Why Isn’t #demo2010 Trending On Twitter?
November 24, 2010 in Students
Today’s massive student protests in Britain have produced a huge amount of Twitter traffic, much of it using the #demo2010 hashtag. But although a number of tags and phrases connected with the demonstrations have trended in the UK over the course of the day, #demo2010 has been strangely — and entirely — absent from the country’s trending topics list.
Why that’s the case still isn’t clear, but a comparison of the relative frequency of tweets using that tag with phrases that have trended in the UK today reveals that it’s not because of lack of traffic.
(click on graph to see it full size)
As of right now — nearly 3 o’clock in the afteroon NYC time — the top four trending topics on UK Twitter are #uksnow, #dayx, #wonderlandwednesday, and McKeith. But as this chart shows, #demo2010 has more than double the traffic of any of them at this hour, and has been far ahead of them all for the entire day. At around noon London time today #demo2010 peaked at 0.2% of total Twitter traffic, according to this chart — more than triple the volume of any of the current top four.
So what’s going on? I’m going to continue to dig, and others are on the case as well. More soon.
Update | Via Twitter, in response to this post, comes the suggestion that #demo2010 can’t trend because there’s a Twitter user with @demo2010 as their handle, and Twitter doesn’t allow strings that are identical to usernames to trend.
Second Update | The above explanation is incorrect. There’s a @dayx account, has been since 2008, and #dayx has been trending all day.
Third Update | Twitter honcho Del Harvey (@delbius) says, in response to a tweeted query, “we aren’t blocking” #demo2010. In followup tweets she guesses that the tag’s failure to trend may be an artifact of Twitter’s algorithms — “The more something is talked about, the more it needs to be talked about in order to be a trend” — but offers no inside info as to what’s going on.
If the trending algorithms are keeping #demo2010 off the list — and if the stats I’ve seen for #demo2010 tweets are accurate — then I must say that the algorithms strike me as seriously flawed. Take a look at what happens when you expand the graph linked above to cover the last thirty days, instead of the last 24 hours…
What this shows us is that #demo2010 spiked high on November 10, and about half as high — but still really high — today. Both spikes were dramatic, and both reflected huge interest in the topic. More importantly, both spikes were spikes — today’s pop in the numbers reflects a rebound from something close to zero between November 13 and November 22.
Compare that to “McKeith,” which did trend today. That term had two huge spikes in the last couple of weeks, and fairly consistent traffic in between and after. It trended even though its traffic today was actually quite similar to the traffic its been seeing over the last ten days.
I’m not taking a conspiratorial view of this, but I do think it’s weird, and it does strike me as bad coding — if in fact coding is the problem.
Fourth Update | Tweeter @adrianshort argues that the “diversity of tweets for TV programmes” versus the “huge numbers of RTs” for #demo2010 may explain the latter’s failure to trend. (Gillian McKeith is currently appearing on “I’m a Celebrity … Get Me Out of Here!” in the UK.) And though there have been a lot of unique #demo2010 tweets today, there have also been a lot of RTs, so this explanation strikes me as the most plausible I’ve heard so far.
In a separate tweet, Short suggests that trending topics may not actually be such a huge deal for an event like today’s demonstrations, since “You get massive network effects from normal RTing a popular tag.” I’m inclined to agree with that, too.
17 comments
November 24, 2010 at 4:24 pm
Jesse
#dayx is no longer trending…
November 24, 2010 at 4:39 pm
Andrius
So what’s next? student will smash a twitter’s head quarter with a psoters f**k twitter ?
November 24, 2010 at 5:08 pm
Adrian Short
@delbius has denied that Twitter are actively blocking the #demo2010 hashtag:
Twitter recently changed their algorithm to be more sophisticated than to just reflect the total amount of traffic around a tag. One hypothesis is that #demo2010 has been demoted because it was re-used from the demonstrations two weeks ago. This may explain
mashable.com/2010/05/14/twitter-improves-trending-topic-algorithm-bye-bye-bieber/
Twitter’s #tt algorithms are private so I doubt that we will get a definitive explanation of what’s going on.
November 24, 2010 at 5:46 pm
Mike Butcher
Twitter says they are not blocking it. More likely it’s just too generic a hashtag and/or has been used before. I daresay there are one or two other 2010 demos going on today in other parts of the world. It may be they prefer them to be relevant to location.
November 24, 2010 at 6:02 pm
Darryl
This is surely less a matter of algorithms, more that Twitter needs some human attention on events that aren’t happening in the US.
November 24, 2010 at 6:04 pm
Adrian Short
Trending topics are localised by city/country, so that isn’t it.
Obviously, global trending topics are going to favour the places with the most (English-speaking) users.
November 24, 2010 at 6:06 pm
SomeRandomNerd
Questions I’d ask;
-How many people (not posts, but posters) were using the hashtag?
-How many of the posts were original, rather than just retweets of the same things? (I recall reading a while ago that Twitter looks at “old fashioned” RT’s in a negative light.)
-Were people tweeting the hashtag a “closed circle” (ie. how many of them followed one another, and how much did discussion break through to a wider audience.) How many of them have Twitter accounts that state they are in the UK?
-Looking at TwitterStreamGraph, the word “trending” was clearly associated with the hashtag (probably because of questions about whether Twitter was suppressing it.) Does this have a negative effect on whether it becomes “trending”?
I don’t know if any of these factors are relevant, but they are the sort of things I’d consider if I were trying to write a “trending topics” algorithm.
November 24, 2010 at 6:11 pm
Bill Miller
I imagine that it would be possible for some tedious tags to hog the list. So they have algorithms that try and find the interesting ones and also to cut out attempts to manipulate the list.
November 24, 2010 at 7:21 pm
Nichol Brummer
There was the same type of discussion last spring about the iranelection hashtag(s). Twitter’s idea seemed to be that they wanted to catch special things, and not those things that are just the standard stuff that excites tweeps around the world all the time, ‘bieber’, for example.
I guess it is not easy to design such an algorithm. And if they say what their algorithm does, some people will try to game it.
November 25, 2010 at 5:17 pm
Sam Neylon
Hi Angus! Did you see this!
www.latimes.com/news/nationworld/wire/sns-ap-eu-italy-student-protests,0,1059534.story
November 25, 2010 at 8:05 pm
Darryl
#ashes doesn’t seem to have trended the past couple of nights, which would bear a lot of this out.
November 28, 2010 at 3:51 pm
The Technology newsbucket: Ireland’s bailout, Oracle’s win, Nokia marketing and more « billytompson
[...] Why Isn’t #demo2010 Trending On Twitter? « Student Activism“At around noon London time today #demo2010 peaked at 0.2% of total Twitter traffic, according to this chart ? more than triple the volume of any of the current top four.” If you can answer this question (and it’s not enough to say “@demo2010 is a username and you can’t have a Twitter trend that’s a username because @dayx is a username but #dayx trended). (Twitter says it’s not blocking it.) [...]
November 29, 2010 at 7:04 pm
La revue du lundi par We Are Social #38 | we are social
[...] #trendingtopics en question Student Activism s’est interrogé pour sa part sur la pertinence de l’algorithme de Twitter pour…. Le site a en effet constaté que les récentes manifestations en Grande Bretagne pour protester [...]
November 30, 2010 at 5:09 pm
Project: Carousel! » How to organise a successful student protest using online tools
[...] student protest on 24 October, the hashtag #demo2010 was used. However, a rumour went around that Twitter was intentionally stopping the hashtag from ‘trending’, so many people started using #dayX during the second day of protests instead. For today’s [...]
December 9, 2010 at 4:13 pm
London on lockdown: How social media spurred student protests | ZDNet
[...] explained how. “Kent Union mobilised 500 students at the University of Kent to attend the #demo2010 on 10th November that was partly undertaken with face-to-face meetings with students, but mostly [...]
December 10, 2010 at 3:05 am
London on lockdown: How social media spurred student protests Health News Now - It's All About Being Informed! - Health News Now
[...] Union, explained how. “Kent Union mobilised 500 students at the University of Kent to attend the #demo2010 on 10th November that was partly undertaken with face-to-face meetings with students, but mostly [...]
December 13, 2010 at 12:00 pm
Twitter, WikiLeaks e la puzza di censura
[...] contro i tagli all’istruzione nel Regno Unito. Gli studenti del University College London su Student Activism hanno mostrato le loro perplessità rispetto all’incongruenza dei dati (sempre su [...]