Twitter’s Feed Algorithm Sucks

I recently did a side-by-side comparison of my Twitter timelines with the ‘latest’ setting (i.e., only people you follow plus promoted tweets) versus the ‘algorithm’ timeline (what Twitter thinks you’ll like plus people you follow and promoted tweets). Before I get to what I found, the reason I did this is that my timeline has become completely unmanageable: there’s simply too much crap that I’m not interested in seeing. Because I use Twitter, in part, as an RSS feed, when I get bombed with too many tweets, it becomes useless.

When I did the comparison (I didn’t do anything fancy like futz around with the API, I just compared the two timelines at the same time*, what I typically found is that around ten to twenty percent of my tweets were non-followed tweets (commercial tweets seem to occur at the same frequency), but many tweets by people I followed were shoved off the timeline (roughly an equivalent amount). One other time I didn’t lose very many tweets, but there was a ton of non-followers stuck between the tweets which nearly tripled my timeline. This isn’t what I want at all.

If Twitter management actually used their software, they would be smarter and let users have a throughput cap–a user gets a certain number of tweet during a given length of time. Any gaps are filled by the algorithm. Because I really don’t want to lose tweets.

I really don’t:

*You can do this is in an informal way by using one timeline on a laptop and using the other on an unsynced phone.

This entry was posted in TWEET!!!. Bookmark the permalink.

2 Responses to Twitter’s Feed Algorithm Sucks

  1. That is why I use lists.

  2. kaleberg says:

    All of the recommendation algorithms are garbage. I have no idea of why the algorithmic results are so bad. The user has even specified what he or she wants to see. At least provide that along with other stuff. That’s why I don’t bother with things like timelines and feeds, except for good old fashioned RSS feeds when they are available.

Comments are closed.