Twitter today killed its “suggested users” feature (which Ev said he’d do way back at Web 2 in October), and replaced it with a more sophisticated approach. In a blog post explaining the move, the company elaborates:
We’ve found that the power of suggestion can be a great thing to help people get started, but it’s important that we suggest things relevant to them. We’ve created a number of algorithms to identify users across a variety of clusters who tweet actively and are engaged with their audiences. These new algorithms help us group these active users into lists of users by interests. Rather than suggesting a random set of 20 users for a new user to follow, now we let users browse into the areas they are interested in and choose who they want to follow from these lists.
Yep. Back in May of last year I wrote:
It strikes me that a few more structured steps in the sign up process could really pay significant dividends for Twitter. Perhaps a “follow wizard” that asks a few questions, and makes suggestions based on input from the new user. Let us drill down by category: Business:Technology:Internet, or Health:Diseases:Cancer. The ontology isn’t very complicated – mapping users to it is a bit more complex, but not impossible.
It took a while, but it looks like Twitter is doing just that and even more, if the algorithms they’ve cooked up prove robust. I look forward to seeing how this changes newbies’ impressions of the service.
But here’s what I wonder – why can’t everyone do this? Is it limited to just new accounts? To my mind, it shouldn’t be. All Twitter users would benefit from this new feature….so hopefully Twitter will open it up to all of us.
Update: You can use the new features by navigating here.
John, unfortunately this only solves 25% of the problems with the suggested user list. It still is closed, opaque, and incomplete. Here’s a blog about why I say that: http://bit.ly/8cZYeV
John,
Thank you for informing me of this. It certainly makes sense for newbies to Twitter…
I hope you have a rocking year!
The Franchise King®
Joel Libava
We’ve build an algorithm to address similar problems for blogs (mapping communities) and it works pretty well.
I doubt Twitter will solve this by just asking for self declared interests. This is a very poor signal, very static and many users will game the system.
To make this work one has to mix:
1- topical analytics (i.e based on what I’m writing, here are my centers of interest) – lots of fresh and deep content here.
2- topology analytics ( who am I already connected to)
3- style and community codes ( ! we’ve a patent pending on this)
Twitter has very little material for 1) – (140 characters is not much … but many users add a link to their site which brings lots of content to mine)
Twitter should be the king of 2… but it only works as people start to connect so no use for newbies
As for 3, well. I doubt the 140 char leaves enough room for expression of style and community codes. Same tactic like 1) could work, but only for the people that publish.
Best
John, R U kidding me? This is an epic fail of a post 🙁
How much did you get paid to write that? — and who paid? Eric Schmidt? Bill Gates? (check the technology suggested users list — and/or see http://english.net.in/did-billgates-ericschmidt-pay-more-than-leola … and more like it in my @nmw twitter stream)
You say “if the algorithms they’ve cooked up prove robust” — well, I think I have just disproven that hypothesis.
The so called “algorithm” for Twitter SUL is not an algorithm — unless you call @ev’s + @biz’s hand-picked chums an algorithm (and that seems to be quite a stretch).
I would say the case is closed with Twitter’s SUL — it’s not “game over” for twitter yet, but the poor and misguided management reeks of writing on the wall.
😐 nmw
With all respect, methinks Twitter’s attempt at suggestions are complete #fail s.
Twitter is about PEOPLE. Almost every single suggestion this service makes is an organization of some sort.
If I want to find interesting follows in science, the last things I want to be inundated with are newsfeeds or pointers to captive blog posts.
This system shows no insight whatsoever as to what *people* are interesting to follow in any given subject area.
Sorry. #Twitterfail .