Twitter is rolling out a new feature to help you discover new people to follow:
The algorithms in this feature, built by our user relevance team, suggest people you don’t currently follow that you may find interesting. The suggestions are based on several factors, including people you follow and the people they follow.
The problem? Twitter badly needs a Matt Cutts.
Here are stats on number of tweets by Twitter Users by RJMetrics from Jan 2010:
- 80% of all Twitter users have tweeted fewer than 10 times.
That means only 20% are active users.
The 2009 Annual Report from Barracuda Labs independently confirms these findings.
- 34% of Twitter users have no tweets
- 73% of users have less than 10 tweets
Now, from the remaining 20% of “active Twitter users”, how many users are spam?
According to TwitSweeper in March 2010: 5%.
These are accounts who tweet "make money fast online!", "multiple sources of passive income", "view my naked pics!", etc.
That leaves 15% of Twitter users who are real and may be considered worth following.
Why This Is A Problem
If Twitter is trying to build a meaningful, relevant social graph, they have to clean up first.
Twitter’s PeopleRank faces the same challenge as Google’s PageRank: Blackhat SEO. These spam accounts are followed by each other and by other fake accounts – all to provide a semblance of a active social user graph and avoid algorithmic detection. These are virtually indistinguishable from real users and will become part of the suggested users ecosystem.
How many times do we encounter spam accounts on Facebook? How many times do we see spam results in the first page of Google search results? In contrast, how many times do we get @replies from spammers on Twitter?
A contaminated social graph or PeopleRank system is harmful to Twitter from an investor, user, and advertiser point of view. It will be great if Twitter is able to suggest whom to unfollow and get rid of all these inactive, fake, and spam accounts.