Strategic ReportStrategic ReportStrategic Report Financial Additional Overview Strategy Performance Governance Statements Information Advanced analytics In addition to replacing a product-centric model with the client-centric approach, we have actively invested our resources in revamping credit management in Retail Banking. Namely, in 2017, we added machine learning tools to traditional statistics used before, and already in 2018, advanced analytics methods, such as client wallet size estimation and a probability of default prediction with machine learning algorithms, became business as usual. This enhanced our capabilities in predictive modelling, through which we aim to further optimise our credit, collections and campaign management in Retail Banking. In 2018, we implemented new analytics tools to identify the best choices for the shadow limits and to predict customer churn. In order to gain a better insight into the behavioural propensities of our Retail Banking clients, we classified them on a loyalty scale based on the variables such as time span, frequency and monetary value of their engagement with Bank of Georgia. Client analysis through loyalty classification allows us to predict client relationships with the Bank and make specific steps to retain them and boost their satisfaction. In 2018, automated decisions accounted for 75% of total consumer loan sales in Retail Banking and 59% of the sales were driven by active campaign management. A share of automated decisions in Retail Business Banking loan sales increased from 11% in 2017 to 47% in 2018 and 29% of the sales were driven by campaign management. In 2018, the campaigns resulted in a total of c.19.5 million actions through phone calls, messages and Next Best Offer (NBO). The year 2018 was also marked by our first venture into Natural Language Processing (NLP). We tried our first high precision sentiment evaluation model using customer satisfaction data. In the future, we aim to use text analytics, an NLP technique, to categorise customer feedback into specific topics, analyse the feedback, extract critical trends from it and find the solutions. Going forward, we also plan to enhance our NBO and make our customer service even more personalised. By harnessing big data and machine learning capabilities, we aim to offer our clients relevant products and services that are logically linked to their needs and lifestyle. We plan to build a recommendation system that self-develops and evolves around client characteristics and habits. Underwriting 37% 62% 100k <5min Pre-approvals in Fully automated Applications Decision time total credits sold decisions processed per for 85% of month applications processed Targeted campaigns through different channels 724 14mln 878k Campaigns Offers Sold products up>5 times y-o-y up 75.0% y-o-y up>3times y-o-y Annual Report 2018Bank of Georgia Group PLC 27