Retail media and paid social do different jobs for a CPG brand. Retail media can place a product near a retailer shopping decision and connect exposure to activity inside that retailer's measurable environment. Paid social can introduce the proposition, shape demand, and direct people to a DTC page, retailer page, or store locator. Neither channel universally wins.
The next dollar should go to the channel whose job matches the brand's distribution, buying moment, creative readiness, data access, and economic constraint. If the product is unavailable, the destination is weak, or the team cannot define the sales denominator, neither channel is ready for more spend.
This is the allocation question for food and beverage teams: not “Which dashboard shows the higher return?” but “Which channel can perform the next necessary job, and what evidence will let us judge it?”
The direct answer: choose the job before the channel
Use retail media when the immediate job depends on a specific retailer environment: helping an eligible product appear around shopping activity, supporting a retailer-specific launch, or learning from retailer-bounded commerce signals. Use paid social when the job starts earlier or outside that environment: introducing an unfamiliar proposition, testing creative angles, building qualified demand, or routing shoppers among DTC, retailer, and local-availability destinations.
That distinction is directional, not absolute. “Retail media” includes onsite sponsored placements, offsite media, and in-store formats with different inventory and measurement. The IAB/MRC retail media guidelines distinguish placements and require clarity about reported outcomes; the IAB in-store standards separately define in-store formats and store zones. Amazon Sponsored Products are one specific product: Amazon describes them as cost-per-click ads for eligible product listings, with placements in shopping results, on product pages, and in other supported destinations (Amazon Ads). They are not a proxy for every grocery retailer network.
Paid social is also not one job. A campaign can send people to an owned checkout, an Amazon product page, another retailer, or a store locator. Those destinations create different observability, margin, and availability conditions. A social click to Amazon can be measured with Amazon Attribution for eligible sellers, vendors, and their agencies, including paid and organic non-Amazon channels (Amazon Ads). That attribution is useful, but it remains attribution inside a defined system—not proof that the sale would not have happened without the ad.
Write a Channel Job Contract before allocating spend
A Channel Job Contract is a one-page agreement about what the next investment must accomplish and how the team will judge it. It prevents a retailer sales report and a social platform report from competing as if they measured the same population.
Write six fields before comparing channels:
Job: the buyer or commercial change the channel must support.
Inventory: the exact media format and eligible product or audience.
Destination: where a person lands and where they can actually buy.
Constraints: availability, margin, promotion, trade, creative, and operational limits.
Data: the observed, attributed, and experimental evidence available.
Failure mode: the condition that would make more media wasteful or the result uninterpretable.
The contract should be narrow. “Grow awareness” is not enough. “Test whether two product-led messages can generate qualified visits to an in-stock regional retailer page without exceeding the approved contribution loss” is actionable. It names a destination, an availability requirement, an economic boundary, and a learning objective.
Retail media vs. paid social: compare the same dimensions
Dimension Retail media Paid social --------- Typical job Influence or capture shopping activity around a retailer or commerce environment Create or shape demand and route shoppers to an appropriate destination Inventory Onsite, offsite, or in-store formats; rules vary by network and product Feed, video, stories, and other social placements; rules vary by platform Destination Retailer search, product detail, brand store, or other retailer-owned experience DTC page, retailer page, marketplace page, store locator, or content Core constraint Product eligibility, retailer availability, product-page quality, retail terms, and data scope Creative strength, destination continuity, event quality, audience reach, and data loss after retailer handoff Evidence Delivery and retailer-bounded attributed activity; sometimes sell-through or test evidence Delivery, owned-site events, matched events where configured, destination signals, and test evidence Common failure Treating one retailer's attributed sales as total or incremental sales Treating clicks, matched purchases, or modeled outcomes as total retail sell-through
This table does not say one channel has better economics. Retail media may put an eligible product close to a shopping decision, but weak product detail, low availability, an uncompetitive offer, or trade deductions can undermine the result. Paid social may create demand beyond one retailer, but a handoff to a retailer can break the observable path and make platform optimization less informative.
Amazon's campaign reporting combines standard and Amazon-specific metrics intended to describe discovery, research, and purchase activity in its environment (Amazon Ads campaign reporting). The IAB/MRC standard calls for disclosure of attribution windows, sales scope, halo logic, data inputs, and methodological limitations (IAB/MRC). These are reasons to read a metric definition, not reasons to dismiss platform data. Platform data can operate a campaign; it simply cannot become the company's complete sales ledger by default.
Five readiness gates before either channel gets the next dollar
1. Distribution and availability are real
Confirm the product can be bought in the promoted geography and destination during the decision window. For retail media, verify listing eligibility, retailer inventory, and the product-detail experience. For paid social sending people to retail, verify that the retailer page or store locator represents actual availability closely enough to avoid routing demand into an empty shelf.
Door count alone is insufficient. A regional brand may have nominal distribution but uneven in-stock conditions. In that case, national demand generation can create exposure where purchase is impossible. The hyperlocal activation framework covers the separate problem of aligning media eligibility with local availability and market operations.
2. The destination can carry the promise
The ad claim, product package, price, variant, and retailer page should describe the same offer. If an Amazon product detail page has unclear imagery, incomplete content, weak reviews, unavailable variants, or inconsistent price, buying more Sponsored Products traffic may accelerate diagnosis rather than growth.
Paid social has the same continuity test. A compelling food creative that routes to a generic retailer homepage creates extra search work. A DTC landing page with a clear bundle may be a better learning destination, but its economics and buyer population differ from grocery retail. Do not compare the two without labeling that difference.
3. Creative is ready for the assigned buying moment
Retail placements near shopping activity often need fast product recognition, accurate pack presentation, and a reason to choose. Paid social may have more room to demonstrate usage, occasion, taste cues, sourcing, or convenience, but the claim still needs to be supportable and relevant to the destination.
Creative readiness is not the number of files in a folder. It is whether the team has distinct, truthful hypotheses and enough production capacity to replace weak work. If both channels would receive the same generic asset with no placement-specific adaptation, delay the allocation argument and repair the message system first.
4. The evidence layers are named
Separate three classes of sales evidence:
Platform-attributed sales: transactions assigned under a platform or retailer's rules, windows, identity coverage, and product scope.
Retailer sell-through: observed sales or units in retailer data, which can still be delayed, aggregated, incomplete, and affected by distribution, price, promotion, and stock.
Incremental sales: the estimated difference versus a credible no-media counterfactual.
Amazon's definitions explain how its reporting attributes purchases and related sales within specified rules (Amazon Ads Help). Amazon Attribution can add visibility for eligible off-Amazon media such as social, but it does not turn an attributed transaction into causal proof. Likewise, Meta's official Business SDK supports sending web, app, and offline events to Meta from a server (Meta Business SDK); a matched offline purchase can improve measurement coverage without proving incrementality.
For the full source-of-truth architecture, use the CPG retail-sales measurement guide. This article owns the channel decision; that guide owns the data layers, reconciliation, and operating cadence after channels are running.
5. Economics and test feasibility are acceptable
Define one contribution basis across both options. Include net recognized revenue and the variable costs that change with the decision: product cost, fulfillment where relevant, returns, marketplace or retailer fees where known, trade and promotion effects, and media spend. Do not compare a retail-media gross-sales return with a paid-social DTC contribution figure.
Then ask whether a credible counterfactual is feasible. IAB's commerce-media incrementality guidance centers credible counterfactuals, bias control, and separating signal from noise, using experiments, model-based methods, or bounded proxies where appropriate (IAB). If the brand cannot create a credible comparison, the plan should state that incrementality is unknown and cap the decision accordingly.
Four common CPG cases and the next decision
No retailer distribution
Retail media tied to retailer inventory is not ready because there is no eligible retail path to support. Paid social may deserve a bounded test if the brand has a viable DTC destination, truthful creative, reliable conversion events, and acceptable unit economics. If DTC fulfillment or repeat economics are also unresolved, neither channel should receive scale funding. Use the next dollar to fix the offer, operations, or measurement.
Amazon is live, but the product detail page is weak
Do not frame the choice as Sponsored Products versus Meta. First repair the product detail and availability foundation. Amazon says Sponsored Products promote eligible listings on a cost-per-click basis and can appear in shopping results and product pages (Amazon Ads). Sending either Amazon ads or paid-social traffic to a weak detail page contaminates the channel comparison because both options inherit the same conversion constraint.
After the page is ready, retail media can test shopping-context demand while paid social can test proposition-led demand. If social routes to Amazon, eligible brands can use Amazon Attribution to observe Amazon activity from those off-Amazon links (Amazon Ads). Use the same product scope and contribution definition, and keep attributed sales separate from incremental sales.
Regional retail doors with patchy inventory
Broad paid social and broad retailer media can both waste impressions outside viable purchase conditions. Define eligible markets or stores from current availability, then determine which inventory can respect that boundary. Retail media may have retailer-specific proximity to commerce data; paid social may offer creative learning and flexible routing. Neither advantage excuses inaccurate availability.
If availability changes quickly or the retailer file is delayed, make a smaller reversible test. Record in-stock coverage, promotion, and distribution alongside spend. A result without those controls cannot tell whether the channel or the shelf changed.
Strong DTC economics but limited retail sell-through data
Paid social to DTC may be the more observable operating path because the brand controls checkout and post-purchase data. That does not prove social creates more total demand than retail media. It means the DTC path supports a clearer immediate decision.
A separate retail-media test may still be justified for retailer-specific learning if the brand can obtain at least the reported outcome definitions, product scope, attribution window, and enough sell-through context to interpret the result. When retailer data is too incomplete to reconcile, label the result as retailer-attributed or directional rather than converting it into a total-market return.
A hypothetical comparison using one net contribution definition
Consider a fictional snack brand choosing between two bounded tests. These numbers are illustrative only. They are not benchmarks, recommendations, or Sharply Labs results.
Option A: retailer media
Media spend: $20,000
Platform-attributed retail sales: $82,000
Retailer sell-through observed for the eligible products and period: $76,000
Estimated returns and spoilage adjustment: $3,000
Trade and promotion adjustment: $12,000
Product and retailer-variable cost applied to the adjusted sales: $31,000
For the operating view, adjusted net sales are $76,000 − $3,000 − $12,000 = $61,000. Net contribution after the stated variable cost and media is $61,000 − $31,000 − $20,000 = $10,000.
The $82,000 platform-attributed figure is not substituted into that equation because its scope may differ from the retailer sell-through file. It remains an optimization and reconciliation input.
Option B: paid social to DTC and eligible retailer pages
Media spend: $20,000
DTC recognized sales: $48,000
Retailer-attributed sales from tagged links: $29,000
DTC returns adjustment: $4,000
DTC fulfillment, payment, and product variable cost: $22,000
Known retailer trade/promotion adjustment: $5,000
Product cost associated with the retailer-attributed amount: $11,000
The DTC net contribution is $48,000 − $4,000 − $22,000 − $20,000 = $2,000 if all media spend is provisionally assigned there. But adding the $29,000 retailer-attributed amount would double-use the media cost and may mix attribution with sell-through. The retailer portion needs reconciliation against retailer sales before a combined contribution figure is defensible.
Now add the counterfactual guardrail. Suppose comparable untreated markets or a valid experiment suggest that much of the observed retail movement would have occurred anyway, but the design is too weak to quantify a stable lift. The correct conclusion is not that Option A produced $10,000 of incremental contribution. It is that Option A shows positive observed contribution under the stated accounting and unknown incrementality. Option B likewise has an observable DTC component and an unresolved retail component.
The next decision may be to improve the counterfactual, not increase either budget. The numbers make the uncertainty visible instead of hiding it behind incompatible ROAS calculations.
A practical decision sequence
Step 1: assign the next job
Choose one: retailer-specific conversion support, proposition learning, demand creation, local availability activation, DTC acquisition, or measurement learning. Do not assign three jobs to one budget and then judge it with one metric.
Step 2: qualify destinations and inventory
Check product eligibility, inventory, geography, product-page quality, price, promotion, DTC experience, and event delivery. Remove any destination that cannot fulfill the promise now.
Step 3: define evidence and economics before launch
Write the platform-attributed, retailer sell-through, and incremental outcomes separately. Create one net contribution definition, list unavailable adjustments, predefine the decision window, and state the counterfactual method or its absence. Our growth service connects paid execution with these operating and measurement decisions, while the food and CPG practice covers the category context.
Step 4: release a bounded test, then reconcile
Release only enough budget to answer the defined question. Monitor delivery and availability during the test. After the data matures, reconcile source scope, returns, trade, promotions, inventory, and costs. Keep, constrain, change, or stop the channel based on the contract—not the most flattering dashboard.
When neither channel should receive the next dollar
Pause both options when the product is materially out of stock, distribution data cannot identify eligible markets, the retailer or DTC destination contradicts the ad, contribution economics are undefined, creative has no testable proposition, or the proposed result cannot influence a real decision.
Also pause when trade or promotion changes overwhelm the media question, or when teams plan to add attributed sales across systems. Media can amplify a functioning commercial path; it cannot make missing availability, broken content, or incoherent accounting disappear.
The right alternative may be product-page repair, inventory coordination, retailer-data access, event validation, contribution-margin work, or a smaller market design. The CPG agency selection framework is useful if an external partner will own those decisions and data handoffs.
The decision to take into the budget meeting
Do not ask retail media and paid social to win the same dashboard contest. Ask which channel has a qualified job, a purchasable destination, fit-for-purpose creative, interpretable data, acceptable economics, and a credible next decision.
Sharply Labs offers food, beverage, and CPG teams with retailer or DTC distribution a focused channel-job and data-readiness review. The output is a decision on what retail media or paid social should do next, an event-and-denominator audit, and a bounded test and measurement plan with explicit budget criteria. It does not guarantee ROAS, velocity, CAC, ranking, or any other commercial outcome.