After you #brokereplies we were promised friend recommendations, so where are they Twitter?

It seems like an eternity ago, but in May 2009 Twitter made what they described in their blog as a “small settings update” to their service unexpectedly and in doing so ‘broke’ an option for how @replies worked for a small percentage of their users. Now it just so happened that the 3% of users who were affected were a vocal lot and managed to create a significant amount of noise on twitter with the #fixreplies hashtag (that is still in use today). The level of noise was disproportionate to the number of users affected, as many users were upset at twitter removing the choice that was available to them previously without warning.

The unexpected change in Twitters service was arguably the right move from a consistency of service perspective. Now users are guaranteed that @reply tweets they send are visible only to people who are following the person in question and not confusing people who aren’t close enough to the conversation stream. That said, an important aspect that was noted in the initial post highlighting the service change on the Twitter blog is “The Importance of Discovery”, many people who were fans of the option to view all @replies used the option to discover new and interesting people to follow who people they followed were interacting with. Essentially the people you were following, when they interacted with others were giving you an implicit cue that this person was worth following too. This act of discovery was important and has not yet been replaced with a suitable option by Twitter despite assurances that this would occur:

“We’re hearing your feedback and reading through it all. One of the strongest signals is that folks were using this setting to discover and follow new and interesting accounts—this is something we absolutely want to support. Our product, design, user experience, and technical teams have started brainstorming a way to surface a new, scalable way to address this need. “

If you have been reading this blog before, you would have no doubt encountered by belief that Twitter holds some advantages over Facebook but when it comes to new user discovery this is one area that Facebook has it in spades over Twitter with its Friend Suggestion feature. Now Twitter does have a suite of tools comparable to Facebook’s Friend Finder tools, but these are limited to finding users who you already have existing relationships with (ether in person, via IM/Email etc). Twitter offers a  Suggested Users page, however this is nothing more than a glorified list of celebrities and news outlets, how is that useful?

It wouldn’t be hard to create an algorithm to examine the relationships you have on twitter and create a list of suggestions based on you you already follow on twitter. The only issue is having access to the dataset and a few CPU cycles to mine through it. This could be offloaded to an external party through the  Twitter API, but involves more data than is feasible to be sent to a third party for processing. (If I follow 1000 people who each follow 1000 people, the API needs to return 1000 x 1000 contacts for cross matching and comparison, which is unfeasible for a third party service.

With Twitter sitting on this gold mine of data, what is stopping them of releasing a user discovery/recommendation engine? The code necessary to create such a feature work is not impossible, just potentially computationally expensive. Twitter has already revealed that the original @reply options meant the service was potentially going to hit a server capacity wall that it could not overcome (A combination of it’s original design and trouble scaling due to Ruby on Rails) is this why? Twitter has grown by 1444% over the last year and this obviously has it’s own set of problems, not only do new users put a strain on the system, but the number of individual interconnects (followers/following) in the network increase the activity of the system. In addition to this, as users increase the number of their followers there is a correlation showing that this increases the average number of tweets per day further putting strain on the servers of Twitter.

It seems like an obvious and relatively straightforward feature to implement, one which Twitter assured us would be forthcoming after they #brokereplies. But is there more to this delay in implementation than Twitter are telling us?

One thought on “After you #brokereplies we were promised friend recommendations, so where are they Twitter?”

  1. Nicely written David,

    I’m going to retweet this over to my twitter pal Daniel @dtunkelang Tunkelang who is deep-dive when it comes information access and discovery.

    I’ve heard about these #lists on Twitter but seen anything yet? I heard some selected users had access to these (which as I understand it are akin to groups in TweetDeck or FriendFeed)

    Speaking of FriendFeed I do find it useful in finding new relevant people to follow on Twitter, by reviewing the threaded discussions from time to time. The authors of said discussions almost always have Twitter accounts.

    The other thing I do which helps work around the #fixreplies debarcle, is occasionally review an interesting user’s Twitter stream for others they are engaging with, then follow those users if warranted.

    Cheers,
    Tony Hollingsworth

Comments are closed.