Your ROAS is lying to you

Open Meta Ads Manager and it will tell you a confident story about how many sales it drove. Open Google Ads and it tells you a different confident story. Add up the conversions each platform claims and you will often find they have collectively taken credit for far more revenue than your business actually booked. This is not fraud. It is the predictable result of every platform marking its own homework.

The number most teams trust to steer spend is platform-reported return on ad spend, and platform-reported ROAS is the single least reliable number in the marketing stack. Understanding why is the difference between scaling what works and pouring budget into a channel that was only ever taking credit for sales it did not create.

Every platform reports itself as the hero

An ad platform has one structural incentive: to show you that spending more with it is a good idea. It is not neutral, and it was never designed to be. The conversions it counts are the conversions it can see and wants to claim, measured with rules it sets, inside a window it chooses. None of that is a conspiracy. It is just what happens when the scorekeeper is also the player.

So when Meta says it drove a 4x return, that is Meta's view of Meta, optimized to make the next dollar look worthwhile. The question that matters to your business is different: of the revenue Meta is claiming, how much would have happened anyway, and how much did Meta actually cause? Platform reporting is structurally unable to answer that.

Where the reported number drifts from reality

Several forces pull platform-reported ROAS away from the truth, and they almost all push in the same direction, toward overstating.

Last-click bias. Whichever channel touched the customer last tends to collect the credit, even when an earlier channel did the persuading. Channels that sit near the finish line look like heroes by position alone.

View-through claims. Some platforms count a sale if a person merely saw an ad, without clicking, within a window. Show enough impressions to people already heading toward you and you can manufacture a large view-through number that reflects timing, not influence.

Double counting. One sale, several platforms, each claiming it. Nobody is reconciling across them, so the same revenue gets counted two or three times in two or three dashboards.

Signal loss after privacy changes. Since iOS app-tracking changes and the steady decline of third-party cookies, platforms see less of what actually happened, so they increasingly model and estimate conversions rather than observe them. Modeled conversions are a guess dressed as a measurement.

Branded demand capture. Spend aimed at people already searching for you converts beautifully and proves almost nothing, because those people were already on their way.

The cost of believing the wrong number

If reported ROAS were wrong in a random direction it would be annoying but survivable. It is wrong in a consistent direction, which makes it expensive. Over-credited channels look more efficient than they are, so they attract more budget. As budget flows in, those channels spend more to reach people who were already going to convert, the reported return stays high, and the real incremental return quietly falls.

The result is a business that is scaling its harvesting and starving its demand creation, while every dashboard reassures everyone that things are going well. By the time growth flattens, the spend has been miscalibrated for months and nobody can point to the moment it went wrong, because the numbers never flashed red. They could not. They were measuring the wrong thing.

Three lenses, and what each is honestly for

There is no single number that tells the whole truth. Serious measurement uses three lenses together, each covering the others' blind spots.

Multi-touch attribution

Multi-touch attribution distributes credit across the touches in a journey rather than dumping it all on the last click. It is useful for understanding sequence and path, and it is only as good as the tracking underneath it. With broken or partial data it produces precise-looking nonsense. Treat it as a tactical view, not a verdict.

Marketing mix modeling

Marketing mix modeling steps back and uses statistical analysis of spend and outcomes over time to estimate each channel's contribution, including channels you cannot cookie at all, like offline and brand. It is privacy-durable because it does not depend on tracking individuals. It is better at the strategic question of how to split a budget than at the tactical question of which creative won.

Incrementality testing

Incrementality testing is the closest thing to ground truth. You withhold a channel from a slice of your audience and measure what changes. It answers the only question that matters for spend decisions: what would not have happened without this. It is the referee the other two lenses are calibrated against.

None of these replaces the others. Attribution shows the path, mix modeling sets the budget, incrementality settles the arguments. Used together they triangulate toward a truth that any one of them alone will miss.

Fix the signal before you trust any of it

Here is the order that matters, because it is the one most teams get backward. Before you invest in any measurement lens, you have to fix the data feeding it. A sophisticated attribution model on top of leaking tracking is a precise answer to the wrong inputs.

That means moving conversion tracking server-side so it survives browser and privacy restrictions, sending clean and deduplicated events through each platform's conversions interface so the signal is restored rather than modeled, and making sure one sale is recorded once, not three times across three tools. This is unglamorous plumbing, and it is the foundation everything else stands on. Restore the signal first. Then measure. Then decide.

Your ROAS is lying to you, but not out of malice. It is doing exactly what it was built to do, which is to make its own channel look good. Your job, and ours, is to measure what your marketing actually caused, and to spend against that number instead of the flattering one.