Meta claimed a quarter of the revenue. We went and found the rest.

Almost nobody gets to compare those two figures cleanly. The ad account sits behind one login and the store sits behind another. The date ranges disagree, the currency conversion disagrees, and by the time somebody has stitched the two together the quarter is over. So the platform number wins by default, because it is the one that arrives already formatted in a dashboard somebody built to make it look authoritative.

Curated Chrome was ours. We built it, ran it for seventeen months and sold it in March 2026. That means we hold the complete campaign export and the store's own revenue total, for the same dates, in the same currency, with nobody to please and no renewal to protect.

The number

Over that window the ad account claimed just over a quarter of what the store took. On the whole life of the account, Meta reported 140 840 euro of revenue against 48 242 euro of spend, a 2.92x return. The store's own number, at the share we measured, was closer to 529 000 euro.

The two wrong conclusions

The first is that Meta was only a quarter of the business, so the budget should come down. That reading assumes the other three quarters would have arrived anyway. They would not have. The account finished with 124 000 followers and 17.6 million reel views, and almost none of that reach was bought directly. It was bought sideways: paid distribution put the product in front of enough people that the organic and the email had somebody to talk to.

The second is that attribution is broken, so the platform figure should be thrown out and the whole store total credited to advertising. That reading is more flattering to us and it is just as wrong. Some of that revenue is repeat purchase from customers acquired long before. Some of it is brand search from people who heard about the product somewhere else entirely.

Both readings are attempts to collapse two numbers into one. The honest move is to refuse. The same gap is what makes an account look like Meta suddenly stopped converting in a week when nothing about the advertising changed.

What the other seventy three per cent actually was

• Organic reach that the paid distribution built. Reels that were promoted, went out to a cold audience, earned a follow, and then reached that person for free every week after.
• Email. 72.7 per cent of email revenue came from automated flows rather than campaigns, and a flow only fires when somebody has already arrived. Ask where they arrived from and you are back at the ad account.
• Repeat purchase. An average order of 253.77 euro across 555 purchases, in a category people come back to.
• Brand search and referral. The cheapest revenue in the account, and entirely downstream of somebody having seen the thing first.

The gap between what the platform claims and what the store banks is not a measurement error to be closed. It is the size of the effect the platform cannot see.

The only test that settles it

No attribution model can answer the question you actually have, which is what would have happened without the ads. Every model is a guess wearing the clothes of a measurement. A spend change can answer it. Hold everything else still, move spend by an amount large enough to matter, leave it long enough to clear the purchase cycle, and watch the store total rather than the ad account.

That test costs money and it is the only honest one available. Most brands never run it, then spend years arguing about attribution windows instead. If you have never once turned a channel down on purpose and watched what happened, you do not know what that channel is worth, and neither does anyone selling you reporting on it.

What we changed in the weekly report because of this

1. The platform figure. What the ad account claims, unedited, including the weeks where it flatters us.
2. The store total. Straight out of the shop, for the same dates, counted once.
3. The blended figure. Store revenue divided by every euro of ad spend across all channels. It is a blunt number and it is very hard to fool.
4. The gap, named. Not buried. If the platform claimed 40 per cent of the store this week and 26 last week, that is worth a sentence, and the sentence goes in.

Three questions worth asking any agency

• Show me a week where your channel's own number and the store's number disagreed. What did you write about it?
• What is the blended return, and is it in the report every week or only when it looks good?
• Have you ever recommended spending less on the channel you are paid to run?

You can see how we answer the first two on the Monday report, and the full account this note is drawn from is written up in the Curated Chrome case study.