What Meta did with 864,000 euros of one client's budget

Method

The source is a single Meta ad account belonging to Real Mad Honey, a direct to consumer consumable brand. The window is the last 365 days. Reporting is pulled in Motion on a 7 day attribution setting, which is the client's own reporting standard, not one we chose for this report. Everything below is published with written permission from the client.

Every ad level number in the two tables is read directly off the reporting screenshots. Where a number is not on a screenshot, it is a calculation from those figures and it is labelled as calculated. Nothing has been rounded up to look better, nothing has been filtered by date range to flatter a single ad, and no ad has been left out of either table. The client bought the media. We produced the creative.

One thing worth saying before the numbers: this is a report about allocation. It is a record of where an optimising system chose to send money over twelve months. It is not a controlled creative test, and it is not presented as one.

The account

Across the 365 day window the account spent €863,979.71 and reported €2,306,295.04 in revenue on 31,809 purchases, a 2.67 ROAS at the account level. That is a real spending account with a real creative library in it, not a small test budget where one ad can dominate by accident. Multiple creative types ran in the same account over the same period, competing for the same budget under the same delivery system.

Five of the ads in that account are Stradvert street interview ads: real members of the public, stopped on a real street, answering a question on camera. The rest of the library is the client's own video creative. Both sets sat in the same account, available to the same optimisation, for the same twelve months.

The five street interview ads

Together the five ads carried about €289,525 of spend and returned about €790,000, a blended 2.73 ROAS. That is roughly a third of the whole account's spend for the year, sitting behind five pieces of creative. RMH603 alone took €107,958.82 of it. It has been live since April and it is still running.

The spread across the five is narrow: 2.50 at the bottom, 2.95 at the top. That matters more than the top line. A single outlier ad is luck. Five ads clustered in the same return band, filmed the same way, is a format doing the same thing repeatedly.

The client's other video creative

Four of the client's own video ads, in the same account, over the same twelve months, received between €225.38 and €371.66 each. Their returns sit between 1.18 and 1.78. These are not bad ads sitting in a bad account. They are ads that never received the budget to find out what they could do.

What that ratio means

€107,958.82 against €371.66 is roughly 290 times more budget behind the top street interview ad than behind the client's next best performing video. No person made that decision. There is no media buyer in this story who sat down and moved a third of a year's budget onto five ads. Meta's optimisation concentrated delivery there, day after day, because those ads kept winning the auction against everything else in the account.

The only manual budget work went the other way. The client raised budgets by hand so that the rest of the creative library would receive any spend at all. That is the detail most allocation stories leave out: the human intervention here was an attempt to spread money away from the street interview ads, not toward them.

The client's own words

"Meta decides where spend goes. If it pushes spend only to the street interview, we have to raise the budget until the rest of the ads get spend at all."

"It's still running and used as a top of funnel ad. The purchases done by other ads and Google as bottom of funnel aren't measured, so the impact of this is actually much higher."

Client statements given during the campaign, kept here as attributed testimony. We have not verified delivery on any date after the reporting window shown.

What this does not prove

This is not a fair head to head test between two creative formats. For that you would need comparable budget behind both sets of ads, ideally split at the campaign level, over a comparable period. That never happened here, because the client's other video ads never received enough budget for a comparison to mean anything. Four ads at a few hundred euros each cannot be measured against one ad at over a hundred thousand.

So the claim in this report is narrow on purpose. The claim is about allocation: given a mixed creative library and twelve months to decide, Meta's delivery system funded five street interview ads with roughly a third of the account and left the client's own video creative in the hundreds of euros. That is the finding. Anything beyond it would be us guessing.

Detail on the top ad

RMH603 is the ad that took €107,958.82 at a 2.77 ROAS. Its cost per acquisition is €25.84. Click through rate is 2.28%. Thumbstop rate is 62.69%, and hold rate is 22.81%. It has been live since April and it is still running.

The thumbstop figure is the one that explains the rest. Just under two thirds of people served the ad stopped on it, which is what gives an optimising system a reason to keep buying more of it. The ad is also used as top of funnel creative, which means purchases attributed elsewhere in the funnel do not appear in its own numbers.

The underlying screenshots

Every figure above comes from account screenshots published in full on the proof page and in the Real Mad Honey account breakdown. If a number in this report does not appear on one of those screenshots, it is a calculation and it is labelled as one.