Twitter is all the rage as illustrated in this report on the State of the Twittersphere report. The information in this report is based on data from Twitter Grader.
Twitter Grader is a tool that measures the authority and reach of a twitter user. It has really taken off since it’s introduction a few months ago. The software has now graded over 900,000 individual twitter accounts and gets used over 20,000 times a day. The Twitter Grade is a score from 0–100 based on the power and authority of a twitter user. It is used to build the Twitter Elite list — a compilation of the top twitter users. (We also generate a list of the top twitter users by keyword and by geographic location ).
The Twitter Grader algorithm has withstood a fair degree of scrutiny — exacerbated by the fact that the rankings generated don’t always agree with some other “top user” lists that rank purely on the number of followers. But, one factor in the algorithm that has generated some debate is the reduction in grade that occurs when a given twitter user has a low follower:following ratio. This is when (on a relative basis) a user has a lower (or equal) number of followers than the number of people they are following. Stated different, on average, users who are following a bunch of people get a lower grade (all other things being equal). The common complaint about this is “why would you penalize users that are following other users — doesn’t that go against the spirit of social media sites like twitter?”. This is a fair question. And, I have a reasonably fair answer: One of the surest signals of a low-authority twitter account is that it has a very low follower/following ratio. Using the follower:following ratio in the algorithm works great — except when it doesn’t. Like when power users like Guy Kawasaki (
@guykawasaki
) don’t make it into the Top 100 list.
Clearly, there was something wrong with the algorithm
. Guy’s a highly connected, highly engaged, highly authoritative twitter user (and I’d say that even if I wasn’t a raving fan, which I am).
The problem is, it can be a bit difficult to tell the difference between a spammy twitter account that is using automated follows and a high quality, authoritative account. It took some near-sleepless nights, but, I think I’ve finally figured it out. Finally, the Twitter Grader algorithm has been improved . Significantly. In addition to the existing factors, Twitter Grader now looks at the degree of engagement a given twitter user has. It looks at how well a given user is fostering conversations in the twitter community. Clearly, Guy is really, really good at fostering conversations. So, Guy is back on the Twitter Grader elite list of top 100 twitter users, where he belongs. So are a bunch of other powerful twitter users — and a few new surprises.
If you’re a power user on twitter and think your grade and ranking was too low before, please try it again. And, if you’ve got ideas for how we might improve the software and make it more useful, please leave a comment. And, if you’re looking for me , you can find me @onstartups . (Oh, and for the record, my Twitter Grade dropped a lot after this algorithm update — which is also as it should be).
Happy twittering.
Webinar: Twitter for Marketing and PR
|
Download the free webinar for tips and tricks to drive inbound marketing using Twitter. |
Inbound Marketing Blog Feed
M Freitas 2:38 AM on January 12, 2009
Yes, Guy is highly connected, authoritative. But you have to agree that most of hist Twitter posts are actually promoting his Alltop service...
DocHobbes 2:48 AM on January 12, 2009
What about creating a way to grade users on location not matching their generic grade, but on how they interact with people around their location?
Dharmesh Shah 2:50 AM on January 12, 2009
DocHobbes: Not sure exactly what you mean. We do show power users within a given location. Perhaps you're suggesting a separate grade given within a location? Or, basing something on tweets to others in that location?
Will Kessel 3:06 AM on January 12, 2009
I'm sorry, I don't get it: how can a guy like @Dennis_Kucinich follow only 2 people (one of whom is his wife), have 628 followers and only a mere 28 updates have a score of 98.5 (up from this afternoon, when I last checked) when I have 208/207 and 4,165 updates with a grade of only 95 -- *and* I am engaged with and in conversations with my followers/friends?
I looked at some of Kucinich's followers: some with 8 updates, some with 5, another with 10, one more with 13... one guy had 125 updates with about a dozen followers, following about 2 dozen... and few of them over 25 followers! How strong is that network?
I'm sorry, it still doesn't make sense; something is still very wrong in the state of Denmark. You're weighing the ratio far too much, if you're touting the engagement idea. Looks to me as if *engagement* has nothing to do with it at all...
Count me as still skeptical.
Dharmesh Shah 3:12 AM on January 12, 2009
Will: Currently, the "engagement" factor is in alpha and only being applied to a limited number of users.
Figuring out the engagement score if resource intensive and I don't have it optimized enough to run it on all users -- yet. Stay tuned, I just need a day or two more.
Arthur C. Van Wyk 3:14 AM on January 12, 2009
is there not a way to make the badge reflect the grade more accurately. mine shows 99, but i'm actually 98 point something.
Dharmesh Shah 3:21 AM on January 12, 2009
Arthur: Yes, I can do this. I need to make the image a bit bigger (or the font smaller). Will add it to the list of future updates.
Dochobbes 4:55 AM on January 12, 2009
Dharmesh - It would probably have to be a second grade. For example, how active am I in the twitter community, but also how active am I with my local community. How many people in my local community do I follow / interact with? Does that make sense?
Destry Wion 8:08 AM on January 12, 2009
Valiant attempts here with ranking the top birds, but I would have to say there's a long way to go yet with the algorithm. That's not to be negative, just a nod towards needing more weight on tweeting quality rather than flock volume (activity), regardless of what the numbers are in the follower:following ratio.
For example, bird X could be following 10 other birds and followed by 5,000 bird X wannabes. Even if bird X never sleeps and holds continuous tweet fests 24/7 to followers and non-followers alike, it means nothing if dialog is a lot of...
birdX My cat just farted.
birdZ @birdX OH. MY. GAWD! birdD @birdX LOL!!
...and the like. That might sound like pre-teen banter, but even PR pros can be just as futile in long stretches (who isn't?), and tweets like that only add to the Twitter carbon dioxide calculations (Yo!).
I know it would be near impossible to do (until Google buys Twitter and adds there own wizardry to things -- heh) but the rankings do need some kind of quality factor; maybe a following:tweet-quality ratio? The challenge, of course, is agreeing on what makes a quality tweet (undoubtedly subjective).
To my mind, a quality tweet is one that points me to something I didn't already know but have interest in -- principally for professional interest (work, research, etc). Presumably I've done the footwork to this point and fly with a flock of birds that have common interests; so all I do now is look for that link or factoid that gives me something I didn't have one tweet earlier. Those kinds of tweets are, to me, high quality tweets and what I give credit for to other birds if I see them. There might be other classifications of quality (e.g., emotional, entertainment...), and so maybe different quality algorithms are needed (i.e., multiple rank types, not just a single "100 top birds"), but one way or the other the rankings somehow need a volume-of-quality-tweets weighting.
Mary Fletcher Jones 9:05 AM on January 12, 2009
We need metrics of some kind, and I think Twitter Grader is a lot of fun. I have a respectable grade of 84. So I think the best use of Twitter Grader is to tell part of the story (as with all quantitative measurement tools). Alone, it can't really reveal how well someone may represent him/herself, a brand, a point of view, etc. on Twitter.
The quality of the Tweets is more important that the quantity. More than 2 tweets a day is excessive, in my book, and posting a tweet that would not be of general interest to your followers is self-indulgent (who cares if you're stuck in an airport or if you like a certain brand of toothpaste? -- real tweets from people I stopped following).
An occasional personal update makes you a real person and engenders trust; constant personal updates make you look like a raving egotist!
I follow people who post useful and relevant links and announcements.
I can't understand why people don't use DMs or even email to carry on some conversations that obviously don't need to be broadcast to their followers. That's why I follow relatively few people, and I had to drop a few.It's funny, I see this kind of competitiveness about rank and elite among certain social media marketers (thank goodness, not all of them), and it just provokes a lot of eye-rolling. Reminds me of certain bloggers who are all about their Technorati ranking. Makes me think, get a life! Because, bottom line, who really cares what your Twitter grade is? If you're a nonprofit org. and you've connected with one person and helped them make a choice that improved their quality of life, or you're a business and you've established an online presence that brought in sales that quarter, that's impact.
That said, please follow me @FletcherPrince ;) Sorry, couldn't resist :)
Destry Wion 9:42 AM on January 12, 2009
Mary Fletcher Jones said:
"Because, bottom line, who really cares what your Twitter grade is? If you're a nonprofit org. and you've connected with one person and helped them make a choice that improved their quality of life, or you're a business and you've established an online presence that brought in sales that quarter, that's impact."
EXACTLY!
And I think that's a relatively new perspective (circa 2008) of how to use Twitter. I'm a spanking new bird and that's how I see Twitter being worth my time (though I'm only human and might fall from grace occasionally). It will be interesting to see if early adopters -- birdX I'm standing in in line for a latte at SB. -- will influence new users in bad ways or if intelligent 'How to...with Twitter' articles will turn tides towards economy and value.
At any rate, Twitter Grader is indeed curious and the near future will probably see some impressive new Twitter measurement tools come out.
Oh, and you've got to follow me @Wion for rare but earth-shattering intelligence. heh!
Destry Wion 9:50 AM on January 12, 2009
Poop. Italics are wonky in that post. Only italicized text should be "birdX I'm standing in in line for a latte at SB." and "heh!" at the very end.
A nice community like HubSpot should have previewable comments for self-editing blunders, no?
Shutting up now.
Liz 10:42 AM on January 12, 2009
Thanks for listening to my ONE complaint about Twitter grading systems, the higher valuation of those who follow the fewest number of their followers (0-10% of followers). I realize that high profile Twitter users can't really follow & respond to thousands of followers' Tweets but at least tey are receiving them and they usually respond to some of them. Having tens of thousands of followers and following less than 100 people just turns a person's Tweetstream into a broadcast, all out-going Tweets and no conversation between users.
IMO, a social community is not about each individual posting their personal diary for everyone to read but about interaction and communication between users on both weighty and lighthearted issues.
I'm sure you'll get a lot of feedback on these changes and it is odd to see some people with the largest followings farther down on the list than you normally see them. But I think occasional tweaking of your grading formula promotes a healthy discussion of how we value our online relationships and measure each individual's contribution to the communal discussion (very difficult elements to quantify).
Twidentify 11:14 AM on January 12, 2009
Glad to hear that engagement is going to become a relevant factor. What, exactly, is going to count as engagement? @replies? retweets?
Also, how much focus is being put on changes that benefit users looking for small communities, rather than the Elite crowd?
At twidentify.com, we chose retweets as the leading measure of influence, as it works best within the context of specific searches. For instance, if a user wants to find other user experience professionals, it's a topic that has a small community but retweets still stand out as being relevant, in or out of the UX community.
Would be good to see grader focus more on user goals and small communities too. It's something we'd like to include in Twidentify.
Cheers!
Dizzledorf 3:25 PM on January 12, 2009
It's its.
Beth 3:26 PM on January 12, 2009
This is all very interesting. I really admire Twitter, so simple, yet so effective. Not sure what my grade is but Twitter is definitely a great way to build online branding.
Online Marketer 9:52 AM on January 13, 2009
Interesting, although I think this will eventually be disregarded or gamed the same as Alexia, Digg votes, and any other social media ranking is. All the SEO freaks figure this stuff out pretty quick and start to game it so that the rankings are always a little biased.