The risk of reading realtime Twitter buzz too closely

There is an interesting article on Techcrunch on this subject today, and it tallies with what our own realtime search engine tells us when we point it at Twitter in an unmediated way. Techcrunch:

Twitter’s near real-time search capabilities and the ability for them and third parties to mine the collective data from user messages for indicators of what’s buzzing online is the intrinsic core value of the company now that’s it has grown to the size it is now, at least for the time being.

We already know Twitter can be quite the source for breaking news, but critics have in the past correctly pointed out that one should be aware of the fact that the mob isn’t always right, and unverified claims on the micro-sharing service - often from a single user or even a single message - can quickly lead to false or incomplete stories circulating rapidly and viciously until the dust settles and the truth surfaces.

And no sooner does this become useful than it is gamed to uselessness. You can't go to a conference today without someone giving you a hashtag and asking you to Twt to the rafters about it. And as techCrunch notes, now digg like viral posse's are taking it to new lengths:

Today, when you look at Twitter’s trending topics, you’ll notice that the large majority of trends are memes started by a single user or a group of users, with the main goal offering entertainment rather than spreading information. That’s all fine and dandy - no harm in having fun - and I realize well that Twitter’s trending topics are not necessarily required to be giving you and me an overview of stuff that really matters, but I can’t help but think it’s a pity that that list is starting to turn into the top 10 of chain letters people used to circulate through e-mail messages in the late nineties.

There is something else as well, which we found when we pointed our own realtine search engine at twitter in unmediated mode (ie not looking for any one thing, just counting the occurrences of things). We found, for example, that those top trending topics are not actually the biggest topics on Twitter at the moment, they are just ones growing fastest. There are some topics that nare much more perennial, with bigger traffic, but their rise and fall (from a higher base) is far less spectacular.

For example, today Twitter says the top trending topics are:

# liesboystell
# liesgirlstell
# 3wordsaftersex
# BGT (Britain's Got talent)

But our system tells us that the following topics, for example are getting more twts:

#yay
#followfriday
#identica
#swineflu
#wikipedia (a lot of the twts are around the scientology banning)

Followfriday is a friday thing, so we suspect its been moderated out of the Trending area. Identica is a big traffic driver on any day you care to look at, as is Swineflu at the moment, and as for Yay - well its just big and random. Our system also tells us that Wikipedia is a bigger story right now than any sex, lies or talent videos.

Now it could also be that we are measuring over different timespans, or counting different things (metrics veterans will know how easy this is...), but what it shows is the Twitter Trend-O-Scope can be focussed in different ways, and will then give different answers.