The problem with the number most accounts run on
Most ad accounts are steered by platform-reported ROAS, which is built on last-click attribution. That number is wrong in both directions at once. It over-credits the platform for purchases that would have happened anyway, and it under-credits the platform for sales that show up later, in a store, on Amazon, or outside the click window.
At the May 2026 Meta Performance Marketing Summit, Meta's own measurement team reported that 31% of incremental conversions driven by Meta get misattributed to other channels. A separate Haus analysis of 640 experiments found Meta's in-platform click attribution under-reports by roughly 15% on a 7-day window.1
So a campaign showing 4x in Ads Manager might be delivering 2.8x in reality, or 5.5x in reality, depending on the brand and the channel mix. Without testing, you cannot tell which, and you end up making budget decisions on a number that is confidently wrong.
The three numbers that replace it
Incrementality-based buying does not throw away platform ROAS; it demotes it to a diagnostic input and decides on three measures a CFO actually recognizes.
1. Blended MER
Total revenue divided by total ad spend across all channels. Because it uses real top-line revenue against real spend and ignores platform attribution entirely, blended MER is much harder to fool than channel-level ROAS. It is the closest thing to a single honest efficiency number.
2. Contribution margin
Profit after the variable costs of making and delivering the sale, measured by campaign and offer. A 3x ROAS that cleared contribution margin in 2023 may not clear it today, because acquisition costs have risen across DTC and tariff exposure has tightened gross margin for many brands. Margin, not revenue, is what tells you whether scaling is worth it.
3. Incremental lift
The sales your ads actually caused, measured directly through conversion lift tests that hold out a randomized group and compare it to an exposed group. Run on a regular cadence, these tests calibrate the dashboards and tell you whether reported ROAS is overstating or understating true impact.
Why more buyers don't work this way
Because it makes reporting look messier in the short term. Triangulating across Northbeam, GA4, Shopify, and Ads Manager, and admitting that the platform number is not the truth, is harder than screenshotting a clean ROAS figure. That short-term messiness is exactly the point: it is the difference between an account that looks fine and a bank account that agrees.
This is the standard every engagement here runs on. You can see how it shows up in practice on the Work With Me page, or start from the basics in what a fractional paid media operator does.
1 Meta Performance Marketing Summit, May 2026 (Damien Wandler, Meta Measurement Products); Haus analysis of 640 incrementality experiments.