Monitoring and reviewing call centre agents’ work was always an incomplete and unsatisfactory process in the past, explained CallMiner’s VP of international, Frank Sherlock. How could it be otherwise, when it has historically been a manual and unscalable job, reliant on chance sampling and listening?
“As an agent, say I make a couple of hundred calls in a week. Traditionally, maybe three to five of my calls are assessed on a weekly or a monthly basis. That is, someone will sit and physically listen to the call — a human being listening to another human being talking to another human being.
“But who would be happy to be evaluated based on 2 or 3% of their actual work? And added to that, manual listening is absolutely subjective — two supervisors could listen to the same call and one could conclude that it was really good, while the other thinks it’s pretty poor. In fact, it’s so inconsistent, ratings from the same person can vary wildly from one day to the next, depending on many factors.”
From random humans to machine-learned consistency
Automated agent performance monitoring solutions, like CallMiner’s Coach, offer in-depth insight and rich analytics to enhance key metrics — while also providing fairness and motivation to the agents themselves. Agents receive real-time feedback on their performance and ranking, including areas for improvement, based on ALL their calls.
“On the surface, you might get a reaction of, ‘oh this is Big Brother…’” Sherlock continued, “but when agents experience these benefits, they really appreciate the insight. Not only are they getting direct feedback on how they can improve and fix specific issues, some of their calls that were previously lost in the noise may get showcased as great examples to their peers. It’s a way of democratising data, making it available to the agent immediately after each call, so it’s all actionable and useful.”
Parsing the content of call transcripts at scale, to meaningfully analyse the way sentiment and emotion is expressed, requires artificial intelligence trained on the nuances of human conversation. “For example, in UK contact centre parlance, there are about 200 different ways that dissatisfaction is expressed,” Sherlock reflected. In addition to identifying the negative feeling — anywhere on the spectrum from profanity to mild sarcasm — the system analyses the agent response, and how the emotional tone of the whole interaction varies and modulates.
Optimising excellence
“As part of the complete contextual analysis, it rates how the agent takes ownership, expresses empathy, and ultimately resolves the issue and delivers the best outcome for the customer.”
