> Data is licensed in aggregate, and typically anything under 1,000 user ids can't be used.
You are correct that the reputable companies I used in my strawman apply practices like this. But it is certainly not the case that this is universal. I've worked professionally with all sorts of a data brokers (for anti-abuse/fraud purposes) and there are many who deal with non-anonymized datasets, especially if we are talking about firms that evolved out the direct-marketing space. Further, there are many ways to make use of semi-anonymized data that while, not strictly joining private information, are able to perform profile appends and data imputation in ways that allow for inferences many would consider privacy violations regardless of the fact the technical construction methods don't directly access specific profiles and are at some level stochastic.
But all that was besides my point, which was perhaps lost with a bad example. HN readers can be both FOR increased use of customer data for acute purposes and AGAINST broader abuses of such data without being hypocritical. There is a relevant distinction to be drawn.
Apologies I always forget that I only have experience with the big guys and haven’t seen what the smaller vendors are doing. I’ve made that mistake before so definitely a blind spot for me.
I do think that what smaller players is doing is the thing people thing of as obviously immoral, but those practices get pinned on everyone else.
You are correct that the reputable companies I used in my strawman apply practices like this. But it is certainly not the case that this is universal. I've worked professionally with all sorts of a data brokers (for anti-abuse/fraud purposes) and there are many who deal with non-anonymized datasets, especially if we are talking about firms that evolved out the direct-marketing space. Further, there are many ways to make use of semi-anonymized data that while, not strictly joining private information, are able to perform profile appends and data imputation in ways that allow for inferences many would consider privacy violations regardless of the fact the technical construction methods don't directly access specific profiles and are at some level stochastic.
But all that was besides my point, which was perhaps lost with a bad example. HN readers can be both FOR increased use of customer data for acute purposes and AGAINST broader abuses of such data without being hypocritical. There is a relevant distinction to be drawn.