From “Heavy Purchasers” of Pregnancy Tests to the Depression-Prone: We Found 650,000 Ways Advertisers Label You

Imatge
Àmbits Temàtics

A spre­ads­heet on ad plat­form Xandr’s website reve­a­led a massive collec­tion of “audi­ence segments” used to target consu­mers based on highly speci­fic, some­ti­mes inti­mate infor­ma­tion and infe­ren­ces

What words would you use to describe your­self? You might say you’re a dog owner, a parent, that you like Taylor Swift, or that you’re into knit­ting. If you feel like sharing, you might say you have a sunny perso­na­lity or that you follow a certain reli­gion. 

If you spend any time online, you probably have some idea that the digi­tal ad industry is cons­tantly collec­ting data about you, inclu­ding a lot of perso­nal infor­ma­tion, and sorting you into speci­a­li­zed cate­go­ries so you’re more likely to buy the things they adver­tise to you. But in a rare look at just how deep—and weird—the rabbit hole of targe­ted adver­ti­sing gets, The Markup has analy­zed a data­base of 650,000 of these audi­ence segments, newly unear­thed on the website of Micro­soft’s ad plat­form Xandr. The trove of data indi­ca­tes that adver­ti­sers could also target people based on sensi­tive infor­ma­tion like being “heavy purcha­sers” of preg­nancy test kits, having an inter­est in brain tumors, being prone to depres­sion, visi­ting places of wors­hip, or feeling “easily defla­ted” or that they “get a raw deal out of life.”

Many of the Xandr ad cate­go­ries are more prosaic, clas­sifying people as “Afflu­ent Millen­ni­als, ” for exam­ple, or as “Dunkin Donuts Visi­tors.” Industry critics have raised ques­ti­ons about the accu­racy of this type of targe­ting. And the prac­tice of slicing and dicing audi­en­ces for adver­ti­sers is an old one. 

But the expo­sure of a collec­tion of audi­ence segments this size offers consu­mers an unusual look at how they and their fami­lies are packa­ged, descri­bed, and cate­go­ri­zed by ad compa­nies. 

Because the segments also include the names of the compa­nies invol­ved in crea­ting them, they also shed light on how dispa­rate pools of perso­nal data—­co­llec­ted by trac­king people’s online acti­vity and real-world move­ments—are combi­ned into bespoke, bran­ded groups of poten­tial ad viewers that can be marke­ted to publis­hers and adver­ti­sers.

How do adver­ti­sers think of you?

In a newly disco­ve­red data­base on the Xandr ad plat­form’s website, The Markup found thou­sands of rows of data that indi­cate sensi­tive consu­mer grou­pings known as “audi­ence segments.”

Select a search term to see a sample of results using this ad term. Or, search the data for any keyword.

Note: Some segments from data suppli­ers Grapes­hot, Peer39, and others found in this data­set are “nega­tive keyword” or “brand protec­tion” segments that are used to control when an ad should not appear, rather than “audi­ence segments, ” which are lists of IDs asso­ci­a­ted with consu­mers.

Test the tool

See the full data here: GitHub