A market for social-media data
Sipping from the fire hose
Making sense of a torrent of tweets
Oct 1st 2011 | from the print edition
MOST tweets are inane, but a million may contain valuable information. Fed through clever algorithms, a torrent of microblogs can reveal changes in a nation’s mood. Hence the excitement about a new market: the sale and analysis of real-time social-media data. MediaSift, a British start-up, will soon launch a marketplace for such information.
Analysing social media used to be a cottage industry. Firms gathered data slowly and patchily, through mechanisms not built for the purpose. Many online services kept their data locked up, because there was no way to make money from them. All this is changing.
Twitter was the first to move because it generates ever more data: the number of tweets per day now exceeds 230m, up more than 100% from the beginning of the year. Twitter would like to turn its popularity into money, but rather than beefing up its own infrastructure, it plans to outsource the task of distributing and selling its data to MediaSift and Gnip, another start-up.
Both MediaSift and Gnip are striving to be “data platforms”. They collect and standardise information from all kinds of social-media services—not only Twitter, but also Facebook, YouTube and others. Both Gnip and MediaSift have built robust networks which can cope with massive amounts of data in real time. And both are enforcing licensing rules: for instance, that a stream of tweets can be analysed but not republished.
Gnip, based in Boulder, Colorado, is more of a wholesale distributor. It charges $30,000 a month for a feed of half of all tweets. Customers can also subscribe to feeds of tweets containing web links or certain keywords. Buyers are mostly social-media monitoring companies, which analyse the data for a fee. Sysomos, a Canadian firm, for example, allows firms to track in real time what people think about certain products.
MediaSift serves both big corporations and individuals. Customers can define sophisticated filters, for instance to find all tweets by men who are interested in a new product and live in London. Charges for DataSift, as the service is called, depend on the filter’s complexity and the amount of data delivered.
The streams from Gnip and MediaSift can be combined with data from more specialised firms that try to extract meaning from social-media data. Lexalytics, for instance, analyses the sentiment of messages and posts. Klout measures the influence of social-media users (some firms give people with a high Klout score preferential treatment).
Having a marketplace such as DataSift has already encouraged other social-media services to open their data vaults, says Nick Halstead, the founder of MediaSift. Financial firms have become interested in feeding such data into the algorithms they use to make investment decisions, says Chris Moody, Gnip’s president. And corporations are increasingly keen on combining social-media data with customer information.
Yet growth in this market could be held back—by privacy concerns. Most people think that tweets are only up to 140 characters long. But those who sip from Twitter’s fire hose can get much more information, including a sender’s location, the biography on his profile page and how many people have subscribed to his messages. Most of this information is freely available on Twitter’s website. But if users realise how their data are used, they may clam up.
from the print edition | Business
