Retail Media ROI: The Formula, the Benchmarks, and Where Brands Get It Wrong

Retail media ROI can look strong in-platform and still fail a finance review. With U.S. retail media spend nearing $70 billion in 2026, the gap between platform ROAS and profit is now too large to ignore. This guide breaks down how retail media ROI should be calculated, where benchmark data can mislead, and how brands can judge incremental profit with more discipline.

TL;DR

• Retail media ROI should be based on incremental profit, not platform-attributed revenue.

• ROAS and ROI are not the same metric, and mixing them leads to bad budget calls.

• Retail media benchmarks are directional because each network uses different attribution rules.

• Margin, trade overlap, halo sales, and incrementality change the answer more than most dashboards show.

• A finance-ready scorecard should show modeled profit next to reported ROAS.

Why retail media ROI often breaks under finance review

Retail media reporting usually starts with ROAS, because that is the number platforms show first. The problem is simple: ROAS measures attributed revenue per ad dollar, while finance teams need profit after costs.

That gap can be wide.

Some estimates put platform ROAS 40% to 60% above incremental ROAS. In one case, a campaign showed 14:1 ROAS in platform reporting, while a broader model put return at just 0.4:1. That is not a rounding issue. It is a measurement issue.

The main reason is that platform reporting gives credit for sales that may have happened anyway. That includes:

• branded search demand

• repeat buyers already on path to purchase

• sales lifted by price cuts or coupons

• halo volume mixed into direct product sales

When those sales are counted as media-driven, return looks better than the P&L shows.

The retail media ROI formula is simple, but the inputs are not

The base formula is:

Retail media ROI = ((Incremental Profit − Retail Media Spend) / Retail Media Spend)

To get there, brands need one more step:

Incremental Profit = Incremental Sales × Net Contribution Margin

That means a brand cannot stop at revenue. It has to convert sales lift into profit after variable costs.

At a minimum, net contribution margin should account for:

• cost of goods sold

• retailer fees

• fulfillment costs

• discounts

• returns

• media spend

• data costs

• production costs tied to the campaign

A short example makes the point.

If a campaign drives $500,000 in incremental sales and the net contribution margin is 25%, incremental profit is $125,000. If ad spend was $100,000, then:

ROI = (($125,000 − $100,000) / $100,000) = 0.25

That is a 25% ROI, even though headline revenue may have made the campaign look much stronger.

Revenue does not settle the case. Profit does. This requires optimizing spend with marketing mix modeling to account for all variables.

ROAS, incrementality, and true ROI answer different questions

ROAS vs. True ROI vs. Incrementality: Retail Media Metrics Compared

A lot of retail media confusion starts here. These metrics are connected, but they are not interchangeable.

ROAS answers: How much attributed revenue did the platform assign to the spend?

Incrementality answers: Would those sales have happened without the ad?

True ROI answers: After costs, did the campaign add profit?

That split matters because each metric serves a different job.

• ROAS is useful for bid and pacing decisions.

• Incrementality is used to test lift.

• ROI is used for budget defense and P&L review.

When teams use ROAS to do ROI’s job, the output gets shaky. A retailer dashboard may help a media team move spend during a campaign, but it does not answer the CFO’s question.

A clean way to think about it:

• ROAS shows platform efficiency

• incrementality shows causal lift

• ROI shows bottom-line return

Retail media benchmarks are useful, but only after normalization

Retail media benchmarks can help show where a network is over- or under-performing. But they do not prove profit on their own.

The reason is that major retail media networks do not score campaigns the same way.

Amazon Ads, Walmart Connect, Target Roundel, Kroger, and Instacart differ on:

• attribution model

• click window length

• halo treatment

• online vs. in-store sales credit

• default reporting logic

So when one network posts a stronger ROAS than another, the result may reflect measurement rules more than business impact.

That is why benchmark tables should be read as reporting context, not as a final ranking.

A few examples make that clear:

• Amazon Ads often uses last-touch click attribution with a 14-day window.

• Walmart Connect includes online, in-store, and halo sales in total sales reporting.

• Instacart Ads uses equal-credit multi-touch attribution by default.

• Target Roundel and Kroger use different closed-loop and loyalty-based methods.

Even small changes in method can move the answer a lot. Some findings show iROAS can vary by 6.5x based on measurement method alone, and in many campaigns the result can flip from positive to negative when the method changes.

That is why a benchmark is a signal, not a verdict.

Where brands get retail media ROI wrong most often

Three mistakes do most of the damage.

Treating platform ROAS as profit

This is the biggest one.

Platform ROAS is built for attribution, not for profit accounting. On branded search terms, that can mean a brand pays to sit in front of demand it already owned. The platform gets credit, but net gain may be thin.

That is how a campaign can look efficient while adding little incremental margin.

Mixing media effects with trade and pricing

If a sponsored campaign runs at the same time as:

• a coupon

• a temporary price cut

• a feature

• a display push

then the sales lift is shared.

If media gets full credit while trade did much of the work, return gets overstated. This double-counting makes retail media look stronger than it is and muddies later budget choices.

Ignoring margin, fees, and returns

Sales volume is easy to report. Profit is harder.

Retail media often gets judged before brands subtract:

• retailer fees

• shipping and fulfillment

• product margin

• return risk

• data costs

• campaign production costs

That missing layer is why a campaign can post good revenue and still disappoint the P&L.

Incrementality is the test that matters most

At the center of retail media measurement is one question:

Would the sale have happened without the ad?

If the answer is no, the sale is incremental. If the answer is yes, the platform may just be taking credit for demand already in motion.

That is why incrementality matters more than attribution when brands need a finance-grade answer.

A clean incrementality setup separates:

• direct conversions

• halo sales

• repeat purchases

• organic baseline

That matters because not every sale should be valued the same way. Some brands count only new-to-brand or new-to-category demand as incremental. Others include repeat purchases, but only after a control group shows lift.

Without that split, a single ROAS number hides too much.

The best retail media incrementality methods are test-based

For many brands, geo-holdout testing is still the best method for retail media lift measurement.

The idea is simple:

• ads run in one group of markets

• matched control markets stay dark

• sales differences are measured over time

Most tests need about four to eight weeks to reach statistical meaning, depending on volume and baseline noise.

Audience holdouts can also work, especially for sponsored products and retargeting, but they depend on platform support and clean randomization. When that is missing, contamination can creep in.

If holdout testing is not possible, some teams use Bayesian Structural Time Series models to estimate what would have happened without the campaign. That can help, but model-based estimates can vary a lot versus observed sales, so finance teams tend to treat them with more caution.

The main point is straightforward:

The tougher the test design, the harder it is for weak incrementality claims to survive.

Halo sales and trade overlap should not be buried in one number

Two items deserve their own lines in any retail media ROI model.

Halo sales

Halo sales are purchases of non-advertised SKUs that increase because a campaign lifts interest in the brand or portfolio.

Halo can matter. But if halo gets blended into direct campaign return, the advertised product may look stronger than it was on its own.

A cleaner setup is to show:

• direct sales lift

• halo sales lift

• total lift after the split

That keeps the reporting more honest.

Trade overlap

Trade overlap happens when promotions and media run together. If a coupon or price drop drove some of the lift, media should not take all the credit.

The fix is simple in concept: remove the trade-funded portion of lift before calculating media ROI.

Without that adjustment, retail media can look like the hero when the result came from a shared effort.

A finance-ready retail media scorecard should show both reported and modeled return

A strong scorecard does not hide the gap between platform reporting and business return. It

What Are the Most Important Things to Know About Retail Media ROI?

Retail media ROI can look strong on a dashboard and still fall apart under finance review. Five points matter most: ROAS can overstate return, incrementality is the only test that holds up, margin decides profit, retailer benchmarks do not line up cleanly, and halo should sit in its own line item. That is the split between a surface metric and a finance-ready number.

• ROAS is not ROI. One documented case found a campaign reporting a 14:1 ROAS while holistic Marketing Mix Modeling showed the real return was only 0.4:1.

• Incrementality is the real test. The core question is causation, not attribution, because spend that does not drive lift may simply tax shoppers who already planned to buy.

• Margin changes the answer. Margin, fees, fulfillment, and returns decide whether retail media produces profit or just sales volume.

• Retailer benchmarks are not directly comparable. Different attribution windows, fee structures, and fulfillment costs can distort side-by-side retailer comparisons.

• A useful ROI calculator splits direct sales, halo sales, trade overlap, and margin. The next sections show how to calculate that number and test it with discipline.

ROAS is not ROI. A high ROAS number can look impressive in a media platform, but finance teams do not book ROAS as profit. They look at what the spend added after all costs were counted. That is why a campaign can post a 14:1 ROAS and still deliver only 0.4:1 when Marketing Mix Modeling measures total business impact. Same spend. Same campaign. Very different answer.

Incrementality is the real question. Attribution reports show where a sale was recorded. Incrementality asks whether the sale happened because of the media. That distinction matters. If a shopper was already on the way to buy, media may be claiming credit rather than creating lift. In that case, retail media acts less like growth and more like a toll booth.

Margin is the denominator that matters to finance. Revenue alone does not settle the case. Gross margin, retailer fees, fulfillment costs, and returns all shape the final number. A campaign can move product and still miss the profit target once those costs hit the sheet.

Retailer benchmarks are not directly comparable. One retailer may use a different attribution window. Another may charge a different fee stack. A third may carry a different fulfillment mix. Put those side by side without adjustment, and the comparison starts to wobble. Benchmarks only help after normalization.

A useful ROI calculator separates direct sales, halo sales, trade overlap, and margin. That split keeps the math honest. Direct sales show the clearest near-term effect. Halo sales may matter, but they should not be blended in without disclosure. Trade overlap can muddy credit. Margin then converts sales impact into a number finance can use.

What Does Retail Media ROI Actually Measure - and Why Does It Confuse Finance Teams?

Retail media ROI tracks the profit a brand gains from retailer media after all costs are counted - not just ad spend. That sounds simple, but it trips up finance teams all the time. Marketing may look at platform sales and call the campaign a win. Finance may subtract cost of goods sold, retailer fees, fulfillment, and returns and land on a very different number. Same campaign, two stories. The gap usually comes down to inputs, timing, and one big issue: platform metrics often show efficiency, while finance needs bottom-line impact.

• Retail media ROI measures incremental profit, not just attributed sales.

• Marketing and finance often use different cost inputs, which leads to conflicting ROI numbers.

• ROAS, true ROI, and incrementality answer different questions and should not be treated as the same metric.

• Retail media networks use different data, placements, and attribution rules, so benchmarks do not transfer cleanly across platforms.

Retail Media, Commerce Media, and Retail Media Networks Defined

Retail media is advertising that runs on a retailer’s owned properties or uses that retailer’s first-party shopper data. Commerce media is the bigger category, with retail media as one part of it. A retail media network is the retailer-owned platform that sells that media access.

That distinction matters more than it may seem. Teams often toss these terms around like they mean the same thing. They do not. And when the terms get blurred, measurement gets messy fast.

Major U.S. retail media networks differ in how much shopper data they share, what placements they sell, and how they attribute sales. Amazon Ads leads spend, while Walmart Connect, Target Roundel, and Instacart each apply their own shopper data rules and attribution methods.

Retail media touchpoints usually sit in three buckets:

• On-site placements such as sponsored search, display, and video on the retailer’s own site or app

• Off-site placements such as audience extension across the open web, social platforms, and connected TV using retailer data

• In-store placements such as digital screens, audio, and physical displays tied back to loyalty or app data

A single campaign can run across all three. That is where things start to get slippery. A dashboard may show neat totals, but the path from impression to purchase can cut across channels, devices, and even store visits. clean measurement is tougher than many platform views make it look, especially as brands adapt to AI SEO and generative search environments.

Those definitions matter because retail media benchmarks, cost structures, and ROI inputs can shift from one network to another. A result from Amazon Ads does not line up neatly with a result from Walmart Connect or Instacart if the data rules are not the same.

ROAS vs. True ROI vs. Incrementality - What's the Real Difference?

This is the part that tends to cause the most friction.

ROAS (Return on Ad Spend) is a channel efficiency metric. It takes attributed revenue and divides it by ad spend. Retailers usually report it in near real time, which makes it useful for fast campaign moves.

True ROI is a profit metric tied to the P&L. It looks at incremental profit after every cost is included. That usually means a modeled view with a six-week lag or more.

Incrementality sits below both metrics. It asks the causal question: did the ad drive the sale, or was that shopper likely to buy anyway?

Here is the cleanest way to separate them:

Each metric has a job. The trouble starts when teams ask one metric to do another metric’s work.

ROAS is good for tactical optimization. It helps teams decide where to bid harder, where to pull back, and which placements look efficient in the platform. But ROAS can mislead when it becomes the only number used to defend media spend. Why? Because attributed revenue is not the same as profit, and attributed sales are not always incremental sales.

That is why any retail media ROI calculator needs to break out a few parts instead of rolling them into one neat output. At a bare minimum, it has to separate:

• Spend

• Lift

• Trade overlap

• Margin

Without those inputs, the number may look clean, but it will not answer the question finance is asking.

A simple way to think about it: ROAS tells a team whether the machine is moving. True ROI tells them whether the machine is making money. Incrementality tells them whether the machine caused the result in the first place.

That difference is exactly why retail media ROI keeps confusing finance teams. Marketing often works from platform attribution because it is fast and easy to act on. Finance works from profit logic because that is what hits the business. Neither side is wrong. They are just using different scoreboards.

With the terms aligned, the next step is turning them into a calculator-ready formula.

What Is the Retail Media ROI Formula - and How Do You Calculate True Profit, Not Just ROAS?

The retail media ROI formula is: (Incremental Profit − Ad Spend) ÷ Ad Spend, where Incremental Profit = Incremental Sales × Net Contribution Margin. That’s the number that shows whether a campaign made money, not just whether a platform reported sales. A sound calculator uses the inputs that change profit and leaves out vanity revenue totals.

The Core Retail Media ROI Formula

The formula has two parts.

Part 1 - Incremental Revenue: Baseline sales × incremental lift % = incremental revenue

Part 2 - True ROI: ((Incremental Revenue × Net Contribution Margin) − Ad Spend) ÷ Ad Spend

That second line is where many teams get tripped up. Platform dashboards often stop at attributed revenue. Finance teams cannot. To get to a profit-based view, net contribution margin has to reflect the full set of variable costs tied to the sale.

Net contribution margin must subtract product cost, retailer fees, fulfillment, discounts, returns, media, creative, data, and variable operations.

In plain terms, revenue is not profit. A campaign can post a strong ROAS and still lose money once those costs hit the P&L.

How to Account for Halo Effect and Trade Spend Overlap

Two adjustments turn a rough estimate into a model that finance teams can trust.

Halo effect tracks sales lift on non-advertised SKUs that move because the campaign increased brand visibility or shopper intent. That lift matters. But it should sit in its own line item, rather than being mixed into the advertised SKU’s return. When teams blend halo into core SKU performance, the result looks better than the base campaign actually was.

Trade spend overlap is another place where numbers can drift. If a coupon, discount, or promo runs during the same period as the retail media campaign, part of the sales lift came from trade support rather than the ad itself. Counting all of that lift as media-driven will overstate ROI. The fix is simple: remove the lift funded by trade before applying the incremental lift percentage.

Once halo and trade overlap are split out, retailer benchmark comparisons become far more useful. Apples-to-apples math matters here.

What the ROI Calculator Should Output

A useful calculator should not stop at one headline number. It should separate platform attribution from modeled profit, halo impact, and final net dollars so each team can see the part that matters to them.

That structure keeps reporting honest. Media teams still get the platform view they need for bidding and pacing. Brand and finance leaders get the number that shows whether the campaign added net dollars.

The calculator takes four inputs - ad spend, incremental sales lift, trade spend overlap, and margin - and returns all four outputs above. That normalized ROI number becomes the right baseline for retailer-by-retailer benchmarks.

What Do Retail Media Network Benchmarks Actually Tell You - and What Do They Leave Out?

Retail media network benchmarks can point a team in the right direction, but they do not prove profit. Each retail media network uses its own attribution rules, sales definitions, and reporting windows. That alone can change reported ROAS without any change in actual business results. Benchmarks only become useful when the math is normalized for margin, trade overlap, and halo.

• Retail media network benchmarks are directional signals, not final proof of return.

• Reported ROAS can change based on attribution windows and measurement rules, even when campaign performance stays the same.

• Walmart Connect, Amazon Ads, Target Roundel, Kroger, and Instacart all use different reporting logic, so direct comparisons can mislead.

• Benchmark tables help show how networks count sales, but they should not be used as a stand-in for incrementality or profit.

How to Use Retail Media Network Benchmarks Correctly

ROAS, CPC, and incremental lift do not measure the same outcome, and there is no shared standard across networks. A 14-day click window and a 30-day click window can produce different ROAS for the exact same campaign. That is a reporting shift, not a business shift.

Methodology can move iROAS by 6.5x and flip results from positive to negative in 83% of campaigns. That number should stop any team from treating platform-reported ROAS as a budget rule. A benchmark is a reference point. It can help flag a campaign that is well below peer norms or one that looks far above expectations. What it cannot do is settle the profit question on its own.

That is the heart of the issue. Benchmark comparisons only make sense after attribution logic is lined up across networks. Without that step, the comparison is apples to oranges. One dashboard may count more post-click sales. Another may give credit to halo. A third may use a longer window and pull in more conversions. The output looks precise, but the inputs are not the same.

Walmart Connect ROAS Benchmarks and What They Miss

Walmart Connect reports last-touch, post-click results across 3-, 14-, or 30-day windows. Its "Total Sales" metric includes online and in-store attributed purchases, plus an "Other Sales" halo.

That setup gives advertisers a broad picture of attributed revenue, but it does not isolate margin or trade overlap. Because of that, Walmart's reported ROAS cannot stand in for true ROI. The number may show that media activity aligned with sales, but it does not show whether those sales were profitable after retail costs, promo pressure, and overlap with trade programs.

That is why Walmart Connect benchmarks work better as incrementality signals than as final proof of return. If ROAS looks weak, the answer is not always poor performance. If ROAS looks strong, the answer is not always profit. The bigger risk is simpler: mistaking reported sales for bottom-line gain.

Amazon, Target Roundel, Kroger, and Instacart Benchmark Ranges

Across 182 brands, the median profit ROI for retail media overall was 1.44, below non-retail media at 1.60. That top-line number hides major retailer-by-retailer swings.

Kroger posts a median profit ROI of 2.31. Target Roundel comes in at 1.93. Amazon sits at about 1.2x for small brands, even though it takes the biggest share of spend in many retail media budgets. For small brands, that mismatch is hard to ignore. They put 71% of retail media spend into Amazon, despite Amazon being their lowest-returning network.

That gap between spend share and return is where benchmark reading gets useful. It can show where budget weight does not match payoff. But even here, the numbers need context. A network with lower reported ROAS may still drive stronger incrementality. A network with high profit ROI in one sample may not do the same for every category, margin profile, or shopper base.

Use the table to compare reporting logic, not to rank networks at face value.

One structural difference stands out. Instacart is the only major network in this group that uses equal-credit multi-touch attribution by default instead of a last-touch model. That means its reported ROAS is built on a different accounting system from the start. It is not higher or lower by definition. It is simply not measuring credit in the same way. Instacart also breaks out "Direct ROAS" and "Halo ROAS" into separate lines, which adds another layer when comparing it with other platforms.

Benchmark tables show reporting logic, not profit truth

A benchmark table can look clean and settled. In practice, it is more like reading five scoreboards from five sports with five rulebooks. The numbers are useful, but only if the scoring system is clear.

Amazon Ads leans on last-touch click attribution with halo at the brand level. Walmart Connect includes online, in-store, and halo sales. Target Roundel uses a closed-loop setup tied to guest ID. Kroger relies on loyalty and match-based reporting. Instacart splits direct and halo effects and uses equal-credit multi-touch. Each system answers a slightly different question.

That is why direct ranking can go off track. One network may look stronger because it counts more downstream sales. Another may look weaker because it reports on a tighter basis. When teams compare those outcomes without adjusting for margin, overlap, and incrementality, budget shifts can follow the wrong signal.

Accreditation limits make every retail media benchmark provisional

One critical caveat applies across the board. As of July 2026, MRC accreditation for retail media networks, including Amazon, Walmart, Instacart, and Criteo, covers delivery metrics such as impressions and clicks. No network has an accredited attributed-sales or ROAS metric.

That matters more than many teams admit. Impressions and clicks may be verified, but the sales credit sitting on top of those actions is still based on each network's own measurement rules. So even when a benchmark range looks solid, it is still directional until the number is normalized for attribution, margin, and incrementality.

This is where brands often get tripped up. They treat benchmarked ROAS as proof of ROI. It is not. It is a reported outcome shaped by platform logic. That distinction changes budget planning, channel ranking, and how success should be judged.

FAQ Section

What is a good retail media ROAS benchmark?

A good retail media ROAS benchmark depends on how the network defines attribution, sales, and halo. Reported ROAS should be read in context, because a stronger number on one platform may come from a longer window or a broader sales definition rather than better business impact.

Why do Amazon and Walmart retail media benchmarks look so different?

Amazon Ads and Walmart Connect use different attribution and sales reporting rules. Amazon uses last-touch click attribution with promoted ASIN and brand halo, while Walmart Connect includes online, in-store, and halo sales in "Total Sales." Those structural differences can change the headline ROAS.

Is retail media ROAS the same as profit ROI?

No. ROAS measures attributed revenue against ad spend. Profit ROI looks at return after margin and other business factors. A campaign can post a high ROAS and still underdeliver on profit if margin is thin or trade overlap is heavy.

How should brands compare Instacart Ads benchmarks with other networks?

Instacart should be compared with caution because it uses equal-credit multi-touch attribution by default, unlike the last-touch models used by many other major networks. Its "Direct ROAS" and "Halo ROAS" split also changes how results appear next to other platforms.

TL;DR Summary

• Retail media network benchmarks are directional signals, not final proof of return. They help spot patterns, but they do not answer the profit question by themselves.

• Reported ROAS can change based on attribution windows and measurement rules, even when campaign performance stays the same. A shift from a 14-day to a 30-day window can change the number without changing the business outcome.

• Walmart Connect, Amazon Ads, Target Roundel, Kroger, and Instacart all use different reporting logic, so direct comparisons can mislead. Each network counts sales and assigns credit in its own way.

• Benchmark tables help show how networks count sales, but they should not be used as a stand-in for incrementality or profit. The table is a guide to methodology, not a final ranking of value.

CTA Block

Brands that need cleaner retail media decisions should start with a Bigeye retail media audit. Our team reviews attribution rules, margin reality, halo treatment, and trade overlap across Amazon Ads, Walmart Connect, Instacart, Target Roundel, and Kroger, then maps reported ROAS to actual business return. **Schedule

Where Do Brands Most Often Get Retail Media ROI Wrong?

Retail media ROI often looks stronger than it is. The biggest reporting errors can make a campaign seem profitable even when margin tells a very different story. Once CFOs start digging into retail media line items, the same weak spots tend to show up first. The issue usually is not the platform dashboard itself. The issue starts when brands read an attribution report like a profit-and-loss statement. Three mistakes tend to do most of the damage.

Treating Platform ROAS as Profit

Platform ROAS is an attribution metric, not a profit metric. That gap shows up fast on branded search terms, where demand already exists before the ad runs. If a brand already owns the top organic position, the ad may not be creating a new sale at all. It may simply be stepping in front of a sale that would have happened anyway.

That distinction matters. A reported return can look strong on paper while net gain stays flat. In plain terms, the brand is paying to collect demand it already had. That is why branded campaigns, especially in retailer search, need a tougher read than platform ROAS alone can give.

Mixing Media Effects with Pricing, Trade, and Promotions

Pricing and promotions can drive the same kind of sales lift that media appears to drive, but standard ROAS often gives the ad full credit. When a Sponsored Product campaign runs at the same time as a temporary price drop or display feature, the sales bump can land in the media column even if trade support did most of the work.

That creates double-counting. It also makes budgeting harder, because teams can no longer tell what the media dollars produced on their own. A blended result may look strong in a dashboard, but that framing tends to fall apart under finance review. Without a clean way to separate trade activity from media activity, the number being reported is not a pure media result.

Ignoring Margin, Fees, and Creative Costs

Revenue is not profit. That sounds obvious, yet retail media reports often stop at sales and ROAS. The costs most often left out include COGS, retailer fees, fulfillment, returns, and creative production. None of those items usually appears in a standard retail media dashboard, and every one of them cuts into net return.

The cleanest check is the contribution margin test: subtract those costs from incremental revenue to see whether the campaign is adding value or slowly wearing it down. Without that step, teams end up optimizing toward a number the CFO cannot use. That is exactly why lift testing comes next.

How Do You Actually Measure Incrementality in Retail Media?

Incrementality in retail media comes down to one blunt question: would the sale have happened without the ad? If that answer is unclear, reported ROAS can look stronger than the business impact it reflects. The first job, then, is to separate incremental sales from credited sales and avoid treating platform attribution as proof of lift.

• Direct conversions are only one part of the picture, and they can include demand that was already there.

• Halo sales, repeat purchases, and organic lift each affect ROI in different ways and should be tracked on their own.

• Geo-holdout tests remain the top method for proving retail media lift, while other methods need more caution.

• A defensible ROI number starts with incremental revenue, then backs out ad spend, COGS, fees, fulfillment, and returns.

What Counts as Incremental Sales and Halo Outcomes

To measure incrementality cleanly, brands need to split sales impact into separate buckets rather than lump everything into one platform-reported total.

Direct conversions are the easiest number to find because dashboards show them by default. The problem is that those conversions can include organic demand that simply picked up an attribution tag along the way. A shopper may already plan to buy, click a sponsored placement, and then get counted as ad-driven.

Halo sales add another layer. These are purchases of non-advertised SKUs in the same brand portfolio that were triggered by an ad for a different SKU. That matters in retail media because a campaign may push one hero product while the real revenue bump appears across adjacent items.

Repeat purchases should also stand apart from new customer demand. For many brands, repeat orders from current customers do not count as incremental at all. Some define incrementality only as new-to-brand or new-to-category sales. That one definition change can alter the ROI story in a major way.

Organic lift has to be netted out against a control group before any lift claim holds up. This is especially important in branded search, where an ad may do little more than intercept demand already in motion. If someone was headed to the product anyway, the ad did not create the sale. It just got credit for it.

Once these sales components are split out, the next step is proving which part of performance came from media exposure and which part would have happened on its own.

The Best Testing Methods for Retail Media Incrementality

Geo-holdout tests are the most credible method available for retail media incrementality. The setup is simple on paper: ads run in one set of geographic markets while matched control markets stay dark, and the brand measures the sales gap between the two groups. To reach statistical significance, these tests usually need to run for four to eight weeks.

That matters because geo-holdouts rely on observed sales difference, not platform credit. In plain terms, they ask the market to answer the question instead of the dashboard.

Audience holdouts take a different route. Rather than splitting by market, they withhold ads from a randomized subset of users within the same market. This can work well for Sponsored Products, DSP campaigns, and retargeting, but only when the platform can support a clean test design. Without that support, contamination creeps in fast and the result gets shaky.

When neither holdout path is possible, Bayesian Structural Time Series (BSTS) modeling can estimate what likely would have happened without the campaign by using pre-campaign sales patterns. It is a fair fallback, but it is still a fallback. Its outputs can vary by as much as 90% compared with observed sales, which makes it a weaker base for finance review and budget calls.

For most brands, the practical split is simple: use geo-holdouts for campaign proof and MMM for annual planning. That line matters because measurement method is not a minor detail. It can swing the answer in a dramatic way.

Methodology choices alone can cause iROAS outputs to vary by an average of 6.5x, and in a study of 42 campaigns, changing the measurement method caused iROAS to flip from positive to negative in 83% of cases. That is not noise. That is a warning sign.

Because of that, brands should ask retail media partners direct questions about test design. Do they use propensity score matching or simple clustering? Did they include historical brand sales as a matching feature? Those choices affect the result before the campaign is even judged. Propensity score matching delivers 12x better match quality than clustering, but it also tends to produce more conservative iROAS numbers. In other words, the stricter method may look less flattering, yet still be closer to the truth.

How to Turn Lift Results into a Defensible ROI Number

Once lift is proven, attributed revenue should stop being the main ROI input. The better move is to swap in incremental revenue and then subtract the costs that finance teams care about: ad spend, COGS, retailer fees, fulfillment, and returns. What is left is incremental contribution margin, which is the number a CFO can actually judge.

That shift changes the conversation. Instead of asking whether media got credit for sales, the brand asks whether media produced margin after all the friction of retail is accounted for.

A useful example comes from Albertsons Media Collective. In 2026, it ran a beta campaign for Mondelez using causal incrementality measurement for in-store digital media. Instead of leaning on dashboard attribution, the test used a control group and found a $2.41 iROAS along with a 14% sales lift. That kind of output lands better in budget meetings because it rests on measured lift, not credited conversion volume.

Two checks help confirm whether those lift results are believable.

TACoS (Total Advertising Cost of Sale) compares ad spend against total brand revenue, not just attributed revenue. If TACoS keeps climbing while ROAS stays flat, the ads are likely collecting demand that already existed instead of creating new demand. That is one of the clearest signs that media may be over-credited.

The second check is the digital shelf itself. Organic rank, buy-box ownership, and out-of-stock rates can shift the baseline underneath the test. If stock runs out in one market, or the product loses buy-box control, the measurement result can get distorted. A campaign is never running in a vacuum. It is running on a shelf with moving parts.

Retailer-to-Retailer iROAS Comparisons Need the Same Measurement Model

Comparing iROAS across retail media networks sounds simple. In practice, it is easy to get fooled.

Before comparing Walmart Connect, Amazon, Target Roundel, Kroger, and Instacart, the methodology behind each iROAS number has to match. Different networks may use different matching methods, different control logic, and different revenue models. If those inputs are not aligned, the comparison is not apples to apples. It is more like apples to shopping carts.

Results should be normalized into one financial model before any budget move is made. That model needs to include SKU-level margin, return risk, and retailer-specific fees. Without that step, one network may look stronger simply because of how it was measured or how revenue was counted.

That is why incrementality is the only fair basis for retailer comparison. It strips away some of the attribution noise and gives decision-makers one cleaner standard: which channel produced sales that would not have happened otherwise?

FAQ Section

What is the difference between retail media attribution and incrementality?

Attribution shows which ad touchpoint received credit for a sale. Incrementality asks whether the sale would have happened without the ad. A platform can claim attribution even when the shopper already planned to buy, which is why the two numbers often differ.

How long should a retail media incrementality test run?

Geo-holdout tests usually need four to eight weeks to reach statistical significance. The exact timing depends on sales volume, market size, and how much noise exists in the baseline.

Are repeat purchases considered incremental in retail media?

Not always. Many brands do not count repeat purchases from current customers as incremental and instead focus on new-to-brand or new-to-category sales. The definition should be set before the test starts, not after results come in.

When should a brand use BSTS instead of a holdout test?

BSTS is best used when geo-holdouts or audience holdouts are not possible. It can estimate the counterfactual from pre-campaign sales history, but its outputs can differ sharply from observed sales, so it is less convincing in finance settings.

TL;DR Summary

• Direct conversions are only one part of the picture, and they can include demand that was already there. Platform dashboards can over-credit media when shoppers were already on the path to purchase.

• Halo sales, repeat purchases, and organic lift each affect ROI in different ways and should be tracked on their own. A mixed bucket hides what the campaign actually changed and what it merely touched.

• Geo-holdout tests remain the top method for proving retail media lift, while other methods need more caution. Audience holdouts can work well, and BSTS can help when testing is not possible, but neither should be treated as equal to a strong holdout design.

• A defensible ROI number starts with incremental revenue, then backs out ad spend, COGS, fees, fulfillment, and returns. That produces incremental contribution margin, which is far more useful than attributed revenue when budgets are on the line.

CTA Block

Need a cleaner way to judge retail media spend? **Bigeye can help audit retail media measurement

What Does a Retail Media Measurement Framework Actually Look Like in Practice?

Once incrementality is proven, the next move is building a finance-ready scorecard that shows what retail media spend actually delivered. That means one brand-owned model across marketing, ecommerce, and finance, not a patchwork of retailer dashboards treated as the final word. Finance-approved margin and fee tables should guide the math, not platform defaults.

That shift matters because the gap between platform-attributed ROAS and true incremental ROAS can land between 40% and 60%.

What Does the Scorecard Every CMO Should Ask For Include?

A strong scorecard answers one core question: did spend create incremental profit? To do that, it needs to map straight to the main ROI inputs: spend, lift, trade overlap, margin, and halo.

Each row should lead to an action, not just a readout. If a lift number appears, it should come with match rate, time window, and confidence range every time. Without that context, a nice-looking result can turn into a bad budget call.

Reporting also needs to split online and in-store outcomes before they are rolled into one brand-level view. Otherwise, in-store ad exposures can get treated like guaranteed sales, which distorts what the media actually did.

A scorecard only works when the data under it is centralized and reconciled. If the inputs do not match, the output will not hold up in a finance review.

How Does Bigeye Help Consumer Brands Connect Media, Analytics, and Commerce?

This is where the framework stops being a slide and starts becoming an operating system. Many CPG digital marketing teams still keep retailer feeds, POS data, and modeled results in separate tools. That setup makes it hard to compare platform ROAS with modeled ROI, and even harder to show finance where the gap sits.

A better model brings those inputs into the brand’s own connected reporting layer, then reconciles in-platform ROAS against modeled ROI so the difference is plain to see. Privacy-safe data joins can then connect media exposure with retail transactions in one model, giving marketing, commerce, and finance a shared view of performance.

What Is the Right Way to Summarize Retail Media ROI Without Misleading Finance?

The right retail media ROI summary does not start with ROAS. It starts with one hard question: did the spend create incremental profit? Platform ROAS can show attributed sales, but incremental ROI shows the profit the media actually caused. That distinction is the baseline for reading retailer benchmarks in a way finance teams can trust.

Finance-ready ROI means incremental profit after ad spend, COGS, fees, fulfillment, returns, and trade overlap. TACoS works as a gut check. If TACoS climbs while ROAS stays flat, the ads are more likely pulling in demand that already existed instead of creating new demand.

Retailer benchmarks only mean much after normalization for attribution windows, fee structures, and return-adjusted sales. Without that cleanup, comparisons can look neat on paper and still point the team in the wrong direction.

A practical scorecard puts platform ROAS and modeled ROI next to each other in one reconciled view. That gives finance a clear line of sight into where the gap sits, what is driving it, and why the spend belongs in the plan. From there, those same inputs can be compared against retailer-specific benchmark ranges.

Ready to See What Your Retail Media Spend Is Actually Returning?

For a finance-ready answer, use the calculator above. Enter spend, incremental lift, trade overlap, and margin. The output separates true contribution ROI from platform-reported ROAS.

When those numbers don’t match, the gap usually points to attribution inflation, trade overlap, or margin erosion. In plain terms, reported performance may look better than the business result that shows up in finance.

That’s when a measurement review makes sense. Bigeye’s marketing analytics and performance intelligence practice aligns platform ROAS with modeled ROI, turning a one-time calculator result into a scorecard finance teams can use. Start with the calculator, then use the linked measurement services to pinpoint and close the gap.