Comment by Just_Smith | Hacker News Reader

Comment by Just_Smith | Hacker News Reader

Parent Full thread

Your comment makes me wonder how you believe this data is viewed in the business. I work in BI, who are usually the developers who turn such data in to insights that can be comprehended by the people in businesses who don't know the first thing about databases. Just from my personal experience in such analysis, I think it's highly unlikely they're looking at the "who" behind the weird listening patterns they're reporting on.

It's much easier to report anamolous behavior in the form of anonymous aggregations than it is to report on such a fine granularity as individual listening patterns. A simple SELECT statement with an DESC order by and a SUM aggregation could yield that Sean Spicer statistic, without every touching a specific individual's data.

Such detailed reporting tends to actually be counter-productive, as it usually leads to information overload. I highly doubt that any analyst within Spotify has ever looked specifically at your history.

In this example, you're absolutely right, and if all they had was my name and my listening history, that wouldn't be so bad. But some of the ads do reference specific individuals. Spotify in particular have made it clear that 1.) listening behavior profiles may be "anonymized" in the sense that they have had the names sanitized from them, but there is a myriad of other potentially identifiable data, such as location, that has not been sanitized, and 2.) they are willing and able to comb through this data to find anomalies.

You're right about some of the ads touching the individual level, but if they're following any "best practices" in their data models, those individuals are likely just unique identifiers.

Granted, everything I'm saying is most certainly speculation. There could most certainly be some folks scouring this data with insensitive intentions. I'm just pointing out that that there are actual technical AND practical limitations to doing such reporting effectively.

It is a matter of appearance. I know how databases work, I am a developer myself. The point in was making is that apparently Spotify is very comfortable with the image of them looking through their customer's data and use it whichever way they find fun/profitable. I expect them to analyse listening behaviour in order to recommend certain sponsored content, I expect them to connect my listening to other's listening in order to recommend music to me, I expect them to use my listening behaviour in order to sell insights to the music industry on how to sell better to people with similar music preferences. What I did not expect however is that they would sift through my playlist titles in order to find out whether there is something embarrassing in there for them to use in their advertisements.

What is next?

Spotify ad: Dear Mosselman, who has listend to his playlist "Oh no Veronica, why did you leave me?" 100 times in the last week. Veronica isn't thinking about you anymore, shes listening to _her_ playlist "Thank god for Mosselman's brother and his good lovin'" since day 1.

This doesn't matter. The public audience of these ads doesn't know any of what you're saying. The important thing is how it appears.

If my job was limited to what was understandable to the average consumer, I likely wouldn't have a job.

Reply on news.ycombinator.com