A leading B2B company with a network of worldwide data centers had problems with its ticket processing—namely, substandard performance in ticket resolution and inefficient staffing loads. The objective was to find the optimal correlation between quality of service, number of human resources needed, and cost. To attack this problem would require addressing issues such as a custom-built, poorly documented CRM system; a mass of unstructured data; and protecting sensitive data (which meant that typical cloud platforms available on the market were out).
Our initial focus was on the data: we ingested everything the client had—including reviews, chats, what worked, what didn’t—structured it, and then applied the analytics to the reformed data. We used advanced NLP techniques (such as Word2vec and LDA) to find and compare similar tickets and implement topic modeling so that the system would “learn” from—for example—interactions that were positive and who did the best work. Our solution involved creating several time-saving capabilities that extended the usefulness of the existing CRM, including automated category assignment and automated technician assignment. These capabilities meant the system could now map a ticket to a problem type and then route it to the technician best able to address it (thus reducing the number of iterations needed to resolve it). At the management level, we developed dashboards to display productivity metrics by ticket type, team, team members, and data center.