Hacker News Reader

Comment by googlryas | Hacker News Reader

Stats get increasingly expensive and useless to store, but one would think mDAU would be one tagged for high precision long retention storage.

View on HN

I mean, no? Why on earth would I want Thursday the twenty thirds mDAU from a year and a half ago? If it's not bringing value, it's costing money, and it's unlikely to provide value, so why keep it?

Because sometimes it costs more money to discard some data than it does to simply keep it.

When I was in mobile gaming we stored that data for way longer.... Do you really think twitter is throwing away user activity data?

I find it very hard to believe that they don't have this information graphed on a dashboard somewhere.

Depends on whether there was UII in the underlying dataset. It's probable that Twitter kept the mDAU numbers and probably shared them, and Elon's people asked for underlying data that aggregates up to that metric, so they can scrutinize the methodology.

However, saving user identifiable information > 90d is a challenge these days - you have to anonymize it, which defeats Musk's purported goal. (And it might be illegal to share that data w/ a 3rd party.)

When your enterprise value is predicated on user activity stats, you keep your user activity stats.

Because there’s no such thing as a regular Thursday at Twitter scale. Any number of things could have happened that Thursday or around it, major world event, Thanksgiving, last day of Ramadan or whatever. And you might have a valid data question regarding these events.

Heck, not only “what happens on Twitter each day of the week” is a very valid data question, but even by the hour.

Your right twitter might just be incompetent and not keep that data, but this would be one of the most basic forms of data you need for analysis when your entire revenue model relies on that statistic.

Reply on news.ycombinator.com