Every performance marketer has had some version of the same conversation with their founders or their CFO. Executive dashboards and presentations show strong ROAS, but the bottom line is showing something a little bit different. Everyone wants to know the real number to trust.
The answer isn’t an exact between picking one and another because in-platform ROAS isn’t wrong by the platform’s own logic. It’s measuring a different thing than the P&L. In Ecommerce, customers are exposed to ads on multiple platforms before they purchase, so the gap between what the platforms claim and what the business actually earned is bigger than most marketers want to acknowledge.
The reason isn’t that platforms like Meta and Google are hiding anything or acting maliciously. It’s that platform-reported ROAS is structurally incapable of telling you what happened in your business, and most teams are still building budgets, scaling decisions, and having executive conversations on it anyway. Change is needed.
What in-platform ROAS actually measures
To understand why the number on your dashboard is misleading, you have to understand what each platform is actually counting. Now, I know, some of this might be obvious but I need to frame it anyway.
Meta and Google both count three categories of conversions. Click-through conversions, where a user clicked an ad and purchased within an attribution window. View-through conversions, where a user saw an ad and purchased within an attribution window without clicking. And modeled conversions, where the platform fills in gaps in tracking with statistical estimates of what probably happened.
Each platform applies its own rules. Meta uses a 7-day click and 1-day view window by default, which shapes everything from CPMs to reported ROAS in Facebook Ads. Google uses data-driven attribution across a 30-day window for most accounts. These attribution windows can, of course, be changed, based on what you think the business needs. Ad platforms credit themselves for any conversion they can plausibly claim, using their own attribution logic, against their own conversion data, inside their own walled gardens.
ds the same customer sees ads on both Meta and Google before buying, both platforms claim the conversion. That double-counting is the rule, not the exception. Modeled conversions compound the problem by filling in tracking gaps with assumptions the platforms have a vested interest in skewing favorably. None of this is a bug. It’s how the systems were designed.
Total cost and total revenue is the better question
The cleaner number is total ad spend across every platform and total revenue from the business. Most modern reporting tools surface this. Triple Whale, Northbeam, and your own custom dashboards can all stitch it together.
This is a better picture than any single platform’s report, but it isn’t a complete one. Even blended revenue can include double-counted attributed orders if you’re pulling revenue from platform reports instead of from your store. Pulling actual ecommerce revenue from Shopify or Google Analytics, divided by total ad spend, gets you closer to what the CFO is looking at. That’s the foundation of MER (Marketing Efficiency Ratio), we’ll get to that another time.
There’s still a haircut to apply on the in-platform numbers themselves before you trust them for tactical decisions. The exact amount depends on how much your platforms overlap, how long your consideration window runs, and whether you’re running structured incrementality tests to validate. As a working rule, assume platform-reported ROAS is overstated to some meaningful degree, with the higher overlap brands seeing more inflation than the lower overlap ones. The only way to actually measure it is to run holdout tests.
What trusting platforms ROAS costs you
The reporting gap is not theoretical. It produces three concrete consequences that shape how brands spend, scale, and report up to the rest of the business.
When Meta claims a 4x ROAS on a campaign and Google claims a 5x ROAS on a related campaign, both numbers can’t be fully true at the same time. Both platforms are usually crediting themselves for overlapping conversions. Teams that allocate budget based on platform-reported ROAS tend to over-fund whichever platform has the more aggressive attribution model.
We audited a brand recently that was running a 50/50 split between Google and Meta at roughly $10K/month total. The platform numbers looked balanced, but our read was that Meta was likely getting credit for purchases they shouldn’t have. Rather than make a sharp cut and risk breaking something we couldn’t see, we pulled Meta spend down gradually while watching for any drop in brand search volume or direct revenue. There wasn’t one. We landed at a 70/30 split in favor of Google, and the same total spend started producing better blended results.
The opposite case happens just as often and we’ve seen it first-hand. Meta drove the impression and even click that introduced the customer to the brand, the customer searched the brand on Google a few days later, and Google’s last-click attribution claimed the conversion. From the in-platform numbers, Meta looks like a weak channel and Google looks strong. The reality is Meta was carrying the top of the funnel and not getting credit for it. Pulling Meta spend down in that scenario would tank the whole account. Reading the in-platform numbers alone gives you no way to tell which version you’re in.
All of this affects how you scale campaigns and when the returns are thinner than the paper they are written on. The 4x in-platform ROAS that’s actually closer to 3x once you back out overlap still looks like room to push spend. The math, when you do it honestly, often says push less aggressively or hold. Brands that scale on platform numbers without checking against blended performance tend to hit walls. Spend goes up, in-platform ROAS holds steady or improves, and contribution margin collapses because the underlying unit economics were never as strong as the metrics suggested.
How should I view ROAS?
The correction isn’t to throw out platform numbers. It’s to stop treating them as the source of truth.
Look at total cost and total revenue first. Pull total ad spend across all platforms and divide by total ecommerce revenue from your store. That ratio is what your business actually ran on. Marketing Efficiency Ratio (MER) is the formal name for this, and it’s the one CFOs trust because it ties directly to the P&L.
Pair efficiency with causality. Blended numbers tell you what happened. They don’t tell you what your paid media caused to happen. Geo holdouts on Meta, conversion lift studies on Google, and structured incrementality testing are the only honest ways to answer the causality question.
None of this is new. What’s changed is that every other measurement layer has gotten worse, which means teams still relying on platform numbers are working with progressively dimmer signals while their competitors move to better ones.
If the platform numbers and the bottom line aren’t lining up, this is the work that bridges the gap.
ROAS shouldn’t be forgotten,
In-platform ROAS isn’t useless. It’s a directional signal at the campaign level, useful for tactical decisions inside Meta or Google. But it’s not the number that should be driving budget decisions, scaling decisions, or CFO conversations.
When it comes to running Ecommerce paid media, Treat in-platform ROAS the way you’d treat any other report from a vendor with skin in the game. Useful, but not the source of truth.
About the Author: Tony Adam
