The talk about how the majority of the traffic on Twitter is pretty much junk is nothing new. Back in March 2009 Google CEO Eric Schmidt told the crowd at the Morgan Stanley Technology Conference that Twitter is a "poor man's e-mail system." David Letterman called it "a waste of time" (although he apparently only found out about it on his July 21, 2009 show, so let's give him some time). It even took a comparatively non-famous gnat like myself a while to warm up to Twitter: I signed up in July 2007 and tweeted barely 17 times between July and October that year. Then I got bored and forgot about it for 8 months. In all, it took nearly a year and half for me to come back and discover how useful Twitter can be to learn more about subjects that interest me, can help me do my job better and connect with like-minded professionals.

Despite Twitter's popularity, the perceived problem with Twitter being a waste of time has not really abated much in the media or on the web: in August 2009 Pear Analytics released an interesting report simply called "Twitter Study" that sampled 2,000 tweets and classified them into six categories: News, Spam, Self-promotion, Pointless Babble, Conversational and Pass-Along Value. (I originally saw the post on the WebProNews blog) The Pear report found that over 40% of its sample was "Pointless Babble" ("Self-promotional" and "Spam" posts falling in at 5.85% and 3.75% respectively)

And almost from the beginning of my Twitter explorations it was very clear to me just how spammy Twitter could be. Sure, there were plenty of tweets in the public timeline trying to sell get rich quick schemes as well as the not surprising number of porn-related tweets. What I discovered early on was Twitter's potential for a more subtle form of spam: Follower Spam. Initially excited to get alerts of all kinds of new people following my tweets, I would look at my Follower list and be surprised to find nearly all were the same multi-level marketing, dating/porn and other unwanted accounts I had no interest in seeing. I suppose I could have just left them alone and hoped that over time these accounts would stop following, but at the same time I became concerned that others viewing my public list would see all these "questionable" accounts following and get the wrong impression of me (am I the only one who thinks like this?). I had to do something about it, so I devised a plan to clear these out on a regular basis ("Twitter-Blocked: It's Nothing Personal, Just Business" January 19, 2009), but at the same time being a metrics guy, I was curious to see how many of my followers would turn out to be spam if I just tracked it for a while.

So beginning in March 2009 I saved all of my new follower notification emails and classed them into 6 buckets:

1. Follower & Following: people who followed me and whose content I value enough to follow back — I don't as a rule follow back everyone


2. Following Only: people whose tweets I like enough to follow but who did not follow me back


3. No Follow: those I did not object to following me but who I could not find any common ground to justify a follow back


4. Follower Unfollowed: those who followed me but had unfollowed before I got a chance to review the request. I don't lose much sleep over these since I suspect many are using auto-follow/unfollow bot services like SocialToo.


5. Blocked: followers that once I reviewed them I found the content of their tweets, their profile, etc. to be objectionable. I often block those that don't have a profile photo or a URL n their profile. Overtly salesy tweets that predominate the timeline (vs. a good mix of informational vs. promotional) I also usually block.


6. Twitter Suspended: Twitter has suspended the account before I get a chance to review the follower.

After 6 months of capturing the alerts, representing 578 follow requests, I was surprised to find that my tracking of follower spam was not really that far off from the Pear study of individual tweet spam: 33.73% of the follower requests I received I ultimately blocked compared to 40.55% Pointless Babble tweets in the Pear study (see the pie chart "Follow Request Composition Mar-Aug 2009"). It's a bit of an Apples to Pears (!) comparison, but if you consider that ultimately the Pear study and my informal tracking are trying to assess value, we're pretty close. I see the Pear study "Pass along", "Conversational" and "News" categories as representing my categories of "Follower & Following" with "Following Only" thrown in — the kinds of tweets I usually find valuable.

Despite the challenges of maintaining a clean and vibrant follower list in an atmosphere of increasing noise, I tend to agree with Scott Bishop of Real Time Marketer Blog who said in a post "There is no Twitter spam, there is no such thing as a bad Tweet, there is no Twitter 'pointless babble.'  Why? Because Twitter is only and exactly how you make it." And I choose to make my own corner of the Twitterverse as interesting and valuable as I can.

If you enjoy talking about this kind of thing like I do, please follow me on Twitter from the sidebar. If you're a spambot, web-cam porn star, MLM schemer or "life coach," don't bother — I'll block you soon after :)

Sources:
Scott Bishop. "Most Twitter Statistics Are Worthless." Real Time Marketer Blog. August 19, 2009.
Ryan Kelly. "Twitter Study Reveals Interesting Results About Usage – 40% is 'Pointless Babble' " Pear Analytics. August 12th, 2009.
Mark Glaser. "How 'Follower Spam' Infiltrated Twitter - and How to Stop It." PBS MediaShift. October 23, 2008.

Technorati Tags: metrics, spam, social media, tips, twitter