Major Telco: Automated NLP Pipeline for Customer Feedback

Situation

Telco, a national communications provider, delivered over 100,000 customer service interactions per week and collected more than 4,000 customer feedback surveys weekly. Each survey included a mix of ratings and free-text responses. While ratings offered a broad sense of customer satisfaction, the sheer volume of free-text data meant Telco could not efficiently extract actionable insights, limiting its ability to deeply understand and improve the customer experience.

Task

Develop an automated Natural Language Processing (NLP) solution to extract actionable insights from free-text survey responses—enabling Telco to discover what influences customer loyalty, which issues most impact satisfaction, and how to prioritize improvements.

Actions

• Processed and cleansed six months of survey text, applying advanced NLP techniques including topic modeling, sentiment and aspect analysis.

• Mapped feedback to topics using a custom-built ontology and semantic similarity techniques.

• Used machine learning (XGBoost, Shapley values) to predict satisfaction drivers.

• Deployed interactive dashboards visualizing sentiment, key issues, and Net Promoter Score (NPS) trends for continuous business monitoring.

Results

• Identified top drivers of positive feedback (service plans, customer service, staff) and negative feedback (network connection, coverage, internet).

• Automated analysis reduced review time and bias, delivering real-time insights at scale.

• Equipped Telco to focus on improvements that have the biggest impact on customer experience and loyalty.

• Recommendations included maintaining the topic ontology and exploring labeled data for future model enhancements.

Impact

The automated NLP solution empowered Telco to unlock rich, actionable customer insights—making continuous experience improvement and data-driven decisions central to their strategy.