A recent study posted in AdAge would have you think so. The study, conducted by an agency called 360i (direct link to the data here) examined 1800 tweets and determined that only 12% of updates posted to Twitter mentioned a brand – and for most of those, the brand was Twitter itself. The implicit conclusion: no one is talking about your brand on Twitter, so brands can direct their attention elsewhere.
I’m certainly not going to argue that Twitter is anywhere near as significant as other channels, and it is true that many marketers use Twitter like a big “easy button” for the Internet. Pushing it tends to give you immediate feedback, which is gratifying, but also tends to summon other marketers.
I’m not willing to go completely down the road of irrelevance based upon this study, however, for one important reason. The AdAge article states that, “after spending six months going over a statistically significant sample of 1,800 tweets,” the researchers were struck by how mundane most were. It is true, by the way, that 1,800 tweets would give you a statistically significant sample – I know, given the metric tonnage of tweets the service throws off every hour that it seems like a small number, but (cue Sly Stallone voice) Sampling Is Da Law.
The problem with the sample isn’t the fact that the N=1800. 1800 respondents would be fine if you sampled them all at once. If I drank from the firehose right now and just grabbed 2000 truly random tweets, you could actually draw some valid conclusions with a 2-3% margin of error on the total. However, that isn’t what this study did – instead, they looked at 1800 tweets over a six month time period. If we assume that they were spaced out regularly, that’s 300 a month, or roughly 10 per day. Even if we accept one month as a valid sampling time frame, what this study gives you isn’t one 1800-person sample, it’s six 300-person samples. Bolloxing them all together gives you a big number, but introduces a longitudinal bias into the data that’s pretty much a non-starter as far as I’m concerned. We weren’t talking about BP six months ago, for instance. And we aren’t really talking about Toyota now.
That’s about as far as I’ll go here to disparage the data, but it does beg the question – why so few? Harvesting tweets is, as far as I know, essentially free and relatively painless. Why not analyze 50,000 of tomorrow’s tweets and see if you get the same result? Yes, it’s work. That’s kinda what makes it worth doing.
Buffer
