Reported ROAS is wrong in two opposite directions at once. Platforms over-attribute by claiming credit for sales they didn't cause, and under-count by missing sales they can't see. For brands selling through retail, the under-counting is far larger.

‍

Average ecommerce ROAS sits around 2.87x, down roughly 4% year over year [1] [2]. Platform-reported figures over-attribute versus incrementality-tested results by an estimated 20–50% depending on vertical [3].

‍

An estimated 15–30% of platform-reported conversions are non-incremental — the customer would have bought anyway [4].

‍

Effect one, weeks 1–4: offline and marketplace purchases enter your ad accounts. Reported ROAS rises. Media performance has not changed.

‍

Effect two, month 2 onward: bidding and audience models train on your full buyer set. Efficiency genuinely improves, and compounds.

‍

What doesn't change: better signal does not make reported ROAS a measure of incrementality. That still requires holdout testing.

‍

What is Reported ROAS, and How Does It Differ from Real Performance?

‍

Return on ad spend is revenue divided by advertising spend. Reported ROAS is that calculation performed using the ad platform's own attribution, for example, Meta's count of conversions it believes it drove, divided by what you paid Meta.

‍

Real performance is whether your advertising caused profitable sales. These are related but distinct, and the gap between them is where most bad budget decisions live.

‍

Three versions of the number get used interchangeably, and they aren't comparable:

‍

Platform-reported ROAS: each platform's own claim. Sums to more than reality, because platforms double-count.

‍

Blended ROAS: total revenue divided by total ad spend. Removes inter-channel attribution arguments; many performance teams now treat it as the north star [4].

‍

Incremental ROAS: revenue that wouldn't have happened without the ads. The only version that answers the causal question, and the hardest to obtain.

‍

The Two Reasons Why Reported ROAS is Often Wrong

‍

#1: Ad Platforms Over-Attribute DTC Conversions

‍

Example: Let's say a customer sees your Instagram ad Monday without clicking. On Weednesday they, search your brand and click a Google Shopping ad. On Thursday, the customer goes on to purchase from a Klaviyo email.

‍

In this situation, Meta claims the conversion on a view-through window, Google claims the click, and Klaviyo claims last-touch. Your dashboard will then show three conversions [1], despite only making one sale.

‍

Beyond double-counting, platforms also claim sales that would have happened regardless.

‍

Incrementality testing consistently finds that 15–30% of platform-reported conversions are non-incremental [4], and one 2026 benchmark index estimates platform-reported ROAS over-attributes by 20–50% versus incrementality-tested or MMM-tested figures, varying by vertical and campaign type [3].

‍

#2: Ad Platforms Can’t See Retail Conversions

‍

Simultaneously, ad platforms miss conversions entirely. Browser-side signal loss from iOS restrictions, ad blockers, and consent gating is estimated to cost 30–50% of conversion data for many advertisers [5].

‍

And for brands selling through retail and marketplaces, there's a much bigger absence: purchases at a Target shelf or inside an Amazon order produce no session, no click, and no pixel fire at all. They aren't lost in transit. They were never visible.

‍

Which Error Dominates?

‍

For a pure DTC brand, over-attribution usually dominates, and their reported ROAS flatters them.

‍

For an omnichannel brand, under-counting dominates overwhelmingly. If most of your volume moves through retail, no amount of double-counting on the DTC slice compensates for the majority of your revenue being absent from the numerator entirely.

‍

This is why "improve ROAS" means something different for you than it does for a DTC-only brand. They need to deflate an inflated number. You need to stop omitting most of your revenue from it.

‍

Benchmarks: What Reported ROAS Actually Looks Like in 2026

‍

These are directional figures, not targets. Keep in mind that these benchmarks vary in both sample size and method, but they are helpful for soft alignment.

‍

Average ecommerce ROAS: approximately 2.87x, down about 4% year over year [1] [2]

Median blended ROAS: around 3.4x in Q1 2026 [6]

Meta, ecommerce: roughly 1.86x–2.2x median for prospecting; 3.6x or higher on retargeting [2] [4] [7]

Google Ads, ecommerce: roughly 3.68x–4.2x, with Search and Shopping campaigns higher [2] [7]

TikTok: roughly 1.4x [8]

Repeat customers: typically deliver 3–4x higher ROAS than new customers across DTC verticals [9]

‍

The primary benchmark that matters is your break-even ROAS. A brand at 70% margins is profitable at 2x, and a brand at 30% margins needs 5x or more to break even [6] [10].

‍

How Sending Retail Buyer Data to Ad Platforms Impacts ROAS

‍

Closing your offline data gap produces two distinct outcomes, and collapsing them into one story is how measurement projects get judged wrongly.

‍

#1: Attribution Recovery

‍

Brands sending offline purchase data to ad platforms at high EMQ scores typically begin to see attribution recovery right away, within their first week.

‍

After verified retail and marketplace purchases start arriving in your ad accounts as conversion events, ad platforms (Meta, Google, and TikTok) match them to people who saw your ads and credit the campaigns accordingly.

‍

As a result, reported ROAS rises and reported CAC falls. Brands often discover that campaigns previously considered unprofitable have in fact been driving retail and marketplace sales.

‍

Note that this isn’t an immediate performance gain; it’s instead a very valuable impact recalculation.

‍

#2: Longer-Term Algorithmic Improvement

‍

Modern ad buying does not rely on manual targeting. Advantage+, Performance Max, and the retrieval systems beneath them decide who sees your ads by learning from the conversions you report. If these algorithms are fed only DTC purchase conversion event data, they learn and optimize on the traits of these DTC buyers. This subpopulation targeted skews toward people who already knew your brand.

‍

After feeding algorithms offline purchase data, three things shift:

‍

Bidding models optimize toward your real customer base, not the slice that happened to be observable.

Audience seeds improve. Lookalike and Customer Match audiences built from your full buyer file (including retail buyers you previously couldn't identify) start from a fundamentally better seed.

Budget allocation corrects, so campaigns driving retail volume stop looking like failures.

‍

This effect does take time, and it keeps improving as the data accumulates. It is smaller than effect one in month one and much larger by month twelve.

‍

In terms of magnitude of impact, Meta reports that advertisers using its Conversions API average 17.8% lower cost per result than pixel-only setups [11]. This is, of course, aggregated across all advertisers in all industry verticals.

‍

How to Report ROAS After Sending Offline Data to Ad Platforms

‍

1. Record a baseline before you change anything. Track platform-reported ROAS, blended ROAS, and MER for the trailing 90 days.

‍

2. Annotate the change date in every dashboard. Every chart anyone looks at for the next two quarters should show when offline events started flowing.

‍

3. Report offline-attributed conversions as a separate line. Keep this distinct, rather than blended into the total. This makes effect one visible as effect one, which protects you from being accused of a reporting sleight of hand later.

‍

4. Track blended ROAS and MER alongside platform figures. Blended is immune to the attribution recovery effect, because total revenue and total spend don't change when the counting improves. If blended ROAS also improves, that's effect #2 as mentioned above.

‍

5. Run a holdout test for the causal question. Geo holdouts or platform lift studies. Signal improves the data; it does not answer whether the ads caused the sales.

‍

Remember: Platform-reported ROAS rising tells you counting improved, blended ROAS rising tells you performance improved.

‍

About Brij Signal

‍

Every step above depends on one input: identified retail and marketplace buyers. No API, agency, or attribution model produces them. Retailers don't share customer files, and marketplaces keep the buyer relationship.

‍

Brij solves that through a mechanism called shelf to signal.

‍

A shopper buys your product in retail or on a marketplace, then scans a QR code on pack or at shelf, registers the product, claims a rebate, or uploads a receipt — identifying themselves directly to your brand in exchange for something they want. Brij verifies the purchase and forwards it automatically, as each purchase happens, as a deterministic conversion event to Meta, Google, and TikTok Ads through their Conversions APIs.

‍

Both effects follow from the same feed:

‍

Effect #1, because the events are deterministic. Match rates average 99.7% [i], since identifiers come from customers who registered rather than from probabilistic matching. Events that match get credited.

‍

Effect #2 because the events are complete and continuous. The platforms aren't learning from a weekly batch of your DTC slice; they're learning from your actual buyers as those purchases happen.

‍

And because the same verified event goes to your CRM and email/SMS platform in parallel, retention work begins on customers you previously couldn't reach — so the LTV side of the equation moves at the same time rather than trading against acquisition.

‍

Brij is SOC 2 Type 1 compliant [i], GDPR- and CCPA-aligned, with SHA-256 hashing throughout and customer data that remains brand-owned.

‍

For the mechanics of getting these events into each platform, see our guide to offline conversion tracking. For the parallel argument on the cost side, see why rising CAC is usually a measurement problem.

‍

Key Takeaways

‍

If you sell most of your product through retail, your reported ROAS has been understating your business, and it’s not just by a few percent, but by most of your revenue.

‍

Fixing that produces a number that jumps quickly. Resist the urge to celebrate it. The jump is your accounting catching up to reality, and it happened without a single media decision changing.

‍

The thing worth waiting for arrives second and quieter: ad platforms that have finally seen your actual customers, spending your budget accordingly. Watch blended ROAS, not the platform dashboards, to know when it lands.

‍

See how Brij turns shelf and marketplace purchases into deterministic signal for Meta, Google, and TikTok Ads — book a 15-minute demo here.

‍

FAQ

‍

What is the difference between reported ROAS and real performance?

Reported ROAS is revenue divided by spend using an ad platform's own attribution. Real performance is whether the advertising caused profitable sales. They diverge because platforms simultaneously over-attribute, claiming credit for sales that would have happened anyway, and under-count by missing conversions they cannot observe, including all retail and marketplace purchases.

‍

How can a brand improve ROAS when most sales happen in retail?

By sending verified retail and marketplace purchases back to the ad platforms as server-side conversion events. This produces two effects: reported ROAS rises within weeks as previously invisible sales get counted, and actual efficiency improves over subsequent months as bidding models optimize against the complete buyer set rather than the DTC minority.

‍

Does higher reported ROAS mean advertising performance improved?

Not necessarily, and usually not at first. When previously uncounted conversions begin being reported, ROAS rises without any change in media performance. That is the measurement correcting. Genuine performance improvement appears later. The clearest test is blended ROAS (total revenue over total spend) which is unaffected by attribution recovery and therefore only moves when performance actually changes.

‍

What is a good ROAS for ecommerce in 2026?

Average ecommerce ROAS is around 2.87x, with median blended ROAS near 3.4x. Meta prospecting typically runs 1.86x–2.2x and Google Ads 3.68x–4.2x. However, the only meaningful target is break-even ROAS: one divided by gross margin, plus a profit buffer. A 70%-margin brand profits at 2x; a 30%-margin brand needs 5x or more.

‍

How much of platform-reported ROAS is not incremental?

Incrementality testing commonly finds that 15–30% of platform-reported conversions would have occurred without the ad. One 2026 benchmark analysis estimates platform-reported ROAS over-attributes by 20–50% relative to incrementality-tested or MMM-tested figures, depending on vertical and campaign type.

‍

Does offline conversion tracking fix ROAS over-attribution?

No. Offline conversion tracking addresses under-counting — sales the platforms never saw. It does not stop platforms from double-counting or from claiming non-incremental conversions. Those remain questions for holdout testing and mix modeling.

‍

Is blended ROAS better than platform-reported ROAS?

For judging overall program efficiency, generally yes, because it eliminates inter-channel attribution disputes and cannot be inflated by counting improvements. Many performance teams now use blended ROAS as their primary metric and platform-reported ROAS only for relative optimization within a channel.

‍

How long does it take to see ROAS improvement from better conversion data?

Attribution recovery typically appears within one to four weeks as offline events begin being credited. Algorithmic improvement builds from roughly month two and compounds as event volume accumulates. The magnitude varies by channel mix, spend level, and how much sales volume was previously invisible.

‍