Case Study: Leading Telecommunications Company achieves automated contact-center categorization and faster issue resolution with Blueocean Market Intelligence

Automating the data categorization process with predictive models

Leading Telecommunications Company engaged Blueocean Market Intelligence to automate the categorization of 6.1 million contact-center chat records into hierarchical categories. The client’s challenge was to extract the right category from duplicated, erroneous and jargon-filled data; Blueocean applied predictive modeling (using an SGD classifier and distributed computing) to build a scalable categorization solution.

Blueocean Market Intelligence implemented a five-step process—data segregation (65/25/10 split), thorough data cleaning, model building with distributed computing, feature engineering/validation, and final testing—to avoid overfitting and tune performance. The resulting predictive model achieved an overall accuracy of 43.7 percent and a hierarchical accuracy of 57.60 percent, enabling quicker, more accurate customer-data categorization, faster issue resolution, and improvements to the customer service model that positively impacted sales and key performance indicators.