The wide range of data the article hints at and the guesses at connecting data to users suggests that this data is probably full of inaccuracies. That is one of the really scary things - companies are making decision about me based on outright wrong data.
This can be brutal for applications like credit worthiness, but I'm still worried about mistakes in data used for more mundane decisions like who to offer a discount to or which passenger to bump to business class.
This type of data isn't really used for business decisions, since an audience like "good credit score" (highly highly regulated and almost never used, fyi) contains like 150MM unique identifiers (cookies, device ids, etc).
Everyone using this data knows it contains inaccuracies, but it is much much better than not having any data at all.
This can be brutal for applications like credit worthiness, but I'm still worried about mistakes in data used for more mundane decisions like who to offer a discount to or which passenger to bump to business class.