Retail Media vs In-House: 2026 Decision Matrix

Retail media is the fastest-growing line in the digital ad budget, and it is also the least trustworthy on its own dashboard. US retail media ad spend reached $60.32 billion in 2025 and is forecast at $71.09 billion in 2026, yet only 15% of marketers say they can measure that spend effectively. The gap between what a retail media network reports and what it actually drives is where budgets quietly leak.

The reason is a single, under-discussed number: incremental ROAS typically runs 30 to 60% below the last-click ROAS that retail networks report. A 6x dashboard figure can be closer to 2.4x in true, lift-over-baseline terms. When 94% of advertisers say they don't fully trust retailer-reported metrics yet most allocation frameworks are built on exactly that data, the decision becomes less about "which channel has the best ROAS" and more about "which channel can I actually verify."

This guide gives you a proprietary, weighted five-factor matrix that scores seven channel types — three pure retail media networks, three owned channels, and TikTok Shop — so you can decide where to allocate first, where to allocate selectively, and what to merely test. Every figure below is sourced and dated; platform ROAS numbers are explicitly distinguished from incremental ones throughout.

US retail media ad spend reached $60.32 billion in 2025 and is forecast at $71.09 billion in 2026, representing roughly 18% year-over-year growth, according to eMarketer's H1 2026 forecast. Retail media now represents about 30% of all US digital ad spending, up from 15% of advertiser budgets in 2022 to roughly 22% in 2025. Globally, the category was approximately $140 billion in 2024 and is projected near $165 billion in 2026, with 277 distinct retail media networks operating worldwide as of November 2025 — a count that changes rapidly and may already be higher.

The concentration is the part most allocation conversations skip. eMarketer projects that Amazon and Walmart will capture 89% of incremental retail media spending in 2026 — about $9.42 billion of the $10.53 billion in net-new RMN investment flows to those two players. Amazon holds an analyst-estimated 75 to 77% of the US retail media market (the exact figure varies by methodology), and its retail media revenue is forecast to exceed $75 billion by 2028. Walmart Connect, the clear number two, holds roughly 6.9% of US retail media spend.

The rest of the field is real but smaller. Instacart Ads surpassed $1 billion in trailing twelve-month ad revenue as of Q3 2025, reaching 95% of North American households across more than 100,000 store locations, and in May 2026 it expanded its Ads Manager to retailers with new self-serve tools. Off-site formats are growing quickly too — eMarketer pegs retail-media-powered social ad spend at $7.76 billion and retail-powered connected-TV spots at $6.10 billion in the US for 2026. Retail media is no longer a tactic bolted onto an Amazon page; it is becoming a full-funnel media layer.

For context on how retail media fits a broader plan, our 2026 channel budget allocation guide covers the cross-channel split this matrix slots into.

Here is the contradiction at the center of retail media in 2026. The category is growing because the dashboards look spectacular — and almost no one trusts those dashboards. Only 15% of marketers report being very or extremely effective at measuring retail media performance, per Skai's 2026 State of Retail Media. Some 75% cite incrementality as their single biggest measurement challenge, 50% measure it only at a basic level, and just 20% describe themselves as proficient at both measuring and applying the insight.

At the same time, the priorities have flipped. According to Osmos.ai, 71% of advertisers now rank incrementality as their number one retail media KPI, ahead of ROAS — but 94% say they do not fully trust retailer-reported metrics. The top barriers they name are accuracy and reliability concerns (44%), difficulty applying measurement across ad types (43%), and limited tools (41%). Forrester found that 86% of commerce media decision-makers in North America and Europe consider strengthening measurement and attribution a high or critical priority.

The verification problem is being worked on, slowly. Kroger Precision Marketing (84.51°) now reports incremental sales for self-service programmatic campaigns; Albertsons Media Collective launched APIs for advertiser integration with measurement models; and Instacart and Criteo have received MRC accreditation. The IAB's Guidelines for Incremental Measurement in Commerce Media define four methodologies — experiments, model-based counterfactuals, econometric models, and hybrid proxies — grounded in three principles: credible counterfactuals, control of bias, and separating signal from noise. IAB Europe's Commerce Media Measurement Standards V2, released in January 2026, carry a six-month grace period running through the end of July 2026; that is a grace-period end date, not a hard compliance switch.

The most important number in retail media is one the dashboards never show you. Per Osmos.ai's 2026 benchmarks, incremental ROAS — the true lift over what would have sold anyway — typically runs 30 to 60% below the last-click ROAS retail networks report. That is not a rounding error; it is the difference between a channel that pays for itself and one that quietly subsidizes sales you were already going to make.

Put concrete numbers on it. Skai reports an overall retail media average ROAS of 6.1x, maintained for five consecutive quarters through Q1 2025 across roughly $9.2 billion in platform activity. Amazon Ads average ROAS sits around 3.4x (down about 3.9% year-over-year), versus a general ecommerce benchmark near 2.87x. Those are platform-reported, last-click figures. Apply the incrementality discount and the picture shifts materially — the chart below shows the same headline numbers alongside an illustrative true range once you subtract that lift gap.

Read that chart with a caveat: the "incremental-adjusted" bars are an illustrative application of the 30 to 60% discount, not a measured per-platform incremental figure. The point is directional, not precise — your real incremental ROAS depends on your category, your baseline demand, and a proper test. But the direction is the whole story. A channel reporting 6.1x that is actually delivering 2.4x in lift is a very different budget decision than the dashboard suggests.

The brands actually measuring incrementality see concrete returns: Skai reports that 54% reduced wasted spend, 49% increased new-customer acquisition, and 29% improved competitive positioning. But there is a persistent gap between aspiration and achievement — profit-margin improvement was achieved by only 24% versus 68% who hoped for it, and customer-lifetime-value improvement by 22% versus 42% who hoped. Measurement maturity, not spend volume, separates the brands that extract value from retail media from those that simply feed it.

This is the same distinction we draw in our media mix vs. attribution decision matrix: last-click attribution flatters the channels closest to the purchase, and retail media sits closest of all. The corrective is holdout testing and modeling, covered in our marketing mix modeling vs. attribution playbook.

Most budget-allocation advice ends at "it depends on your objectives." That is true and useless. Here is the actual framework: five factors, explicit weights, scored 1 to 5 per channel, with the weighted total mapping to a verdict. The weights reflect the 2026 reality that incrementality and measurement trust — not headline ROAS — are what separate good allocation from waste.

The table below applies the five-factor matrix to seven channel types: three pure retail media networks, three owned channels, and TikTok Shop. Owned channels (marked with a dot) tend to score higher on incrementality confidence and measurement clarity precisely because you control the attribution stack and can run your own holdouts.

The pattern is worth interpreting rather than just reading. Google Shopping, an owned channel, tops the matrix not because its raw ROAS beats Amazon's, but because you own the measurement stack — you can run holdouts, you see the full path, and you aren't reliant on a retailer grading its own work. Amazon Sponsored Products and Walmart Connect land in "allocate selectively" territory: unbeatable on audience fit and at-scale intent, dragged down by the measurement-clarity penalty that 94% of advertisers feel. TikTok Shop and open programmatic/DSP fall to "test only" — promising, but harder to verify and easier to waste on.

The forward read is that this ordering is not fixed. As Kroger, Instacart, and Albertsons ship genuine incrementality reporting and independent API access, the measurement-clarity scores for those networks should rise — which would pull them up the matrix without any change in their ROAS. The brands that win in 2026 and 2027 will re-score this table every quarter as measurement infrastructure matures, rather than locking in an allocation built on today's verification gaps.

The measurement-clarity factor in the matrix deserves its own breakdown, because it is the axis where the networks differ most. The comparison below maps five measurement capabilities across the four most-cited platforms. It is a snapshot of a fast-moving area — clean rooms are now effectively standard in retail media deals, and accreditation and RCT support are being added quarter by quarter — so treat the cells as directional and re-verify before committing budget.

The notable takeaway is that the biggest networks are not automatically the most measurable. Instacart and Kroger Precision Marketing (84.51°) have moved early on incrementality reporting and independent integration, while Amazon and Walmart — despite their scale advantage — still report incrementality with limited outside verification, which is exactly why senior buyers describe them as taking full attribution credit. Tooling is consolidating around this: The Trade Desk, Pacvue, and Skai announced a unified activation and measurement integration across more than 250 commerce media partners in April 2026, and 85% of buy-side stakeholders cite access to retailer first-party data as the primary opportunity driving retail media investment.

There is a cost most allocation posts celebrate as a feature: platform diversity. The average brand manages about 6 retail media networks today, projected to expand toward 8 to 11 by the end of 2026, per Skai and Fugo.ai. Without consolidated reporting, each additional network adds disproportionate rather than linear complexity — a new portal, a new taxonomy, a new attribution model, and another place where measurement can't be reconciled against the others.

The operational reality backs this up. Only 12% of brands have integrated media and commerce operations, and 75% cite internal resource constraints as their top performance barrier. Roughly 14% are still not measuring incrementality at all. Meanwhile 52% are moving display budgets from traditional programmatic to retail DSPs, adding yet another surface to manage. This is precisely why operational lift earns its own weighted factor in the matrix: a channel with a strong ROAS that requires a dedicated specialist and a standalone portal can quietly cost more than it returns.

Retail media's structural advantage is intent proximity: the ad sits inches from the buy button, on a surface where the shopper is already searching. That advantage assumes a human is doing the searching. The emerging risk — still forward-looking, not current reality — is agentic commerce: AI assistants that curate the shopping experience and complete purchases on a person's behalf. If the interface shifts from a search bar to an agent, the sponsored-product placement loses much of its leverage.

This is the "watch this space" framing rather than a 2026 prediction. But it should shape how much you lock in versus keep flexible. A measurement-first allocation — one that proves incremental lift rather than trusting reported ROAS — is also the allocation best positioned for a world where the shopping surface changes underneath you. If you can demonstrate that a channel drives real, baseline-adjusted lift, you can defend it through a platform shift; if all you have is a last-click number from a surface that may not exist in the same form in three years, you are exposed.

For brands ready to put this matrix to work — re-scoring it against their own category, building the holdout tests that correct last-click inflation, and consolidating reporting before adding the next network — our paid media services and analytics and measurement engagements start with exactly this kind of incrementality-first allocation review. Creative still matters inside every one of these channels; our 2026 ad creative benchmarks by channel cover what actually moves CTR and ROAS within a placement.