Real-time Customer Insights, Segmentation & Social Network Analysis
CORTEX aiNSIGHT is an end-to-end integrated AI-based solution that provides real-time customer insights, segmentation and social network analysis. The focus is on developing derived customer insights to contribute towards a 360-degree view of the customer where structured and unstructured data is mined, natural groupings is detected in the customer data via unsupervised clustering techniques as part of a segmentation analysis, and customer interactions and community structures are analyzed utilizing information about connections between individual customers. These valuable insights are also used as derived features in other customer-centric AI-based applications such as risk scoring, affordability scoring, fraud and anomaly detection, response prediction, recommendation engines, and churn prediction. So, CORTEX aiNSIGHT is typically also used in combination with some of the other AI solutions within the CORTEX AI Library. In many applications models developed on segmented data sets allows for more accurate results than ones that are generalized over a single non-segmented customer data set. For example, segmentation use case for credit risk where a segmented scorecard outperforms a single scorecard.
Single Versus Segmented Scorecards
Segmentation analysis is dependent on the types of data available and is typically focused on a combination of demographic, geographic and/or behavioural segmentation. In turn this can feed into more specific types of segmentation analysis such as market segmentation or risk segmentation where aspects such as stability, willingness and affordability are relevant. In financial applications, demographic segmentation typically utilizes Know Your Customer (KYC) data and can be further enriched for telecommunication customers by adding event level data that allows for richer geographical information through the knowledge of tower usage. Behavioural segmentation can for example be used to identify groups differentiated by value, volume and product usage. For telecommunication customers, transaction data enriches behavioural segmentation substantially by enabling the calculation of additional features which speak to the particular way in which the customers use their phones, wallets and related services. Some example features include proportion of data usage in business hours, proportion of SMS messages sent to core contacts versus non-core contacts, proportional overlap between calling partners weekdays versus weekends, variation in airtime balance as at time of Top-up, etc. If these types of features are tracked over time, it enables the development of perspectives on customer lifecycle and maturity, and associated opportunities at various points in this lifecycle. Related to this is the ability to identify different trajectories. For example, if a customer displays behavior type A this month, what behavior(s) can we expect with what probabilities next month and the month after? The type of analysis also allows the measurement of segment size at a portfolio level over time, and the identification of emerging submarkets.
Social Network Analysis
Social Network Analysis focuses on customer interactions and community structures which are analyzed utilizing information about connections between individual customers. As an example, changing phenomena within the market tend to propagate through the community structure. A particular behaviour might originate in one section of the community and spread to others. By understanding how the community is structured, one can anticipate how long it will take for each trend to reach different sub communities of customers. A prime example of this is the adoption of new products, as often this is driven by word of mouth referral. For telecommunication customer use cases, this type of analysis is dependent on event level data, as it is only at this resolution that one can see the individual connections. The transactional-level data allows for creating various graphical representations of customers. Some use cases for insights from social network analysis include detecting lending circles, syndicated schemes and fraud rings as well as profiling of individual customers based on their social interactions.