Network status by market in 2026
| Market | Networks | Access model | Key 2025-2026 development |
|---|---|---|---|
| United States | Amazon DSP / AMC, Walmart Connect, Kroger KPM, Target Roundel, Instacart Ads | Spectrum: fully self-serve (Amazon console, Target Roundel) to managed-only for premium placements | Walmart acquired Vizio (CTV hardware, SmartCast OS) for $2.3B to own the living-room OS for shoppable CTV ads. Kroger KPM self-serve expanded mid-2026 with TikTok integration. IAB Europe V2.1 Commerce Media Measurement Standards published January 22, 2026. |
| Australia | Woolworths Everyday Rewards Media (Cartology), Coles 360, Amazon AU | Managed-service dominant for off-site / custom audience activation. Both use LiveRamp Safe Haven. | Cartology ad revenue +19.5% FY; Coles 360 media income +13.5% (normalised ~20.5% FY). Rapid growth; both networks investing in data clean room and off-site capabilities. |
| Canada | Loblaw Advance (Loblaw Ads), Walmart Connect CA, Amazon CA | Loblaw Advance managed for custom audience off-site; Walmart/Amazon operate as in US | Loblaw Advance: 500+ in-store digital screens actively rolling out; MediaAisle DSP for first-party data activation. Grocery duopoly (Loblaw/Empire/Sobeys) makes participation partly mandatory. |
| Singapore | Shopee Ads SG, Lazada Ads SG | Self-serve CPC bidding via platform portals | No data clean room access, no off-site programmatic equivalent to Western grocery networks. CPC inflation during 11.11 / 9.9 / 12.12 events is the primary budget-defensibility challenge. |
| Malaysia | Shopee Ads MY, Lazada Ads MY | Self-serve CPC bidding | Same model as SG. No verifiable published ROAS or CPM benchmarks for SG/MY networks. Cross-border commerce dynamics with adjacent markets add complexity. |
ROAS reality: what the numbers actually mean
The retail media industry benchmark figures circulate widely and are regularly misapplied. Two numbers need to be understood together to use them honestly.
Platform-reported blended ROAS
6.1x
Skai Q1 2025 Retail Media Trends Report. Includes view-through attribution (credit for purchases that may have occurred organically). Amazon-specific: ~3.4x, down ~3.9% YoY.
Incremental-adjusted ROAS
2.4-4.3x
Skai / Osmos.ai, 2025-2026. Measures only lift attributable to the ad, excluding organic velocity the platform claims via view-through windows. Use this range as your decision benchmark.
General e-commerce last-click baseline
2.87x
Skai Q1 2025. Context: retail media at 6.1x platform-reported looks compelling vs. 2.87x generic last-click. At 2.4-4.3x incremental-adjusted, the margin is narrower.
Retail media's core structural advantage is deterministic purchase-signal matching. Amazon knows you searched for a product, saw an ad, and bought it (or did not) within their platform. Woolworths and Coles know your loyalty card purchase history across 52 weeks. Walmart knows what you put in your physical cart, not just your digital one. None of this individual purchase-level data is available to Meta or Google. This gives retail media a more accurate bottom-funnel attribution signal than either probabilistic or modeled approaches.
The structural risk: platforms grade their own homework. View-through attribution windows (typically 1-day post-view, 14-day post-click per IAB/MRC guidelines) allow platforms to claim credit for purchases in-flight organically. The ROAS figure in your platform dashboard is not independently audited. IAB Europe published V2.1 of its Commerce Media Measurement Standards on January 22, 2026, with updated guidance on attribution and measurement for retail media, but standardisation across all major networks is not yet complete.
The halo effect problem
Retail media campaigns lift organic sales among non-exposed shoppers through shelf-space confidence, stock signaling, and brand salience. Platform-reported ROAS does not capture this halo. This means true ROI may be higher than reported, but it also means you cannot credit the platform with organic lift it did not cause. A geographic or temporal holdout test (dark period) is the only way to measure true incrementality accurately.
CPM ranges and budget architecture
| Placement type | CPM range | Context vs. open programmatic |
|---|---|---|
| On-site display (standard) | $5-$25 | 3-5x premium over open exchange ($1-$5 CPM). Premium justified by deterministic purchase signal. |
| Off-site / CTV | $25-$60 | Amazon DSP off-site specifically: $12-$18 for display/OLV; $25-$60 for CTV. Do not apply a blended off-site figure. |
| In-store digital screens | $10-$50 | Emerging channel. Loblaw Advance (Canada): 500+ screens rolling out. Format-specific CPMs vary widely. |
| Sponsored products (effective CPM) | $10-$40 | CPC-billed, converted to effective CPM for cross-format comparison. Category competitive intensity drives the range. |
Minimum spend to note: Target Roundel self-serve (Roundel Media Studio) has no minimum spend and no platform fees for Target Product Ads. This is the most accessible entry point for brands testing retail media. For managed-service placements (premium on-site, off-site programmatic, CTV, in-store), practitioners report $50,000-$100,000+ per campaign minimums at most major networks. These are not officially published.
The margin math brands keep getting wrong: A 3.0x platform-reported ROAS does not mean profitability. Example for a brand with 30% gross margin: the platform ROAS is measured against gross revenue. From that revenue, the retailer takes a wholesale deduction (typically 15-20% of revenue for consumer goods). Platform fees or CPC costs further reduce effective margin. After both deductions, a naive 3.0x platform ROAS on a 30% margin product can produce negative Net Margin Contribution. Calculate NMC rather than ROAS as the decision metric, and calibrate your ROAS floor to account for both the wholesale structure and the platform cost.
The six mistakes brands make on retail media
Entire budget on on-site placements only
On-site inventory exhausts within a category. Once you reach saturation on impression share, additional on-site spend returns diminishing results. Off-site programmatic (using the retailer's purchase data to target shoppers on CTV, display, and social) is where the audience scales. The catch: off-site CPMs are 3-5x higher than on-site, and off-site conversion rates are approximately 20x lower (on-site roughly 10-12%; off-site roughly 0.5% per practitioner estimates). Off-site retail media works when audience targeting is sufficiently precise to justify the CPM premium and the conversion rate difference.
Applying the wrong ROAS math
Platform ROAS is reported net of platform fees but before wholesale deductions and cost of goods. Applying the same ROAS target you use for Meta or Google to retail media ignores the wholesale margin the retailer takes before you see revenue. Calculate Net Margin Contribution as the performance metric: revenue from attributed sales minus cost of goods minus retailer wholesale deduction minus platform fees. A 4.0x platform ROAS on a 25% margin product with a 15% wholesale deduction may be marginally profitable or break-even, not the growth engine a 4.0x headline suggests.
Not using AMC to exclude recent purchasers
Amazon DSP off-site campaigns without exclusion audiences retarget users who purchased in the last 30 days at CTV CPMs. You are paying $25-$60 CPM to show shoppable CTV ads to customers who already bought. Amazon Marketing Cloud's exclusion audiences are the fix: build an audience of recent purchasers and suppress them from off-site DSP. This requires SQL proficiency and query-compute budget, but for accounts spending over $200,000 per year on Amazon media, the wasted spend on non-incremental impressions typically exceeds the AMC operational cost.
Accepting platform ROAS without holdout validation
Platform-reported ROAS includes organic sales credited via view-through attribution. Without a holdout test (geographic pause or temporal dark period), you cannot know how much of your attributed revenue would have occurred anyway. The mismatch between platform-reported and incremental-adjusted ROAS (6.1x vs. 2.4-4.3x at the industry level) represents the gap between what the platform claims and what the ads actually caused. Run a geographic holdout test on at least one major platform per year to calibrate your actual incrementality.
Treating SG/MY retail media identically to US/AU retail media
Shopee Ads and Lazada Ads in Singapore and Malaysia are CPC-bidding search placement platforms, not full-stack retail media networks. There is no data clean room, no off-site programmatic equivalent, and no independently verified attribution methodology. Budget defensibility during major sale events (9.9, 11.11, 12.12) requires day-parting, bid caps, and category exclusions during the peak bidding windows. The post-event attribution window inflates apparent ROAS because high-velocity organic purchase behavior coincides with the campaign run dates. Use temporal holdout testing (pause before or after a sale event and measure the organic baseline) to separate paid lift from event-driven organic velocity.
Not aligning CMO and CRO on Trade Spend vs. Media Spend
Retailers deliberately structure their commercial relationships to encompass both JBP trade commitments (slotting fees, promotional funding, booked as revenue deductions by the Sales team) and digital media placements (booked as marketing OpEx by the Marketing team). When Marketing funds a retail media campaign and Sales simultaneously agrees to a virtual endcap with the same retailer, the brand bids against itself within the same retailer's ecosystem using budget from two separate P&L lines. Brands addressing this in 2026 are moving to a unified Commerce Budget, jointly owned by the CMO and the Chief Revenue Officer, with a shared investment framework across trade spend and media spend.
Five-market strategy differences
United States: full-stack retail media. The US is where the complete retail media stack exists: on-site search, off-site programmatic, CTV via Walmart (Vizio/SmartCast OS) and Amazon DSP, in-store screen inventory, and data clean room access via AMC and Kroger's 84.51-degree data science. Brands operating at scale in the US should be using AMC for audience deduplication before running off-site campaigns, and should have a clear on-site / off-site allocation strategy that accounts for the on-site impression share ceiling.
Australia: grocery duopoly as distribution infrastructure. Woolworths Cartology and Coles 360 together represent the Australian grocery retail media duopoly. Both are growing rapidly and both require managed-service engagement for off-site and custom audience activation via LiveRamp Safe Haven. The strategic frame for AU brands: evaluate retail media spend as part of the joint business plan with each retailer, not as a standalone media performance decision. The question is not only "what is the ROAS" but "what does this investment mean for shelf placement, distribution velocity, and the commercial relationship."
Canada: grocery duopoly plus in-store screens. Loblaw Advance is the primary retail media network in Canada, with the MediaAisle DSP for first-party data activation and a rollout of 500+ in-store digital screens. The Loblaw/Empire/Sobeys grocery structure in Canada creates a similar dynamic to Australia: retail media participation is partly mandatory for shelf coverage. No Canada-specific ROAS benchmarks are publicly available; US figures are directionally applicable.
Singapore and Malaysia: CPC defense, not programmatic scale. The SG/MY retail media landscape is Shopee and Lazada, operating as self-serve CPC-bidding marketplaces. No data clean rooms, no off-site programmatic equivalent, and no independently published ROAS or CPM benchmarks exist for these networks. The strategy in SG and MY is fundamentally defensive: maintain on-site search impression share in your category, manage bid caps to contain CPC inflation during event periods, and use temporal holdout tests to measure organic baseline. Do not apply US or AU retail media frameworks to SG/MY planning; the structures are materially different.