Case Study: a UK bank improves customer service with Tiger Analytics NLP-powered conversation intelligence

a UK bank achieves 80%+ intent accuracy with Tiger Analytics

a UK bank faced challenges in analyzing its high volumes of unstructured customer interaction data from phone and chatbot channels. Noisy transcripts, unclear tagging rules, and a reliance on third-party vendors limited its ability to gain consistent insights for improving service quality and first-call resolution. To address this, the bank partnered with Tiger Analytics to build an in-house NLP capability.

Tiger Analytics implemented a solution that ingested and cleaned transcripts, then used BERTopic for intent modeling and semi-supervised fastText classifiers for call-resolution classification. This approach, refined through active learning, provided the bank with clear reasons for calling and transparent resolution outcomes. The project achieved over 80% accuracy for intent and over 70% for resolution, leading to reduced repeat calls, improved agent benchmarking, and unlocked revenue opportunities.