The online tool SNALizzner monitored and detected the social network peak of the motorcycle event (18/08).
The SNA solutions of Social-3 allow Telepizza to identify the real value of its fans in social networks.
Social-3 analyzed the chat behaviour of the mobile community of the soccer world App 'MyMadrid'.
More details here can be found here: SNA focus to MyMadrid - Real Madrid official Mobile Community for its Fans
JBC's active fans form a community and influence with a positive buying behavior.
Contact us at email@example.com or via twitter sna_social3 for the full customer case!
Several centrality measures identify influencers, the best known measure being eigenvector centrality. However, people in a social network will be only be influencers when several conditions are met: (1) They should have a large network of friends and (2) they should be known in the community to give useful advice, e.g. having a high authority or reputation.
To identify a large network, measures such as degree centrality are the most suited. This indicates that a person reaches out to many other people. Depending on the type of business question to answer, only out-degree centrality might be a better measure.
To identify persons in the network with a high reputation, several measures can be used: eigenvector centrality, pagerank or senderrank. Pagerank measures the authority of a person, because it mainly uses incoming links to calculate the measure, while senderrank uses outgoing calls to calculate the measure. Eigenvector centrality uses as well incoming as outgoing calls and also might include the edge strength. Therefore, eigenvector centrality is the best candidate to identify the most reputable persons in the network.
Several models are now available to predict churn using social network information extracted from call data records. However, before using those models, the first step is to identify the level of churn due to the friends in your network from historical data.
All models based on churn in social networks are based on the premise that the more churners you have in your network neighbourhood, the more likely you will also churn at a later date. Therefore a first step will be to check if the churners are found in the neighbourhood and if the date of churn is later than their own churn date. Enough instances need to be present in your data before next steps, such as predicting future churn, are undertaken.
If a large proportion of the churners have churners with an older churn date in their network neighbourhood further analysis is warrant to identify the level and probability of churn.
Several publications (Dasgupta et al, 2008, Kawale et al., 2009, Karnstedt et al, 2010) create a graph that shows the probability of churn with increasing neighbouring churners. The probability is calculated by dividing the number of all current churners that have k churned neighbours by the total number of subscribers that have k churned neighbours.
If this relation still holds, the obvious next step is to use a predictive the model that takes the above observations into account.
Figure 7 from Churn in Social Networks: A Discussion Boards Case Study, M. Karnstedt et al. , IEEE International Conference on Social Computing (2010). DOI 10.1109/SocialCom.2010.40