Improvement of call center customer service in a thai bank using disco fuzzy mining algorithm

The main objective of the study was to benchmark performance, control discrepancies, and investigate variations of a bank customer service call center data dealing with incoming calls of its clients and customers. To do this, initially an event log consisting of a total of 625,767 process instances (i.e., events) was collected from a private bank in Thailand for the month of July 2015. Using Disco fuzzy mining technique as a process mining tool enabled us to simulate and create authentic visual models/maps from the collected event log in form of fuzzy mining graphs. To better investigate the behavior of the bank's clients/customers and administrators/operators dealing with inappropriate (not successful) customer service and calls, only “Failure” or “Not Responded” types of process instances/events were chosen and selected for the study. The findings showed that the number of the incoming calls made into the call center (customer service section) due to the “over card limit” problem was the highest (i.e., with 4,667 process instances in total) compared with the other problems within the call center event log. On the other hand, the results showed that almost 32%of the “Over Card Limit” type of the problems were not solved at the first attempt (i.e., clients/customers have to re-dial and re-contact the operators in charge of the section again for the second, third or sometimes multiple-times). Similarly, our results showed that the problems occurred due to the reason why the “card number is already activated” allocated the second highest number of problems (i.e., with a total of 2,068 process instances in total) within the call center event log. Interestingly, 10% of the incoming calls facing the “card number is already activated” were not solved/fixed at the first attempt and clients/customers need to re-contact the call center operators again afterwards. With the same token “record not found” (with 52% of the calls not solved at the first attempt), “account not found” (with 68% of the calls not solved at the first attempt) and “account does not exist” (with 61% of the calls not solved at the first attempt) types of the problems allocated the third, fourth and fifth most frequent incoming calls made into the call center customer service section. Eventually, the results of the study can be used in order to enhance and improve the performance of the customer service processes in a more efficient, effective and timely manner.