You would be hard pressed to find a person aged 15 to 35 who has not heard of social media sites like Facebook, Twitter, MySpace and so forth. Facebook alone counts more than 550 million users as of mid-2010 and the number is still growing. Put into context, these 550 million users, if regarded as a country, would be the third largest on the planet – after China and India, and just ahead of America.
How to Mine Social Media for Business Success
Clearly, the latent potential of social networking in business, government and society requires research. It was therefore apt that a group of academics and industry practitioners gathered to share their knowledge on social network mining at a workshop organised by Singapore Management University's School of Information Systems. There, participants shared their insights into the issues.
Professor Jeffrey Xu Yu of the Chinese University of Hong Kong discussed the applicability of graph theory to address problems of nodes on the internet being reachable from each other. Each of these queries, known as ‘Reachability Query’, tries to use classic shortest distance determination via search trees based on its location on the graph to find a pattern matching over the connected set of nodes in the network modelled as a graph.
In the context of mining the social network on the internet by modelling the connected nodes in it as a graph, the method tries to answer questions such as: which nodes are connected to which other nodes on a social network? How many hops does it take for one node to reach another?
Bangalore-based Vineet Chaoji, senior associate scientist at Yahoo! Lab in India, lent an industry perspective. Part of his job is to analyse and find patterns within Yahoo’s own network of users, many of whom already have informal social networks.
Chaoji presented a case study from Yahoo! which examined the use of social influence for targeting advertisements to a small network of friends as a means for getting users to sign up to a paid service of PC-to-phone calls.
The study was a means of getting traction to increase user adoption for a paid premium service, by observing how a network of influences could better target their constituencies of followers. “As you probably know well by now, social media has the potential of making internet-based marketing much more effective than it now is – which is [currently] more like a hit and miss affair,” said Chaoji.
Depending on the click-through rates and views of the relevant advertisement, total viewership is usually in the region of 0.1% to 0.001% – or even lower – of all site visits where there are advertisements. The theory is that people are more likely to click on an ad and respond to it when they know trusted friends are already using or endorsing that service or product.
Using data from Yahoo Messenger, Yahoo! mined the density of connections in a sample of users. Chaoji sought answers from a focused set of questions: Do individuals wield influence over their friends in online social networks? Are highly connected individuals the same as social influencers? How far reaching is the influence in the network? Is it local to its social neighbourhood or is it global? Which targeting methods are suitable?
What he found was that a small number of highly credible or trusted influencers could indeed have a disproportionate influence on their circle of friends – depending on the level of connectivity of the key ‘index users’, as measured by the number of people they connected to via their Messenger chat threads.
However the more important attribute was closeness within the network neighbourhood. In other words, highly ‘influential’ users, while effective in reference to their nearby connections, will experience an exponential drop in their level of influence beyond the first hop most of the time. “Overall, we are more likely to trust our friends and their interests and recommendations, especially if these also happen to be the same as ours,” Chaoji added.
How should advertisers reach these first-hop networks of friends? Another finding from the research was that the advertisements had to be non-intrusive and must blend into the sites that these people are likely to visit.
The study explored two complementary targeting/marketing approaches: direct and social neighbourhood. Direct marketing aims to target the users that are most likely to adopt a service, whereas social neighbourhood-based marketing targets the neighbourhood of likely adopters.
Social targeting was found to be more effective since it can influence a larger set of users through the homophily effect (that is, the tendency of individuals to associate and bond with others with similar interests and attributes).
Yahoo! had run pilots wherein advertising and promotions for a movie, for instance, were served via social networks. The advertisement contains hooks allowing users to forward the movie trailers to external social media, such as Facebook, Google Buzz, Plurk, Twitter, and so on. Campaigns using social networks shows some evidence of greater user interaction with the ad, and a larger audience reach.
Trust and relationships
During the workshop, Professor Lim Ee-Peng of SMU’s School of Information Systems delivered a paper on rules and behaviours within “trust mining,” discovering the drivers for rules regulating networks of nodes in social networks that have reciprocal levels of trust.
Starting from the premise that people care about trust enough to use it as a parameter for decision making, Lim mines the network of trusted relationships in websites like Epinions (www.epinions.com), which uses a network of reviewers on anything from food to electronics to cars. Reviewers rate products with stars (1 to 5) and write reviews about the products. Some reviewers eventually develop trust relationships with the website’s users.
“We tend to trust people we know and have a relationship with. Within the context of the anonymous web of reviews made by people we cannot see, we have to learn who to trust,” Lim noted.
According to his research, new trusted links between users are formed based on some existing trusted links. However, different sets of rules apply for the formation of distrust relationships.
Lim’s study showed, for example, that trust within social networks is transitive: in other words, if A trusts B and B trusts C, then A trusts C. But, transitivity does not work for distrust, i.e. if A distrusts B and B distrusts C, then A may not distrust C. A core reason for the transitive nature of trust in social networks is that it is propagated via each pair of trusted connections, he observed.
The power of transitivity also varies with the type of network. These relationships tend to work better in strong web trust links (for example, Web of Trust) than weak ones (for example, Twitter followers, almost all of whom have no real personal relationship with the opinion leader they are following).
Limits of online trust
To be sure, there are issues regarding the general robustness of trust mining as a way to discover rules within social media. Frequent users of sites like eBay, for example, suspect that the system for rating the sellers can be unreliable.
In a trust-based system like eBay, sellers are rated on their reliability by buyers who expressed opinions on the match between what was advertised and what was delivered. In such cases, since the online identities of both buyer and seller are unknown, there is the possibility of a single seller impersonating many buyers, who recommending that seller as a trustworthy seller.
Hence, the fact that 20 people recommended “JohnTAN345” (for illustration) as a trustworthy seller – increasing the probability that someone will buy from him next -- may be meaningless. “JohnTAN345” could have easily sent in recommendations for himself under different aliases with glowing testimonials. As the popular adage goes, “on the internet, nobody knows you’re a dog.”
Nevertheless, whatever the application and area of interest, it was clear from the workshop that social media has insinuated itself into the collective consciousness of many users. Savvy corporate decision-makers should be looking at it as a means of outreach, whatever the direction social networks will take in the future. It is likely that it is companies that get the right mix of intent and subtlety to keep users engaged that will be more profitable and successful.
About the Author
[email protected] is an online resource that offers regularly updated business insights, information and research from a variety of sources, including interviews with industry leaders and Singapore Management University faculty. The resource can be accessed at http://knowledge.smu.edu.sg.
Read more on