Making Data Science work for improving customer service Qorus Banking Innovation Awards 2022

Traditionally, application of analytics has always been used for product cross-sell or value enhancement. This is a ‘never before’ approach of applying analytics for providing best-in-class customer service with no narrow boundaries in terms of leads, conversions and lifts. We use data signals to capture service related needs from customer transactions. We additionally apply multiple analytical approaches to identify the relevant service needs for each customer. We have created a one-view of all service related needs applicable to every customer of the Bank covering Individuals, Businesses, Proprietorships, MSMEs (Micro, Small and Medium Enterprises) and NRIs (Non-Resident Indians). This encompasses 50+ service triggers broadly spanning across 5 domains as follows – • Nominating Beneficiaries: Going a step further, we not only want to safeguard the interest of our customers but also ensure that their families are financially protected in case of any unfortunate event. We guide customers to declare a nominee in all their accounts and investments so that their loved ones can inherit the funds hassle-free and stay financially protected. • Unclaimed Funds: There are huge sums of unclaimed deposits lying in dormant bank accounts, some even for more than a period of 10 years. We try to reach out to the customer or the family members to activate the account to claim or invest the proceeds. • Compliance/Regulation: To ensure compliance with local rules and regulations, it is mandatory to submit certain documents like proof of identity and address proof at regular intervals of time. We reach out to customers who have not submitted their documents. This pre-empts them from any unintended discontinuation of Banking services. • Monetary Benefit: Through the monetary benefit section, we guide our customers with simpler digital options to earn maximum rewards and also to redeem unclaimed rewards before their expiry. • Tax Saving: While customers try to save tax by investing in certain financial products, many customers miss out on other modes of tax saving may be due to lack of awareness of these options. For instance, NRI customers can avoid paying double tax by submitting NRI Double Tax Avoidance Agreement (DTAA). Nearly a million customers are reached out to every month with Service based enablers. Innovative use of technology: • Statistical computational techniques: Application of specific statistical computational techniques for determining certain value thresholds. For example, setting a lower or upper limit of the customer’s account balance for reaching out to ‘To be Dormant’ accountholders. • Algorithms: We use methods based on Deep learning comprising Natural Language Processing (NLP) and Algorithms, such as – word2vec, fuzzy logic, phonetic matching, levenshtein and jaccard distance. For example, to identify customer address mismatch as many customers relocated from cities to their hometowns due to the pandemic. • Rank-ordering mechanism: Implemented topological sorting, a rank-ordering algorithm that grades the services available for every customer in order of their importance. This was primarily effectuated to empower the frontline channels in their conversations with customers by speaking about the most impactful service element first. • Prescriptive Analytics: We interweaved a communications framework for the frontline channels in order to guide them on how to speak to their customers for availing the services. A KEA (Knowledge, Enabler, Action) framework was devised to provide context to the services cited. For example, i. Knowledge answers what is the service, which is further classified under a Category and Sub-category. ii. Enabler talks about the background of the service highlighting expiry dates or regulatory deadlines, if any. iii. Action refers to the process for availing the respective service focusing on the seamless and convenient use of digital mediums to avail the respective service. For example, to help customers save tax especially Advanced age customers (>=60 years), the Bank provides an Application Form that can be submitted to avoid Tax Deduction at Source on interest income earned. This is applicable if their total income is below the taxable limit. A lot of customers don’t submit the form mainly due to lack of awareness. This initiative helps customers save significantly in taxes. We identify customers who are eligible to submit Form 15H basis their age, FD amount, tenure, interest earned, annual income and thereby the income tax slab they fall under. Knowledge: Segment: Tax saving Category: Due for Sub-category: Form 15H Enabler: ‘Customer had submitted Form 15H last FY to save tax. Customer has not submitted Form 15H this FY.’ Action: ‘Customer is a Senior Citizen and has invested in Fixed Deposits with the Bank. Guide customer to submit Form 15H to avoid Tax Deducted at Source (TDS) on Fixed Deposit, if applicable. Digital Path: Online NetBanking>>Accounts>>Request>>Form 15H or customer can visit the nearest branch.’ By capturing the required information, Relationship Managers are well-equipped with the right pitch pertaining to a service before they engage with the customers. We also collaborate with the Marketing team to send customised e-mails and sms campaigns to reach out to the customers through the digital route. Technology stacks: The solution was supported with a best-in-class tech stack that consisted of ETL for data processing and model development  -Open-source Python: Because of its low difficulty/functionality ratio we are using it for the below activities a. To create Customised ETL scripts to process the Data efficiently between different layers. b. To create customised scripts to automate the dynamic quality checks for all the ETL pipelines to ensure the best data quality throughout the data lineage process. c. To create algorithms for Descriptive analysis, Predictive analysis, and Prescriptive analysis which is used to drive decision making  -Jupyter (32 core/512 GB RAM): A web based interactive computing platform hosted on our Linux machines within our premises. It is used for Python script development and Testing activities.  -Linux: Operating system where most of our solutions are hosted currently. We are leveraging the Linux Cron statements to schedule the execution of the commands at the desired interval.  -Airflow: Open-source workflow management platform for data engineering pipelines. It helps to set the dependencies between predecessors and successors of the ETL pipelines.  -Postgres (16 core 80 TB DB server): It is our DataLake layer of around 80 TB capacity. Acting as the centralised Data repository for all the analytics needed within our department. And currently both OLAP (Online Analytical Processing) and OLTP (Online Transactional processing) are taking place in PG.