Malcolm Owen On August 18, 2009 at 12:13 am

From the department of news-gathering labelled only as “Did this really need to be researched?” comes: Twitter is used for pointlessness…

Twitter Fail Whale

Yes, someone with way too much time on their hands (Pear Analytics) has found out that out of a random sample of 2000 Tweets, that 40% of them are classifiable as “Pointless Babble”. The rest consists of conversation, spam and self promotion.

Although I would generally say the findings are probably accurate, I do feel that the 2000 sample Tweets checked are simply too few to confirm the findings. 2000 might be a lot, but there are many times more tweets sent on a daily basis, dwarfing the sample. Add in the fact that it’s only the public timeline tweets (ignoring private ones) captured every 30 minutes over a 2 week period, and it becomes really quite meagre. I would have thought that 20,000 or 100,000 Tweets would be a fairly decent amount to check from, but 2000? Seriously?

Pear Analytics have said that they aim to be doing this regularly. I hope next time they monitor more tweets, say, like ours?

3 Responses

  1. Ryan Kelly says:

    Malcom, 2000 tweets was a statistically significant number. We based it on roughly 3m tweets sent per day in the US (although there are no accurate estimates on this). If your into math, we’d be happy to show you how we arrived at that number :)

  2. Malcolm Owen says:

    I am intrigued as to how 2000 is a significant number for this, based on the potential 42 Million (ish) posts to choose from. Alas, I also realise that the more in the sample size, the more time is spent deciding what to class each posting as, and the more expensive the research turns out to be.
    Granted, there will still be people unhappy with the sample size even if it were 20,000 and so I can’t really complain too much about it. The answers probably still scale, though…

  3. moskito says:

    What can Twitter actually do for my business?