An Event Driven Approach to Customer Relationship Management

An Event Driven Approach to Customer Relationship Management in e-Brokerage Industry Dickson K. W. CHIU Wesley C. W. Chan Gary K. W. Lam Franklin T. Luk Dept. of Computer Science & Engineering, Chinese University of Hong Kong Shing-Chi CHEUNG Dept. of Computer Science Hong Kong University of Science & Technology scc@cs. ust. hk kwchiu@acm. org, cwchan@yahoo. com, kwlam@cuhk. info, luk@cse. cuhk. edu. hk 1

Introduction and Motivation n CRM n n Most recent CRM work n n n Helps a firm to streamline customer services Centralize its customers data for analysis purposes Critical to the success of a business Concentrate on data mining Construction of customer behavior models This paper n n n Turns knowledge into business action Carry out appropriate actions effectively and efficiently Detailed system architectures Implementation methodologies for CRM activities enactment “Event-driven approach” 2

Event-driven Approach n Motivated by the active database paradigm n n n Event - occurrence of something interesting to the system itself or to user applications Event driven execution of rules in event-conditionaction (ECA) form ECA (active) rules: On event if condition then action Exceptions and alerts are events too (action = handler) Ensure efficiency and timeliness Business Events Client Events Brokerage’s Events Environment Order Placement Staff Turnover Market News Complaints Service Change Stock Price Change Service Exceptions Change of Personal Profile 3

Brokerage Industry in Hong Kong n n n Has high potentials in collecting valuable client information (HKSFC requirement) - brokerage must keep client transaction records for 5 years little room to increase its revenue through cross- or upsale trading Changes in industry n n n Endorsement of removal of minimum commission rule Extension of trading hours of stocks and futures Tightened requirements of liquid capital for margin trade loan – CRM for risk management 4

SME Brokerage in Hong Kong n n n n Most are SME firms (therefore our target of study) Competition from big firms and banks - drop of market share from 23. 01% to 19. 04%. Reduce operational cost Increase revenue Reduce client attrition rate Move to e-commerce Internet platforms Become more robust and cope with the changes retain existing clients and to increase their satisfaction through effective coordination and enactment of CRM activities 5

An Event Driven Approach to CRM ICQ / email / SMS Internet Alert Managerial Application Client Portal Sender CRM System Active Rule Engine market data Client Call Center Broker Front-end Environment Listener Data Warehouse Analytical Engine Back-end 6

Heart - Active Rule Engine n n Separate the active rule engine from the analytic engine Knowledge discovery in the analytic engine is resource intensive and computational expensive Triggering events from external sources - detected by the environment listener, the analytical engine or user input from the front-end subsystems Time events generated by the system clock helps tracking deadlines (e. g. , payment due dates) 7

Backend - Data Warehouse n n n Oracle 8. 1. 7 Database Server on a Windows 2000 Server platform For Online Analytic Processing (OLAP) Client transaction and holding records retrieved from OLTP system 8

Backend - Analytic Engine n Important changes or alerts are detected, these events will be forwarded to the active rule engine for processing n n n n Client Value Estimation - expected profit from a client (Domingos and Richardson) Client Attrition Alert - account balance, statistical behaviors Client Risk Analysis - related to margin trade Client Segmentation - transaction amount, frequency, recency, stock type, risk shouldering and other demographic data Client Channel Analysis - trend of client contact channel and specifically for web presence by studying clients’ click-through Marketing Campaign Analysis - studies the successfulness of a marketing campaign Key Performance Indicator (KPI) measurement 9

Backend - Environment Listener n n Input from different market and news data sources E. g. , Reuter’s services via Java Message Service (publishand-subscribe mechanism) Relevant data are stored into the data warehouse for later processing by the analytical engine or other retrieval purposes Relevant events are passed to the active rule engine for processing. n n n Evaluation of several variables and of several values per variable Related attributes can be evaluated similarly for speeding up (Cheung et al 2002) Need to take in account of the dependencies among the issues 10

Rules triggered by events related to client behavior Event Condition Action R 1 Trade input (from client portal / broker call center) Abnormal transaction amount 1. Alert client – cancel or continue 2. Alert broker for attention R 2 Payment due (time event) Payment unsettled 1. Notify client for payment or reduce holding 2. Alert broker for attention R 3 Client withdraw cash (from client portal) Withdraw amount exceed threshold 1. Alert broker for possibility of client attrition 2. Alert also the manager if important client R 4 Client withdraw cash (from client portal) New account balance < unsettled payment 1. Remind client for payment 2. Alert broker for attention R 5 Complaints (from client portal / broker call center) - 1. Record details in data warehouse 2. Alert broker to handle (possibility of client attrition) 11

Rules triggered by events related to client profile Event Condition Action R 6 Client attrition alert Valuable client (from analytic engine) 1. Alert broker for possibility of client attrition 2. Alert also the manager if important client R 7 Client change in segmentation (from analytic engine) - 1. Alert broker for important clients 2. Subscribe the client to new information sources and news 3. Suggest cross-sale R 8 Client change interest (from client portal) - 1. Alert broker for important clients 2. Subscribe the client to new information sources and news 3. Suggest cross-sale 12

Rules triggered by events related to the market environment Event Condition Action R 9 Particular stock price change (from environment listener) User-defined alert price met Alert the clients R 10 News for particular stock arrived (from environment listener) R 11 Severe stock risk signal, e. g. , total holding value drop by 10% (from analytic engine) Margin clients 1. Notify client for payment or reduce holding 2. Alert broker for attention R 12 Severe stock risk signal (from analytic engine) Cash clients 1. Notify client (as service) 2. Alert broker for attention for important clients Send the news to the clients who have subscribed to the information of the stock 13

Rules triggered by events related to the brokerage firm Event Condition Action R 13 Broker resignation (from managerial applications) - 1. Notify affected clients about the change in broker for them 2. Alert the new responsible broker for attention for important clients R 14 Important news / changes in policy of the brokerage firm (from managerial applications) - Notify all clients and brokers 14

Front-end – Broker Call Center n n n Relevant alerts - notify the broker to carry out follow-up actions Demographical data, estimated trade limit, event logs Follow-up service activities, personalized recommendations Exploring cross-selling opportunities of stock derivatives Record clients’ complaints, common queries, special requests 15

Front-end – Managerial Application n n Common reports with pre-defined queries / customized reports Events and alerts related to high-valued clients directed to managers too 16

Front-end – Client Portal n n Consistent style and usability instead of eye-catching effects Avoid graphical and multimedia interactive contents 17

Lessons Learnt n n n Change of workflow and information management procedures Should roll the online CRM web interface to the clients only after the CRM system’s operation has become smooth internally Phased approach in system development and deployment Verification of the required data from the legacy system is difficult and time-consuming SME brokerages in Hong Kong are not quite familiar with advanced information technologies 18

Benefits of Event Driven Approach n n n Business rules, in general, can be naturally modeled as ECA rules, which can be used for activity enactment, monitoring and exception handling Implemented with JMS or other contemporary technologies, such as Web services Rules can be added to, deleted from and modified for a system more easily than traditional software development approaches Business environment keeps changing, users expect new features or functionalities Abandon of a pure artificial knowledge discovery approach - turning understanding into action is the key to deriving real benefits 19

Conclusions § § § Pragmatic event driven approach to CRM for the e-Brokerage industry Practical system architecture and a working prototype – unified platform for automated actions and human expert attention Discuss the use of different categories of business events from the clients, the firm and the market environment Focus on efficient and timeliness of CRM activity enactment ECA-rule paradigm § § systematic specification of handling asynchronous business events enables effective enactment of the specified handlers (actions) deliver the most appropriate care to clients Suitable for implementing CRM system for other e-commerce sectors by turning knowledge into business actions 20

Future Work § § § Scale up CRM solution with J 2 EE Mobile CRM with J 2 ME Application in other service industries, e. g. , insurance Integration with workflow management system B 2 B integration / Electronic Contracting n n A Data-driven Methodology to Extending Workflows Across Organizations over the Internet (HICSS 36) An Architecture for E-Contract Enforcement in an E-service Environment (HICSS 36) On e-Negotiation of Unmatched Logrolling Views (HICSS 36) Enterprise Document Management n A Watermarking Infrastructure for Enterprise Document Management (HICSS 36) 21