changing one element at a time, and reading the result on the conversion metric rather than the click. The discipline matters more than the tooling. As Flywheel Digital put it in their February 2026 creative testing guidance, you should focus on only one or two creative elements per test, and results are "measured in weeks, not quarters." Retail media gives you faster feedback than brand advertising, but only if the test is built to isolate a single cause.

Start by defining the win condition. Decide upfront whether success is a lift in conversion rate, units per order, new-to-brand share, or repeat purchase over the following month or two, then hold the test to that metric. Without a pre-set criterion, you will cherry-pick whichever number looks good after the fact.

Then isolate the variable. Test headline against headline, or image against image, or call-to-action against call-to-action, but never several at once, or you cannot attribute the difference. Let each variant accumulate enough impressions before you read it; a thousand impressions per variant is a reasonable working floor, not an industry-certified threshold, so treat it as a starting point and require more volume for low-CTR placements.

For rotation, you have two modes. Equal-weight rotation splits delivery evenly and is the honest setting while a test is running, because it gives every variant a fair sample. Performance-weighted rotation shifts delivery toward the winner and is the setting you switch to once a test has resolved, to harvest the result. The common mistake is running performance-weighted rotation during the test itself, which starves the challenger of data and locks in whatever happened to win early. Finally, read the verdict on the right metric: CTR tells you the creative earned attention, but conversion-to-order and ROAS contribution tell you it earned revenue, and those are the numbers that decide the winner.

What Are the Most Common Creative Mistakes That Waste Retail Media Spend?

The most common mistake is reusing one generic asset everywhere, which violates the placement-native rule and quietly drags down every format it touches. Skai names this directly: "Many brands use generic product images and templated ad formats, resulting in lower engagement and reduced ad performance." The fixes below are the recurring offenders worth auditing first.

• One creative across all placements. A display banner and an upper-funnel video need different assets. Using the same file for both undersells both.
• A weak main image on a promoted product listing. Sponsored product ads pull the listing's main catalog image, so when you promote a product, a moody lifestyle shot or a pixelated image there breaks the intent match and the conversion. The fix is catalog content, not an ad upload.
• Missing brand lockup on display. Without a clear logo, the unit cannot contribute to brand recall measurement, so you lose the one thing display does well even when clicks are low.
• Spec non-compliance. Wrong dimensions, oversized files, the wrong aspect ratio, or a non-white background where the platform requires one will get the asset auto-rejected before it ever serves. Every major retail platform publishes its own spec sheet, and the fastest way to waste a launch window is to skip it.
• Set-and-forget creative. Static assets fatigue fast in the concentrated, highly targeted audiences typical of retail media. Performance degrades within weeks, so a campaign left untouched for a month or two is usually decaying.
• Offer mismatch between creative and landing page. If the ad promises a discount the product page does not honor, you pay for the click and lose the conversion, and you erode trust on top of it.

How Do You Measure Creative Performance Properly?

You measure creative properly by reading three metrics in order: click-through rate as the appeal signal, conversion rate as the intent signal, and attributed ROAS as the verdict. CTR alone is a trap. A high CTR with low conversion means the creative is attractive but mismatched to what the shopper actually wanted, and you are paying for clicks that do not convert. The retail media CTR lift of 9% in Q4 2025 (Skai Q4 2025) is encouraging at the market level, but your own funnel is what matters, and the click is only the first step of it.

The metric most operators get wrong is view-through. Click-through attribution credits a sale to a shopper who clicked the ad; view-through attribution credits a sale to a shopper who saw the ad, did not click, and converted later. Display and video creative do a lot of their work through view-through, so judging them on clicks alone systematically undervalues them and tempts you to cut formats that are quietly contributing. Read display and video on view-through plus assisted conversions, and reserve last-click strictly for the bottom-funnel formats where a click is the expected action.

Operators and brands also read the same report differently, and a good measurement setup serves both. A brand looks at ROAS, conversion rate, and new-to-brand share to judge its own spend. An operator running the network looks at fill rate and yield to judge the health of the inventory. The reporting layer has to expose both. Our Display Ads reporting gives full-funnel visibility across campaigns, inventory, and individual banners, down to detailed order logs showing the SKUs and quantities behind each attributed order, which is what lets you trace a creative decision all the way to revenue. For the wider ranges that tell you whether your numbers are healthy, our retail media performance benchmarks give the comparison points.

How Do You Manage Creative at Scale Across Hundreds of Brands and SKUs?

You manage creative at scale by automating the review workflow, standardizing what each advertiser can change, and templating production so cycle time stays flat as volume grows. Manual review is the bottleneck that caps throughput first, and the data shows the gap: nearly two-thirds of advertisers now use generative AI in retail media, yet creative and messaging requests are only 3% of those AI prompts (Skai Q4 2025), which means the creative operation is still mostly manual even as planning gets automated. That is the gap an operator has to close.

Automated compliance is the first lever. On our platform, ControlHub Content Cop runs an AI-powered content quality check that validates creative against spec and brand rules before it enters the human queue, replaces email approval chains with two-way in-platform communication, and works on mobile so approvals do not stall on desktop access. That cuts the spec-rejection cycle that otherwise consumes a trafficker's day.

Governance is the second lever. Brand Jukebox lets an operator set tiered creative exposure, standardizing which formats and creative features each advertiser tier can access, so brand-inconsistent or off-spec creative never reaches the review queue in the first place. Pair that with centralized brand guidelines and template-based production, where advertisers build inside pre-approved frames, and you compress creative turnaround from days to hours. The throughput payoff is concrete: traffickers on our platform manage 32% more campaigns each, revenue per account executive rises 4x, and time-to-completion improves 40%, which is how a network scales creative across hundreds of brands without scaling headcount at the same rate. These same operational pressures shape retail media platform selection when an operator decides whether to build or buy. Across our primary markets in the US and India, the format taxonomy is consistent: Flipkart, JioMart, and Meesho run the same sponsored-product, display, and video building blocks, so the operating playbook ports across regions.

Frequently Asked Questions

What is the best image size for retail media display ads?There is no single best size, because each platform sets its own. The standard display dimensions you will see most often are 300x250, 728x90, and 160x600, and supporting those covers most placements. Beyond those, Amazon, Walmart, and Target each publish their own required dimensions, file-size limits, and image rules, so the reliable move is to pull the current spec from each platform's official advertising spec page before you build, rather than trusting a third-party list that may be stale.

How often should you refresh retail media ad creative?Refresh on the signal, not the calendar. Watch the CTR trend, and when it declines consistently over a week or two, the creative is fatiguing and should be rotated. For high-frequency placements that hit the same concentrated audience repeatedly, a working cadence of every 30 to 45 days keeps assets from going stale, but the CTR trend is the real trigger and should override any fixed schedule.

Does creative quality affect retail media auction ranking?Yes, indirectly but materially. Most auction models reward relevance, and the clearest relevance signal a creative sends is its click-through rate. Higher-CTR creative tends to earn better effective CPMs, which means a better asset can win more impressions at the same bid. Creative quality is not a separate ranking input you set; it shows up in the auction through the performance the creative produces.

What is a good CTR for retail media sponsored product ads?Sponsored product CTR usually sits in the low single digits, varying by category, query, and platform, while display CTR runs lower because the intent is weaker. One unit caution: the roughly 10% figure Equativ reports for Amazon Sponsored Products is a conversion rate, what share of clickers buy, not a click-through rate, so do not confuse the two. For grounded ranges on both, our retail media performance benchmarks break them out by format.

How do you get brands to submit spec-compliant creative?Make the rules unambiguous and check compliance before a human ever looks at the asset. Clear, current spec documentation plus an automated pre-submission check catches most violations at the door. On our platform, ControlHub Content Cop runs that automated quality check and routes anything that fails back to the advertiser through in-platform communication, so rejections resolve in the tool instead of in a back-and-forth email thread that costs days.

Want to see how a full creative surface and the operating layer behind it work together? Explore Adscape and ControlHub, part of the Osmos retail media operating system.

Sources

1. Skai. "Q4 2025 Digital Advertising Quarterly Trends Report." January 2026. https://skai.io/press-releases/quarterly-trends-report-q4-2025/
2. Tinuiti. "Retail Media Trends and Outlook for 2026." March 2026. https://tinuiti.com/blog/commerce/retail-media-trends/
3. Flywheel Digital. "Retail Media: Creative Testing Lab for Brands." February 2026. https://www.flywheeldigital.com/blog/retail-media-creative-testing-lab
4. Amazon Ads. "Dynamic Creative Optimization (DCO) Definition and Guide." https://advertising.amazon.com/library/guides/dco-dynamic-creative-optimization
5. RMIQ. "Dynamic Creative Optimization (DCO) Definition." September 2025. https://www.rmiq.net/glossary/dynamic-creative-optimization/
6. Equativ. "From Click to Conversion: Sponsored Product Ads Explained." August 2025. https://www.equativ.com/blog/from-click-to-conversion-sponsored-product-ads-explained
7. Skai. "The 5 Biggest Retail Media Challenges for Beauty Brands." Updated December 2025. https://skai.io/blog/retail-media-challenges-for-beauty-brands/
8. Recast. "Creative Performance Drives 47% of Sales." 2022 (citing Nielsen, 2017). https://getrecast.com/creative-performance/