"Attributed sales are the sales your ads brought in"

“The report says ads drove this much in sales. So that's what ads are worth.”

Or the opposite, from the skeptics: "Half of that would have sold anyway. Ads just take credit for organic orders."

Both are claims about the same number, and both are usually made without measuring it.

Attributed sales are read as caused sales. If an ad was clicked within the attribution window before a purchase, the purchase belongs to the ad. Remove the ad, and that revenue disappears. ROAS and ACOS are then treated as the return on the ad.

Every conclusion below rests on these. Reject one and you can see which conclusions fall with it.

• ✅Attribution records a real touch. The click happened. That part is accounting, not opinion.
• 🔶Without the click, the purchase would not have happened. This is the incrementality question. Attribution does not answer it; it can't, because it never sees the world without the ad.
• ⚠️The overlap with organic is small. It depends on who the ad reaches. Ads that reach existing customers overlap heavily; ads that reach new ones overlap less.
• ⚠️The answer is the same across campaigns. It isn't. Account-level totals can hide one campaign type that is pure harvesting.

Conclusion: there is no single "cannibalization metric." I stack four rulers, from weakest to strongest evidence — buyer mix, organic response, touch vs attribution, and controls with placebo arms. Only a holdout — actually switching ads off — speaks to causation.

The full step-by-step decomposition, including how AI joins the reports, is in the course.

Experience

The strongest evidence I have is a holdout I didn't design as one: I switched off all advertising on one of my smaller brands for two weeks. Retail revenue fell 59% by the second week, and the weekly revenue lost was almost exactly the revenue ads had been credited with before the shutdown (about 99% of it). So on that brand, attribution was close to incremental. But the true incremental return sat about 25% short of break-even. Attributed ≈ incremental did not mean profitable.

Verdict: ⚠️ partly. Attributed sales can be close to incremental when ads reach people who don't already buy from you; they overstate it when ads reach people who do. Which case you are in is a measurement, not an assumption — and it differs by campaign type.

Type, in detail: Experience (one account-level shutdown; one campaign-type split).

Have data that says otherwise? Share a case. Reversals are published.