Blog

March 30, 2010

Twitter Authority and Relevance

You may have noticed that Google has begun to display tweets and other real-time communications that are related to search queries. But how does Google ensure that the results its displays are quality tweets from quality Twitter users? Well, only Google knows for certain, but it is likely very similar to their model for ranking web sites, PageRank; though there are likely some disparities because of the inherent differences between the forms of communication.

At the most basic level, PageRank uses two methods to determine the quality of a page for a particular search query: (1) the quality and number of sites that link to a specific page and/or domain and (2) the relevance of the search query to the page, e.g. the number of times the search query is used on the page.  Moreover, Google’s Twitter model is likely very similar.

Authority

Regarding the first method, determining the authority of the Twitter user, there are likely a myriad of factors, but a few stand out as being likely dispositive: (a) the number of followers, (b) the quality of the followers, and the (c) ratio of followers to following to tweets.

(a) Google likely looks to the number of followers to determine whether the Twitter user should be given authority, e.g. a user with 100 users will generally be deemed as having far less authority than a user with 10,000 followers.

(b) However, just like PageRank the authority of each of those followers almost certainly has a large effect as well. For example, if Twitter User X is being followed by Twitter User Z, and Twitter User Z is being followed by 2,000,000 other followers, Twitter User X will likely being deemed to have substantial authority. This is very similar to PageRank in that a site that is linked to by another site that in-turn has a large number of sites linking to it, will impart substantial PageRank to site. For example, if Site A is linked to only by Site B, but Site B is linked to by 10,000,000 sites, Site A will generally be deemed an authority on the subject for which it is being linked to by Site B. The biggest likely difference between the Twitter algorithm and PageRank is that Twitter does not provide any context that Google can use to determine the subject for which one Twitter user is following another.

(c) The last factor is more speculative, but seems entirely plausible & likely: the ratio of followers to following to tweets. There is tacit understanding amongst Twitter users that if User B follows User A, User A will reciprocally follow User B. As such there are many users that have only tweeted a handful of times but have thousands of followers. Further, the quality of those handful of tweets are, at the very least, quite low. As such Google doesn’t likely regard these Twitter users as imparting much authority despite having a large number of followers.

Relevance

In regard to the second method, determining the relevance of the search query to the tweet, there are two factors that are determinative: (d) the incorporation of hashtags and (e) the use of signals.

(d) One of the conventions of Twitter is the use of hashtags to denote the subject of a Tweet. Unfortunately, this convention has been exploited by spammers and can no longer be reliably used as an indicator of the subject of the tweet. In response to this exploitation, Google has indicated that it generally gives less credence to tweets that incorporate hashtags. However, any penalty for the use of hashtags is likely mitigated, at least partially, by the authority of the Twitter user.

(e) Google likely monitors sites with high PageRank that are updated regularly to determine whether Twitter search results should be incorporated into the regular search results page, the use of these site is typically called monitoring signals. For example, if 10 of the largest news sites post updates on their sites at the same time and regarding the same subject, Google will possibly include real-time search results from Twitter, for at least a short period of time.

When developing and implementing a Twitter strategy it is important to take all of the above factors into consideration in order to maximize the effectiveness of the strategy.