Why your ROAS is lying to you.

Pull up your weekly performance dashboard. The big number at the top says ROAS 4.8x. You feel good. The board feels good. The account manager at the agency feels great because that number justifies more spend, more retainers, more everything.

Here's the bad news: that number is almost certainly wrong. Not directionally wrong — quantitatively wrong, often by 30 to 60 percent. And the worst part is that the people inside the dashboards know this and don't tell you, because the dashboards are the product they sell you.

1. Last-click attribution is a flat lie

The default model in most ad platforms gives 100% of the credit to the last paid touch. So if a customer hears about your brand from a friend, sees you on the podcast their boss listens to, gets a display impression on the way home, and then searches your brand name and clicks the Google Ad, the Google Ad takes all of the credit.

Brand search ROAS is the most overstated number in marketing. It shows up everywhere. It's the reason your “branded campaigns” look like miracles. They're not miracles. They're a tollbooth on the highway that your other channels built.

2. Walled gardens grade their own homework

Meta and Google attribute conversions to themselves. They both count the same conversion. You'll see Meta claim 1.2x ROAS and Google claim 1.4x ROAS on the same customer. Add them up, and you'll show 2.6x of attributed revenue against a customer who'd have bought anyway.

They have no business incentive to deflate their reported performance. You should not be using their numbers as the system of record. You should be using your own.

3. View-through is doing heavy lifting it didn't earn

View-through attribution credits a conversion to an ad that the user saw but never clicked. Combined with broad audiences and cheap CPMs, this becomes a credit-laundering machine. Your programmatic vendor will happily claim a $300k uplift you cannot verify.

The cheap test that tells you the truth

You don't need MMM. You don't need MTA. You need a holdout.

Pick a geo. Turn off paid in that geo for two to four weeks. Watch organic revenue. Compare to a matched control geo where you kept spending. The delta is what your paid was actually doing.

Most clients are terrified to run this test. The ones brave enough to run it discover that 20 to 40 percent of their paid spend is credit-stealing from organic and brand — and they reinvest that spend into channels where the lift is real.

What we do at Elkhawaga

Every paid engagement we run includes a quarterly incrementality test on at least one channel. We've killed our own retainers this way — but we've also doubled spend on channels where the number held up. The math doesn't care which way it goes.

If you're running paid and you've never done a holdout, your ROAS isn't lying to you. You're just letting someone else do the lying for you.