Beyond traditional clustering and predictive models lies social network analysis. It can help describe customers’ behaviors in new ways, but what exactly is it and how can businesses use it?
To find out more, I interviewed Carlos Andre Reis Pinheiro. He’s been working in social network analysis around the world for many years looking at distinct types of problems in a variety of industries. He recently developed a course, Social Network Analysis for Business Applications to share his knowledge and help businesses improve performance.
What is social network analysis?
Social network analysis is a set of tools and algorithms based on graph theory to reveal relationships between entities. Everything is linked somehow, and the network analysis approach helps us in understanding the correlations behind different problems and scenarios. The term social is employed once the network analysis is performed under datasets comprising people. How subscribers communicate to each other? How social media users are connected to each other? How authors and co-authors are related to each other? However we can apply network analysis to understand how bank accounts are linked, or countries and governments are connected, or claims are associated, or tax payers are related over time.
What types of businesses can benefit from social network analysis?
Theoretically, any type of business and industry. We can apply social network analysis to reduce churn or to increase sales in telecommunications. The viral effects when customers decide to leave or to buy are identified and used in a way to diminish the first and improve the second. We can also deploy network analysis to detect abuse in insurance and utilities, reveal fraud in banking, or suspicious activities in tax payers. We can use this approach in even sociology, biology, and even medicine.
How do you collect information to perform social network analysis?
Very often the transactional data is the main source for the social network analysis. In telecommunications, calls and text and multimedia messages are one important source to build networks. Even though, for optimization purposes, switches, usage and human motion may be used to create a graph of the physical network. In banking, transactions between bank accounts (or between banks and countries) can be used to build the network and therefore to compute the metrics that explain the relationships comprised on it. In insurance, all claims are considered in connecting the business events to all types of entities (suppliers, policy holders, drivers, witness, repairs, etc.). In utilities, consuming transactions can be used to build the network that connected supply and demand. Emails connecting people, likes, tweets, messages, all kind of relationship between two distinct entities may be considered to build the network and analyze it afterwards. Basically, there is no limit for business applications by using social network analysis.
What kind of insights can you gather by performing social network analysis?
When thinking about complex problems, I believe there is always some network behind them. All kinds of problems can in some way be explained upon the concept of the network science. Most of the business problems are handled by looking at transactions and individual attributes. Social network analysis can distinguish between relationships.
In telecommunications for instance, most of the analytical models take into account subscribers attributes to better understand the market and to predict business events such as churn, bad debt or purchase. Network analysis can understand the subscribers in the most important way, based on their relationships. Telecommunications companies provide a way for people to get connected. Social network analysis is therefore the proper approach to understand how they get connected over time and then use this knowledge to improve business performance. Same happens in banking, insurance, utilities and retail. Thinking in insurance for example, a person may occur in one claim as a policy holder and in another claim as a witness. Or a particular physician occurs in many suspicious claims. The same repair takes place in many exaggerated claims. Everything is linked. We have to understand how these connections affect our business.
You created a new Business Knowledge Series course, Social Network Analysis for Business Applications. Why did you create the course and who can benefit from taking it?
I truly believe that everything is linked somehow. And understanding all these connections, visualizing the network behind complex problems, it is a good approach to better understand the business scenarios and then to provide robust solutions. I have been working in social network analysis for many years, in many countries, looking at distinct types of problems, and in different industries. A course like this is a great opportunity to share the knowledge. Not just the instructor’s knowledge, but mostly the students’ knowledge. Each person comes into the course with a particular experience, in different business, industries, scenarios and purposes. This exchange during the course is unbelievable. We can learn a lot from each other, like a network. We can realize that problems in different industries sometimes are quite similar and solutions may be just adjusted for distinct scenarios. We evolve as a network by interchanging our knowledge during and after the course.
What’s your take on the future of social network analysis?
Social network analysis is part of a discipline which is growing solid and fast. Network science comprises methods from graph theory, mathematics, statistics, physics, data mining and information visualization. It might be used in many business problems, no matter the industry. Based on network science we can better understand social phenomena, political and economic partnerships, diplomacy, business problems, human mobility, international trade market, biological systems, spread and pandemic diseases, internet, communication and collaborative networks, among many others. The use of network analysis to solve business problems is straightforward and may reveal more than we can expect. Intrinsic knowledge behind social relations might disclosure the proper information to better understand a very specific problem and solve it onward.
