Black Friday ROAS: What Platforms Report vs Orders Shipped

In the December review you will be holding three Black Friday ROAS figures from three systems and one order count from the store, and you will be asked which is true. The Price of Being Found gives a defensible number four properties. Platform-reported ROAS has none of them, and the review is decided by whether you bring a number that has all four.

The loss in play is a specific one: a budget you ran well, re-graded by a report you did not write. Octalysis calls the reflex Rightful Heritage. The fix is not a better argument about whose ROAS is right. It is a different kind of number.

Why the three figures cannot be reconciled

The book states it as a category error. The ad platform reports conversions it can associate with exposure to its own inventory, inside a window it defines, under a rule it sets. The analytics tool reports sessions it could resolve to a source, after consent and cross-device joins. The commerce platform reports orders, the only figure that is a fact about money. Taking the middle one for the board pack, which the author admits to having done on the agency side, is not reconciliation. It is choosing which question to answer without saying so.

Peak week makes it worse. Every channel is on, every window is open, and the same Cyber Monday order sits inside Meta's 7-day click window, Google's, and the email platform's at once. Your Black Friday ad budget is finite. The platforms' claims are not. explains the arithmetic.

The four properties

A number you can defend, per the book:

1. It comes from a source that does not sell you media. The store, or finance.
2. It states its coverage. Revenue of X, of which the systems can attribute Y, coverage of Y divided by X. The gap is reported, not hidden.
3. It states its design. Experimental, quasi-experimental, or observational, and if observational, that it is a description rather than a causal claim.
4. It states its interval and its floor. Not "the channel delivered 12% lift" but "12%, interval 4 to 20, from a design whose minimum detectable effect was 8.3%."

Reported ROAS fails the first property by definition. It has no coverage statement, no design statement, and no interval. It is a single confident number, and the review will treat it as one until someone brings the alternative.

The table to bring instead

Four of the five rows need no vendor and take an afternoon. The fifth needs either a holdout that started by 2 October or a causal read on the GA4 export, which produces per-channel incremental ROAS with intervals from data you already hold.

Why the smaller number wins the room

The book is candid that presenting the five numbers will, the first time, make marketing look worse: coverage lower than expected, claim ratio above one, an anchor date that reads "never". It is also candid about why that is the right first result. The alternative is a set of confident numbers wrong by an unknown factor in an unknown direction, which is the same position with worse information about itself. The organisations that get past this are the ones where somebody senior says out loud that a smaller number they can defend is worth more than a larger number they cannot.

That sentence is the one to get said before December, not during it.

What to do this week

• If you have to defend the number: build the table above for Cyber Week 2025 as a dry run. The first four rows are already available.
• If you own the budget: decide now which number the December review will be graded on. If it is the store's order count, say so in September, and every platform report that follows becomes context rather than verdict.

Four questions your CFO will ask about the Black Friday budget turns this table into the answers. The calendar has the dates.

As of 9 September 2026. The four properties, the five-row structure and the three-systems identity are from The Price of Being Found (Edition 2.10), Chapters 9 and 19.