.svg)
.png)
Meta says a campaign drove $180K in revenue. Google says $220K. The retail media placement you added last quarter says $60K. Add it up and you're reporting $460K in "attributed" revenue on a client whose actual sales for the month were $400K. Nobody lied. Every platform just did exactly what it's built to do: claim as much credit as its own attribution model allows.
This is the reconciliation problem, and it's getting worse, not better, as agencies add more channels to the mix. Here's what's actually causing it, and a framework for getting to one number your clients can trust.
Each ad platform tracks conversions using its own attribution window and its own rules for what counts as "influenced by this ad." When a customer sees a Meta ad, searches the brand on Google two days later, and buys, both platforms often claim that sale. That's not a bug in either system — it's how walled-garden attribution is designed to work, and The Drum's reporting on this shift puts it plainly: every platform claims credit for the same sale.
The gap is also widening because of specific attribution changes, not just channel proliferation. Meta lowered the threshold for what it counted as an "engaged view" on video ads from 10 seconds down to 5 seconds in early 2026, reasoning — per its own announcement — that nearly half of purchase conversions tied to Reels now happen within the first two seconds of attention. Meta later folded this into a broader "engage-through attribution" model that also counts non-click social engagements like likes, shares, and saves. Meta says billing didn't change, only what gets counted as a conversion. Third-party analysis of the shift, from ad-reporting vendor 1ClickReport, estimates it inflates reported ROAS by 15–25% on many accounts — that's the vendor's own estimate, not a figure Meta has published, so treat it as directional rather than exact. Short-form video under 15 seconds is affected most. Separately, Meta deprecated its 7-day and 28-day view-through attribution windows in the Ads Insights API on January 12, 2026 — confirmed in Supermetrics' own technical documentation, which links back to Meta's original announcement of the change. If a client's reporting tool still requests that deprecated data, it can silently return blank rather than an error, understating real numbers without any alert.
One agency founder described running into exactly this kind of mismatch: a client split spend 50/50 between Google and Meta at roughly $10K per month, and the platform numbers looked balanced on the surface. Suspecting Meta was getting credit it hadn't earned, the agency gradually shifted budget toward Google while watching for any drop in brand search or direct revenue. None appeared, and the account settled at a 70/30 split with better blended results on the same total spend. That's one agency's account of one client, not a universal rule — the same imbalance can run the other way, with a platform quietly carrying top-of-funnel awareness that a last-click model never credits. The only way to know which situation you're in is to check.
None of this means your platforms are broken. It means platform-reported numbers were never designed to add up to your client's actual revenue, and treating them as if they should is where the trust problem starts.
Attribution and incrementality answer two different questions, and conflating them is a common source of client confusion.
Attribution assigns credit across touchpoints a customer interacted with before converting. It tells you where a conversion showed up. Incrementality asks a harder question: would that sale have happened anyway, without the ad? As Rokt's Marc Allsop put it to The Drum, attribution tells you what happened; incrementality tells you what caused it.
Neither one replaces the other. Marketing mix modeling, multi-touch attribution, and incrementality testing each answer part of the picture, and marketers are split on which to trust most — in one industry survey, 27.6% rated MMM the most reliable method, 19.4% picked multi-touch attribution, and 18.9% pointed to unified measurement approaches, per eMarketer's incrementality FAQ. No single method captures the full picture on its own, which is why the reconciliation step below matters as much as any individual measurement method.
The fix isn't picking the "right" platform to trust. It's building one number that no single platform can inflate: blended ROAS, sometimes called marketing efficiency ratio (MER).
The formula is simple: total revenue divided by total ad spend, both pulled from one reconciled source rather than from each platform's own reporting. Revenue comes from the client's payment processor, ecommerce backend, or CRM — not from any ad platform's claimed conversions. Spend comes from summing each platform's actual billing, not the in-platform reporting view. Karbon Analytics' worked example shows the mechanic clearly: say a store brought in $150,000 in real revenue against $50,000 in total ad spend across Meta, Google, and TikTok — a blended ROAS of 3.0. Individually, those same three platforms might report a combined $172,000 in claimed revenue, a $22,000 overclaim on money the store never actually made. Blended ROAS counts each sale once, by design.
This isn't just a spreadsheet exercise — it's the same principle retail media networks are now applying at scale. Albertsons Media Collective and Mondelēz ran a matched-market test that isolated real incremental impact rather than relying on platform-claimed credit, and it produced a $2.41 matched-market incremental ROAS with a 14% lift in in-store sales across 116 locations, according to eMarketer's coverage of the test. That's the same reconciliation logic, applied with a real budget and a real result.
For a client conversation, this is the number to lead with — not because platform ROAS is wrong to look at (it's still useful for tactical, in-platform decisions like which ad set to scale), but because it's the only figure that reflects what the client's business actually made.
Blended ROAS tells you what happened. It doesn't tell you whether the spend caused it. That's where incrementality testing comes in — and it's become a lot more accessible to agency-sized budgets than it used to be.
More than half of US brand and agency marketers, 52%, already run incrementality tests, and 36.2% plan to increase that investment over the next year, according to the EMARKETER/TransUnion survey cited in eMarketer's FAQ. But adoption has outpaced maturity: the same research, drawing on Skai and the Path to Purchase Institute's State of Retail Media report, found 44% of marketers question the reliability of their incrementality results, 43% struggle applying it consistently across ad types and retailers, and 41% say their tools aren't sufficient to run tests well. Retail media is where the pressure is highest — 71% of advertisers now rank incrementality as their single most important retail media KPI.
The good news for smaller accounts: testing no longer requires an enterprise budget. In late 2025, Google lowered the minimum spend needed for its incrementality experiments from roughly $100,000 down to $5,000, using Bayesian statistical models that need far less data to produce a usable read — a change reported by Search Engine Land, PPC Land, and SEOteric. That puts controlled testing within reach of accounts that could never have justified it before.
Two things worth setting expectations on before you start. First, most practitioner guidance lands on a holdout group of roughly 5–10% of the audience — large enough to detect a real effect, small enough that withholding ads from it doesn't meaningfully dent performance during the test window. Second, incremental ROAS numbers typically come in lower than platform-reported ROAS, because incrementality sets a stricter bar for what counts as a real result. Kroger Precision Marketing makes this same point in its own measurement guidance, per eMarketer — a lower incremental number is expected, not a sign the campaign is underperforming.
The reconciliation math matters less to a client than what it means for their budget. Lead with the reconciled number and the business implication, and keep the methodology to a short, available-on-request explanation rather than the centerpiece of the conversation.
A few things tend to help. Show the platform-reported numbers next to the blended number side by side, so the gap is visible rather than asserted. Explain the "why" in one or two sentences — which attribution change or overlap is driving the difference — rather than walking through the full mechanics of each platform's model. And when a channel looks weaker under blended or incremental measurement than it did in-platform, check whether it's actually underperforming or whether it's carrying awareness that a last-click model further down the funnel is claiming credit for. Cutting a channel on platform-reported numbers alone is how agencies accidentally break the part of the funnel that was working.
Pulling the platform numbers, the CRM revenue, and the blended calculation together every reporting cycle is its own recurring task — one that gets heavier with every client and every channel added to the mix.
Rebuilding the reconciled number above by hand — logging into Meta, Google, and every retail media portal, then stitching it together — is exactly the kind of recurring work that gets harder to sustain as a client roster grows. Adriel connects a client's ad channels into a single dashboard your team and the client can both see, so that reconciled view is something you open, not something you rebuild every reporting cycle.
What's the difference between attribution and incrementality? Attribution assigns credit to touchpoints a customer interacted with before converting. Incrementality tests whether the sale would have happened without the ad at all. Attribution tells you where a conversion appeared; incrementality tells you what caused it.
Is a gap between platform-reported ROAS and blended ROAS normal? Yes. Because platforms use overlapping attribution windows and can both claim the same sale, summed platform revenue routinely exceeds a client's actual revenue. Blended ROAS, calculated from total revenue divided by total ad spend, removes the double-counting.
Do I need a dedicated measurement platform to do this? No. Blended ROAS only requires revenue data from the client's payment processor or CRM and spend data pulled from each platform's billing. A dedicated cross-channel reporting tool makes it faster and easier to keep current, but the calculation itself doesn't require one.
How much budget do I need to run an incrementality test? As of late 2025, Google's own incrementality testing can work with budgets as low as $5,000, down from roughly $100,000 previously, thanks to Bayesian statistical methods that need less data. Geo-based and holdout tests from other providers vary, but testing is no longer limited to enterprise accounts.
Why does incremental ROAS look lower than my platform-reported ROAS? Incrementality sets a stricter bar — it only counts sales that wouldn't have happened without the ad, while platform ROAS counts everything the platform can plausibly claim. A lower incremental number doesn't mean the campaign is failing; it means you're finally seeing the real number.
.png)
.png)
.png)