Why no simulated attribution reports?

One of the problems with the attribution cartel at W3C is, as Rick Bruner explains inThe W3C Is Making A Mistake About Measuring Advertising Effectiveness, a “structural bias toward channels positioned closest to observable conversion activity, including search, retail media, retargeting and click-oriented social advertising.”

The attribution cartel companies haven’t released any data to back up a claim to the contrary. They claim benefits to legitimate sites, but so far, it’s all hypothetical.

The weird part is that several of the attribution cartel companies already have extensive tracking data. So where is the study showing how the attribution reporting would have come out, if the users in the history dataset had had attribution tracking turned on?

Absence of evidence is not always evidence of absence, but absence of a report that (1) could have been produced based on existing data and that (2) could make a strong case for the data holder’s position is at least sus.

A good example of the simulation approach is Inferring Users’ Demographics and Sensitive Interests Using the Topics API by Athicha Srivirote, Muhammad Abu Bakar Aziz, Jeffrey Gleason, Desheng Hu, and Christo Wilson. That study did not use the in-browser Topics API. Instead, the researchers calculated what the Topics API data would have shown based on a conventional browsing history data set—and got the results that many expected (but that Google somehow chose not to checkat the time).

Personally, I would bet that the structural bias shows up in an unavoidably clear way, just because if it wasn’t there we would have heard by now. They have the data to get it through based on data—so why aren’t they using it? Why are they relying on the same kind of rush tactics thatCraig Newmark and the Count warned us about? W3C doesn’t have to YOLO the Attribution proposal. An accurate simulated attribution report would help to “think before you click” and make an informed decision.

Bonus links

Meta workers can opt out of being tracked at work up to 30 min by Laura Cress and Osmond Chia.

Hackers Used Meta’s AI Support Bot to Seize Instagram Accounts by Brian Krebs.