What questions are my customers asking?: Towards Actionable Insights from Customer Questions in Contact Center Calls
This blog refers to our paper accepted at Interspeech 2023, Dublin.
Background
In the dynamic world of customer support, contact centers serve as the unsung heroes, managing a diverse array of customer interactions across various channels, including chat, email, and phone calls. Among these, phone calls remain a preferred method, cherished for the immediacy of human interaction and swift solutions.
Within the symphony of call center conversations, customer questions emerge as the keynotes that carry profound insights into their needs, expectations, and preferences. By analyzing the tapestry of questions exchanged, call centers can unlock a treasure trove of valuable information about customer behavior, guiding them towards optimizing their services. It’s a profound revelation — a single question can illuminate a path towards enhancing the customer experience.
Empowering Business Decisions
Imagine this: each query posed by a customer during a call is a pixel in a larger picture that unveils the customer landscape.
A frequently asked question isn’t just a repetitive blip; it’s a signal pointing towards a recurring issue. For instance, if numerous customers inquire about a specific problem with a product or service, it’s a clear signpost for improvement. The call center can promptly diagnose the issue, leading to rapid resolutions and averted complaints — a recipe for elevated customer satisfaction.
Beyond surface-level analysis, the types of questions customers ask can offer deeper insights. When agents encounter questions they struggle to address, it could indicate areas of knowledge deficiency. Such revelations become catalysts for proactive training interventions, nurturing agents to become even more adept problem solvers.
In the quest to transform questions into actionable insights, we’ve crafted an end-to-end framework that marries technology and human expertise. Let’s walk through its facets:
- Contextual Question Rewrite: We begin by extracting raw question spans from call center conversations. But that’s not all — we take it a step further, delving into the realm of co-references and disfluencies to ensure contextual accuracy. It’s like decoding a secret language, making the data ready for comprehensive analysis. We do all this through our in-house contact-center LLM (large language model).
- Streamlined Analysis: Questions don’t exist in isolation; they’re part of a larger narrative. We harness the power of grouping customer inquiries, weaving them into coherent categories. Alongside this categorization, we knit the fabric of call meta-data — response times, origins of questions — enriching our understanding and guiding strategic decisions.
- Human-in-the-Loop Excellence: Our framework doesn’t stop at analysis; it’s a dynamic ecosystem. Through a human-in-the-loop process, we fine-tune our Knowledge Base, infusing it with novel customer questions and refining responses. The result? Quicker responses, satisfied customers, and an ever-evolving support system.
Conclusion
Our innovative system extracts, rewrites, and groups customer questions from millions of contact center calls, enabling businesses to analyze customer needs and pinpoint improvement areas. By identifying knowledge gaps and facilitating efficient knowledge base updates, our solution empowers businesses to enhance customer service and reduce operational costs through agent coaching based on response times for specific questions.
This work has been accomplished through a collective effort involving Varun Nathan, Devashish Deshpande, Jithendra Vepa, Cijo George
Learn more about how we’re changing conversation intelligence for contact centers around the world at Observe.AI.
