Custom Ad Format Benchmarks in Retail Media (2026)

Last updated: August 2026.

Reviewed by Kunal Damgude, Growth and Product Marketing Manager.

Retail media does not have a native ad unit that can be benchmarked on its own. Sponsored products are assembled by the platform out of the product listing itself, so they already present as part of the shopping surface, and the format story actually worth benchmarking in 2026 is a different one: retailers are building their own ad units and charging a premium for them. Criteo launched auction-based display in June 2025, and as of its Q2 2026 results the unit is live with more than 85 retailers globally, up from 60 in the first quarter, per PPC Land's coverage of those results. For the full return-on-ad-spend matrix by platform, format and vertical, our ROAS Benchmarks by Platform and Ad Format (2026) hub carries it. This page covers the units a retailer builds itself.

Retail Media Has No Separate Native Ad Unit to Benchmark

Native advertising is a paid placement styled to match the design, tone and behaviour of the page carrying it. On the open web that distinction earns its keep, because the alternative is a banner that looks nothing like the content around it. Native display is a large planning category on exactly that basis: eMarketer puts US native display ad spending at $147.98 billion in 2026, growing 13.1% year over year.

Inside a retail media network the distinction collapses. A sponsored product ad is assembled by the platform from the catalog listing the retailer already publishes, the image, the title, the price and the star rating, so the advertiser never uploads a creative and never opts into a native version of the unit. Nielsen reaches the same conclusion from the buy side, describing sponsored products as "the most native of all ad formats on RMNs" in its analysis of how brands split retail media budgets.

That is a statement about one format, not about onsite advertising as a whole. Sponsored brands, onsite display and onsite video all run on an asset the advertiser uploads, and those are the formats where creative control, and therefore format design, actually sits. The standard five-format taxonomy is set out in Sponsored Ad Formats in Retail Media.

The published native benchmarks are real, they simply measure something else. eMarketer sets the performance threshold for a native placement at a click-through rate above 0.4% and reports an 18% lift in purchase intent against standard display. Both are open-web display comparisons, drawn from environments where a native unit is being measured against a banner. Neither says anything about what a shopper does on a product listing page, and neither transfers to a retail media rate card, which is why an advertiser presenting them as the benchmark for your inventory is presenting the wrong benchmark.

The practical consequence for a retail media team is that there is no native line to put on a rate card and nothing to benchmark against display or video. What does exist, and what moved measurably in 2026, is the unit a retailer designs and builds itself.

The Evidence: Retailers Are Already Building and Pricing Their Own Ad Formats

A custom or differentiated ad format is an ad unit the retailer designs and builds on its own surfaces rather than selecting from a vendor menu, which makes the inventory unavailable anywhere else and therefore priceable on its own terms. Three 2026 builds put evidence behind that.

Criteo auction-based display: reading the adoption curve

Criteo introduced auction-based display in June 2025 to bring programmatic bidding into retail environments. The slope matters more than the total here: the unit went from 60 retailers in the first quarter of 2026 to more than 85 in the second, roughly 25 additions in three months, and PPC Land reports that Criteo describes it as its fastest growing advertising format. Standard formats rarely move that fast in a year.

The platform financials belong next to that number rather than instead of it. Criteo's own Q2 2026 release reports Retail Media revenue of $47.9 million, down 21%, or 22% at constant currency, from $60.9 million, reflecting a $21 million headwind from previously communicated scope changes with two specific Retail Media clients, partially offset by strong growth across the broader retail partner base. Excluding that impact, the release states, Contribution ex-TAC, meaning contribution after traffic acquisition costs, grew 20% across the underlying client base.

A vendor can lose two large accounts and keep adding retailers to a new format in the same quarter. Revenue and adoption answer different questions, and for a retailer deciding whether differentiated inventory actually sells, adoption is the one carrying information.

Two more retailer-built units shipped in 2026

Costco Velocity opened beta access to Reserved Display in Q2 2026, a Moloco-powered unit placed on the homepage and search results (eMarketer, 20 March 2026). Home Depot Orange Apron Media introduced high-impact Category Leaderboards on category landing pages at its third annual InFronts on 22 April 2026, alongside a self-serve Brand Page Portal and geotargeted banner ads (Home Depot corporate newsroom).

Both builds get their fuller account in Sponsored Ad Formats in Retail Media, where they sit inside the format taxonomy. The shape of the evidence is what matters here: two of the largest retailers in North America put engineering behind units that exist only on their own surfaces, in the same half-year that a third-party version of the same idea crossed 85 retailers. Where each surface can carry a bespoke unit, and how the build sequence runs, is covered in Play 3: Create differentiated, high-impact ad formats.

The Benchmarks Behind a Rate-Card Floor (2026)

A rate card only holds against numbers the buyer already accepts, so the starting point is what the market is paying and who is collecting it.

The demand pool and where it concentrates

US advertisers spent $60.32 billion on retail media in 2025 and will spend $71.09 billion in 2026, according to a December 2025 EMARKETER forecast, a rise of 17.8% year over year. For a retailer setting rates, the concentration matters more than the total, because it says how much budget is in play outside the three networks that set buyer expectations.

Share figures per eMarketer. Amazon sets the reference the rest of the market is measured against, and the reference keeps rising: Amazon reported $19.8 billion in advertising services revenue for Q2 2026, up 26% year over year from $15.7 billion, its fastest growth across the six quarters in its supplemental disclosures (PPC Land, 31 July 2026).

Amazon also disclosed first performance data for sponsored prompts inside its conversational shopping surface, which became billable in March 2026. Shoppers who click a sponsored prompt "convert to a sale 48% more often and spend 21% more on average than those who don't," chief executive Andy Jassy said. A surface the retailer built, reported as its own line, with its own performance story. That is the same move Costco and Home Depot made, at a different scale.

The platform numbers advertisers will quote at you

An advertiser negotiating a rate for a unit you invented will anchor on what it pays elsewhere. These are the numbers it brings, with their true vintage attached, because the age of a benchmark is part of its weight in a negotiation.

Sources for the table: RMIQ Walmart Connect guide, RMIQ Instacart guide, Sequence Commerce 2026 Amazon benchmarks and Improvado's aggregated network data. The Sequence Commerce figures are an agency aggregate published as directional rather than as a census, and the Walmart video multiple is RMIQ's own best case for optimised campaigns.

Two things are worth taking from the table beyond the numbers. Inside a single platform the CTR ladder falls by roughly a factor of five between search-adjacent placements and display, which is why a high-impact bespoke unit cannot be sold on a display benchmark and cannot be sold on a search benchmark either. And the Instacart line has not moved in over a year: no 2026 refresh of Instacart ROAS or CPC has been published by Instacart or by a top-tier analyst, so an advertiser quoting $5.25 in late 2026 is quoting first-half 2025 data. Say so in the room.

Offsite performance, as context rather than a custom-format claim

Offsite results published in 2026 tell a retailer what its extended inventory is worth, with one qualifier: they come from a supply-side platform brands buy through, not from a retailer building its own unit. TripleLift's June 2026 vertical results report a beauty campaign delivering 11x higher ROAS, 79% lower eCPMs and 121% higher incremental reach against an Amazon Publisher Direct benchmark, and a home and living campaign delivering a 1.87x increase in ROAS against benchmark with a CTR 73% higher than goal.

"Offsite retail media is no longer a supplemental tactic, it is a core growth engine for brands that want to meet consumers where they are across the open internet," said Taylor Stewart, VP Growth and Emerging Channels at TripleLift (PR Newswire, 22 June 2026).

The pricing-model depth behind onsite and offsite rate cards, CPM against CPC against CPA and the revenue-share structures underneath them, is in retail media monetization models. The full ROAS breakdown by platform, format and vertical stays in the ROAS benchmarks hub.

How to Price a Differentiated Format Without an Industry Multiple

No independent benchmark for what a custom, retailer-built ad format clears has been published. Premium multiples do circulate, but they trace back to vendor and publisher copy rather than to a dataset, and a rate you cannot source is a rate you cannot hold in a renewal conversation. Four anchors hold up in place of a multiplier.

1. The floor comes from your own numbers. The conversion rate your existing placements deliver, your average order value, and the return an advertiser needs to keep buying, together set the most a click or an impression can be worth to that buyer. Price under it and you are giving inventory away. Price over it and the campaign stops working, the budget leaves, and the rate was never real. This arithmetic runs entirely on data you already hold, which is the point: a bespoke unit has no external comparable by definition.

Worked on illustrative numbers, so the shape is visible: if your sponsored placements convert at 10% and your average order value is $45, a click is worth $4.50 in order value to the advertiser. At the 4x return that advertiser is holding you to, it can pay up to roughly $1.12 for that click before the buy stops working. Note where that lands, a few cents under the $1.18 to $1.22 Amazon reference band. That is the conversation: your floor is a function of your own conversion rate and basket, and if it sits below the reference band, the gap is what you have to close with performance rather than with a price argument.

2. Published floors exist for adjacent inventory. Instacart sets a $15 CPM minimum on display and video. It is a real, published number for shoppable display and a defensible place to open a conversation about a richer unit on your own surfaces.

3. The market already pays more for controlled inventory. In retail and ecommerce display, Digital Applied's 2026 benchmark set puts Google Display Network CPM at $2.10, open programmatic at $4.30 and private marketplace at $5.95, up about 5% year over year, and reports that brand-safe inventory rated viewable under Media Rating Council (MRC) standards carries an 18% to 22% CPM premium buyers accept. That is an open-web display benchmark, not a retail media rate card, and it should never be quoted as one. What it establishes is direction: the more control and scarcity attached to a placement, the higher it clears.

4. Adoption is the willingness-to-pay evidence. Media spend inside Criteo's Retail Media business reached $527 million in Q2 2026, up 31% at constant currency, with more than 4,500 global brands active on the platform, per PPC Land's reporting. Brand money is funding differentiated retail inventory at scale. That is the argument to bring to a rate conversation, and it is sourceable, which a multiplier is not.

The pool a bespoke unit competes inside is also smaller than the headline market. Per Nielsen's most recent published breakdown, from mid-2025, brands globally split their Amazon retail media budgets 40% to sponsored products, 24% to sponsored brands, 20% to display and 16% to video. A high-impact custom format competes for that display and video third, not against sponsored products, and demand in the third is real: 72% of retail media advertisers say video is one of their main off-site spending areas, per eMarketer.

Proving ROI on a Format You Built Yourself

A unit nobody else sells has no benchmark to be measured against, which makes the proof the product. Three layers, and an advertiser renewing at a premium rate will want all three.

1. Direct ad revenue. Delivered volume times the rate, net of any demand-partner cost. This is the line the media P&L sees, and the easiest of the three to produce.
2. The advertiser-side signal that decides renewal. Conversion rate, return on ad spend, and new-to-brand share, meaning shoppers new to the advertiser's brand rather than new to your store. This is the layer the brand's own reporting runs on, so if your unit reports on a different basis than the advertiser's other buys, the comparison goes against you by default.
3. Retailer gross margin. What the unit did to the merchandising P&L, not only to the media one. A format that moves more of a high-margin category is worth more than its rate card says, and that is a retailer-side argument no vendor benchmark will ever make for you.

The measurement basis is now settled in one important respect. IAB Europe's Commerce Media Measurement Standards V2.1 set a 30-day lookback window as the default, and the six-month grace period during which retailers and ad tech partners could comply with either the first or the second version of the standards ran to the end of July 2026 and has closed. If a custom format reports on a longer window than 30 days, that is a disclosure to make up front, not a footnote.

Because no external comparable exists, the only credible proof that a differentiated unit created demand rather than captured it is a test run on your own inventory: hold a portion of the eligible audience out, compare, report the difference. The method is set out in incrementality testing in retail media, and what closed-loop reporting will and will not tell you network by network is in closed-loop attribution across Walmart, Amazon and Instacart.

Where Osmos Fits: Measuring and Scaling the Formats You Own

A format you invented generates events no external reporting stack recognises, so the measurement has to come from your own event stream. Osmos records view, click, add-to-cart and purchase as first-party events across onsite, offsite and in-store retail media through the Osmos API Hub. That is what lets a retailer report a bespoke unit's performance without depending on third-party cookies or an identity graph, and it is what makes the holdout test runnable on a unit that has no industry comparable to borrow.

Per Osmos's own platform data, retailers configuring their own ad formats see an 11% increase in yield and a 14% increase in brand wallet share. On the yield side the same platform data shows a 38% increase in fill rate and a 40% improvement in brand adoption, and on the operations side 4x revenue per account executive with 32% more campaigns handled per trafficker. That last pair is usually what decides whether a custom-format programme survives past its first advertiser, because a bespoke unit is trafficked by hand where a self-serve sponsored product campaign is not.

Frequently Asked Questions

Is native advertising a separate ad format in retail media?

No. The catalog listing is the creative for a sponsored product, so there is nothing for an advertiser to opt into and no distinct line for a retailer to price. When an advertiser asks for native, it usually means one of two concrete things: an in-feed placement on your own site, or a content placement bought offsite. Name the unit it is actually buying, price that, and the negotiation gets shorter.

What counts as a custom or differentiated ad format?

A unit the retailer specifies and stands up on inventory it already controls, rather than taking a standard unit off a vendor menu. The working test is whether a brand could buy the same placement on a competitor network. If it could, the format is standard inventory with your logo on it, and it will price like standard inventory no matter what it is called.

How much can a retailer charge for a differentiated ad format?

No independent benchmark for a custom-format premium has been published, so the defensible number is derived rather than quoted: a floor from your own conversion rate, average order value and the return the advertiser needs, then a premium justified by measured performance on the unit itself. Write the first term as an introductory rate, measure it, and re-price at renewal against your own data rather than against someone else's multiple.

Which retailers have shipped their own ad formats in 2026?

Criteo's auction-based display reached more than 85 retailers by Q2 2026, up from 60 in Q1. Costco Velocity put Reserved Display into partner beta in the second quarter of 2026. Home Depot Orange Apron Media announced Category Leaderboards on 22 April 2026. All three attached the new format to a surface the retailer already owned and already had traffic on, rather than buying new inventory to resell.

How do you prove ROI on a format that has no benchmark?

Run a holdout. Split the eligible audience, withhold the format from one group, and report the difference in conversions and sales between the two. Report the lookback window alongside the result, because 30 days is the default in IAB Europe's current standard and anything longer is a material difference the advertiser's own analyst will find on their own.

Do differentiated ad formats work in-store as well as onsite?

Yes, though the commercial logic changes: in-store inventory is sold by place and time rather than by person, so the pricing and targeting model is not the onsite one. The build-out is covered in in-store retail media monetization.

Sources

1. PPC Land, Criteo cuts full-year guidance as Q2 revenue falls 11% to $428 million (5 August 2026)
2. Criteo, Criteo Reports Second Quarter 2026 Results, PR Newswire (5 August 2026)
3. eMarketer, FAQ on retail media networks: how marketers should allocate budgets in 2026 (9 January 2026)
4. eMarketer, FAQ on native advertising: formats, AI opportunities, and the best metrics for 2026 (15 January 2026)
5. Amazon.com, Amazon.com Announces Second Quarter Results (30 July 2026)
6. PPC Land, Amazon advertising gains 26% to $19.8 billion as sports inventory sells out (31 July 2026)
7. eMarketer, Costco Velocity Reserved Display launch coverage (20 March 2026)
8. The Home Depot, Orange Apron Media Unveils New Partnerships, Onsite Offerings, and Certification Program at Third Annual InFronts (22 April 2026)
9. TripleLift, TripleLift’s Offsite Retail Media Delivers Proven Results Across Industries, PR Newswire (22 June 2026)
10. RMIQ, 2026 Walmart Connect Advertising Guide (February 2026)
11. RMIQ, Instacart Ads Advertising Guide 2026 (February 2026)
12. Sequence Commerce, Amazon Advertising Benchmarks 2026: What’s Good by Category (June 2026)
13. Improvado, Top Retail Media Networks 2026: Rankings and Benchmarks (2026)
14. Digital Applied, Display Advertising Benchmarks 2026 (20 April 2026)
15. Nielsen, The Future of Retail Media (June 2025)
16. IAB Europe, Commerce (incl. Retail) Media Measurement Standards V2.1