Tom Webster, writing and speaking

Twitter Encourages Data Butchery

Added on by Tom Webster.

One of the dangers of getting your information from Twitter is the fact that most survey research data is loaded with caveats and constraints, and 140 characters isn't enough to list them. So yesterday, when one of my Twitter pals posted this:

Sharing on Facebook is now more popular than sharing via email (via @mashable) http://bit.ly/dygLI

...I naturally had to follow up. It turns out that this finding was amongst users of a content sharing widget called AddToAny, and is based upon internal data from users of their sharing tool. That doesn't invalidate the finding, mind you, and when I followed up with AddToAny I learned that the same survey last year showed EMail in the lead, so within the sample frame a shift is a shift. You can't project it to the gen pop, of course, who are still doing lots of sharing via good old cut-n-paste, but as long as you know that, you know how to process the data from this particular report.

Trouble is, you don't have room on Twitter to post the sample constraints, only the headline. And though the original poster probably knew the difference (I am sure he did, in fact) his legions of re-tweeters probably don't. Eventually, the link gets left off of the retweet, and only the misleading headline (with a "credit" to the original tweeter) remains. 99% of the issues I see with survey research--particularly social media research--are in the reporting, not the survey. Twitter lets you post the headline, but the butchery is in the details.

My point here is not to be a cranky Datasnob--merely to encourage you to always click the links and follow through before you retweet. Fight innumeracy!