Thoughts on Twitter’s New Ranking System for Search

image userpic
May 8, 2009 | BY Brad Wellman
  


Brad Wellman
Spyder Trap Online Marketing

Yesterday I wrote about how Twitter will soon begin indexing links for search. Ending that post I asked how you think Twitter should base their reputation ranking system; should it be based on the number of followers a user has? What about the number of retweets they have? Maybe some other way?

Twitter is great for finding out that something is happening right now. Its great for getting headlines and brief information about current events, but where it struggles is allowing users to find more information relative to these topics. It is hard for users to sift through the millions of tweets to find more detailed information.

In a post by Stan Schroeder earlier today, he mentioned, “this reputation ranking system is far more complex than a simple combination of factors such as followers and retweets. The system needs to be contextual; it needs to recognize which tweeple are important for a certain keyword or phrase.”

This is an important thing to consider, because if Ashton Kutcher were to tweet about something related to Google, even though he has the most followers on Twitter, it shouldn’t matter as much as if say, Matt Cutts were to tweet something related to Google, given his role within Google and his familiarity with the topic. Tweets relating to Hollywood news or upcoming movies? In this context is where Ashton is relevant and can shine. Much as tweets from the White House are more relevant coming from Obama, rather than say, myself (even though Obama has 1.1 million followers compared to my 700, that’s beside the point).

This is a complex problem because the Twitter search algorithm needs to analyze keyword density within each tweet and connect that to the number of retweets, the user tweeting it, how many followers they have, etc. Hopefully it is something the folks at Twitter are working on ironing out. How do you think they should handle the reputation ranking system?

Tags: , , .
Bookmark the permalink.
  • http://www.sjhalestorm.wordpress.com Scott Hale

    Brad,

    I prefer the idea that retweets provide a good portion of the weight in a ranking system. I’m less concerned with how many followers a user has than I am with the amount of people passing on the information. Naturally, the power users will have more retweets, but Twitter has been fairly successful in providing a platform where generally useless links die relatively quickly.

    As you mentioned, the search needs a contextual element that will set it apart from a popularity contest. On the other hand, indexing users on Twitter will inevitably give specific individuals power in controlling content. I do not necessarily want links from Matt Cutts when I run a Twitter search on Google.

    On Google’s level, the problem is not so clear, but consider a local specialty store. Should the owner get the contextual jump in the ranking system? I want to know the good, the bad, and the ugly about this store and the owner is not likely to provide the latter.

    Thanks for the post, another solid piece of work from the Spyder Trap team. Excited to see what happens.

  • http://www.twitter.com/bradwellman Brad Wellman

    Thanks for the comments Scott, much appreciated. You definitely bring up some good points of concern regarding the Twitter ranking system, and the crew over at Twitter definitely has their hands full trying to figure out one that will work. I liked your point about the local specialty store, as it will be difficult for Twitter to discern who is relevant for a given topic beyond those that work for the “local specialty store.”

    Thanks again! It should definitely be interesting to see how it all pans out…