Social Targeting: Social Network Analysis in Telecommunications Industry Application
The Department of Network and Data Science welcomed András Kárpáti (Lynx) to deliver a presentation on March 5, 2019. He shared his perspective on the application of graph theory in general and social network in particular in business development. As a product company focusing on the use of big data graphs in dealing with complex business problems, Lynx proposes a better campaign targeting solution for generic products.
One of the ongoing projects is with a telecom company in Asia that attempts to increase postpaid subscription. Social targeting in this case provides more reasonable strategy in identifying who will be more likely to subscribe to a telecom service or to buy a telecom product by targeting communities rather than individuals. It implies that what friends buy matters in individual purchasing behavior as a customer is much more likely to buy if his friend does. The presence of peer-effect, therefore, is negligible.
To tap this rising opportunity, discovering community structure of social network remains intriguing. A notion of ‘strength of weak ties’ becomes the key concept in facing community detection in large graph such as telecom customer network with 18 million customers (nodes) and 220 million calls between them (edges). Community falls into two categories, namely, strongly connected component and neighborhood. Meanwhile, customer selection criteria could be classified into outlier in community and new adopter in community.
Taking these two features into account, the conversion rate from prepaid to postpaid per month reaches 3.96% percent in total. It is 7.8 times better than random at selecting people to change their subscription from prepaid to postpaid. Moreover, the best selection criterion is the combination between neighborhood and outlier in community as it gives 5.58% conversion rate with 1,098% uplift from 0.51% in baseline scenario (random selection).
In the end of the talk, Prof. János Kertész looks forward to the prospect of collaboration between Lynx and the Department of Network and Data Science. András believes that Lynx with its international network ranging from its European headquarters in Budapest and its Asian branch in Singapore has many opportunities to offer including career development.
Blogpost by Rafiazka Hilman