The first step in designing the methodology was to specify the problem: what are differentiating and discriminating factors that make people leaders? How can we identify these factors through an analysis of LinkedIn data?
A dedicated app embedding a LinkedIn Connect was then built, allowing user data collection and storage. Using a large spectrum of the available data (skills, ego graph, past experiences, educational background, etc.) our algorithm computed a series of 4 indicators, that were then used as building blocks for the Global Leaders Index.:
Each of these indicators evaluates a specific facet of leadership evaluation. We leveraged various social networks data analysis metrics to calculate each one of those. We particularly took advantage of:
- The number and nature of relationships (internal friends / external friends, i.e. inside or outside the current company) happened to be a real discriminating feature;
- Proxys of "betweenness centrality", a metric allowing us to evaluate if users where central for communication in the graph of relations;
- Variability inside a same function, a same company, or inside a particular sector. This follows the idea that, for instance, accountants with lots of connections might have better capacity to create relationships than salesmen with the same amount of connections.
The output of this work on granular LinkedIn data was the GLI, that helped hundreds of IMD alumni calculate their leadership capabilities, and identify which skills to improve to accelerate their careers.