Combining different data sources to maximize Telco value

In this short demo video Incelligent presents how heterogeneous big data processed by its advanced machine learning algorithms can make huge difference for Telco decision support and thus enable the materialization of a tremendous competitive advantage. 

This case combines a diverse set of data sources: network service availability & performance metrics, customer retention & churn logs, revenue & profitability data, and marketing actions log (own and competitors’).

The use case assumes a Telco aiming at maximizing company value, modelled here via a complex index that combines revenue, profitability in short and long run, and of course customer experience and retention.

At the beginning we present tools for the navigation through big data that enable the identification of problems or opportunities by arbitrarily correlating data sources and parameters, as for example network unavailability indices with time dimensions and customer churn. Later we analyze the impact of network service quality metrics (data throughput and network unavailability) with customer acquisition rate, churn and profitability. A 50-day weighted average reveals a quality degradation that may be related to commercial and financial efficiency drops.

Aiming to analyze if there is an actual or circumstantial relation between these issues, we proceed testing the correlation of key relevant metrics with two methods, one linear (Pearson) and one non – linear (MutualInformation). The linear does not reveal something remarkable but the non – linear clearly justifies the relation between quality and profitability, quality and churn, quality and acquisition potentials. In this case it illustrates a relation that is shifted in time but well justified.

Following, we proceed to a predictive analysis of potential actions for the next month, in our case January 2016, assuming that we are at December 2015. Using a complex mid-term customer value index we elaborate on different combinations of marketing actions from within our telecom operator and its competitors, as also actions in the area of network management in relation to the marketing actions. In this scenario we experience a dominance of the choice to go for an offer of a high number of call minutes, combined with service overprovisioning, i.e. the green areas of the game theory – like matrix.

We continue elaborating on the predictions for the 4 key indices (churn, acquisition, profitability and network quality problems) assuming different tactics, starting from the “no action” or business as usual mode, comparing it with our base lines, i.e. business targets. Right after we receive a notification about a prediction report delivery, that reveals potential reactions of the competition assuming our selected policy. We continue elaborating on different potential policies we can follow, what will be their impact and how the competitors are predicted to match our actions. We navigation through the evolution of the key selected indices based on each scenario till we select our strategy with the justification of its detailed impact.

This simple scenario presents how our machine – learning powered technology changes the rules of the game in the telco business enabling advanced and proactive decision support, at short, mid and long term levels.