Marketing on Twitter: Survey reveals most engaging (and most annoying) Tweets

Twitter has become a powerful marketing tool, but what techniques actually work when it comes to micro-blogging, and what tactics should be avoided? A study based on 43,000 responses to Tweets has uncovered what people like (and loathe) about Twitter posts.

The report, “Who Gives a Tweet: Evaluating Microblog Content Value” is the culmination of a year’s worth of analysis conducted by the researchers Paul André of Carnegie Mellon, Michael Bernstein of MIT, and Kurt Luther of Georgia Tech as they set to find out what factors contribute to effective Tweets.
Last year, the team created a site, Who Gives a Tweet that asked users to designate a selection of tweets according to the emotional responses they provoked (“positive,” “neutral,” “negative”).
The team then analysed the 43,000 crowd-sourced responses they’d collected from the site, looking for patterns and takeaways.
The survey indicated that tweets that are informative or funny evoke the best responses.
The results indicated that tweets that contain stale information, repeat conventional wisdom, offer uselessly de-contextual news do much to justify themselves.
In the study’s sample, respondents considered only 36 percent of the tweets on display worth reading — and another 39 percent barely worth the effort.
“These results,” the authors note, “highlight the need for better awareness and presentation of valued content.”

Being ‘boring’, repeating old news, cryptic, or using too many # and @ signs were cited as the main reasons for ignoring tweets and deciding to ‘unfollow’ users.
For each tweet, the researchers determined a content such as Me Now (current mood or activity), Presence Maintenance (e.g., “Hullo twitter!”), Self Promotion (e.g., sharing a blog post the author just published), and Information Sharing.
Using the study’s findings, Megan Garber of The Atlantic set about crafting the most annoying tweet possible, republished below for your enjoyment:

Read the full report here (PDF)