How to Use Voice Mining to Tune In to the Voice of the Customer

For businesses with call centers to thrive in today's economy, it's more important than ever to tune in to the voice of the customer. The amount of conversational data being collected digitally in call centers is growing faster than our ability to deal with it, so customer insights buried in this data are missed. Voice mining technology can now pull all of this audio data together, analyze it and report on it. Using voice mining technology, Knowledge Center contributor Mary Grace Crissey explains how you can analyze audio data to increase customer satisfaction and loyalty.


Voice mining technology digs into mounds of stored but often unused data, giving you ears to hear your customers. Your customers are telling you and your competitors what they want to buy and what prompted them to choose you. But manually rereading call center notes or replaying phone conversations can't possibly deliver customer intelligence in a timely fashion, so valuable information remains stuck and inaccessible.

Empowered with analytics, though, you can gain the advantage of being able to spot trends in everyday conversation threads to predict and respond to market opportunities and pitfalls. Four components are required for organizations that wish to take full advantage of voice mining technology.

Three of the four components are commercially available: audio voice-to-text convertors, text mining analytics, and business intelligence reporting and performance dashboards. The other component is the person who drives this system. Adding even a single component is noteworthy, but those organizations who implement all four in harmony will reap the most from voice mining technology.

Component No. 1: Audio voice-to-text converters

Before conversational audio data can be processed by a computer, it must be translated into electronic format. The "analysis" involved at this step is performed by a phonetic index search that automatically transforms a captured audio signal into a sequence of phonemes.

Many voice transcription systems today are supplemented with spell-checkers because words can be taken directly from a dictionary. What makes the spoken text harder to understand is the lower accuracy of the words chosen by the transcription algorithms when it comes to deciphering every "um" or "ah" sound.

Even with great advances over the past five years, there are times when these translations are impossible to read. Because the reason you are trying to "read" these spoken words is to understand and comprehend what is being said, you can get much further by taking this data and performing pattern detection and concept-clustering techniques. These are available in the second component of voice mining: text mining analytics.