8 Unconventional Uses of Speech Analytics

Speech-to-text and speech analytics technologies are often focused on capturing, understanding, and harnessing the voice of the customer to improve products and services. Time and again, speech analytics are used to unearth customer sentiment, cultivate better customer satisfaction, identify contact drivers, monitor customer views on competitors, and identify customer issues with communications channels (website, ecommerce, self-help). It’s becoming more and more apparent, however, that the speech analytics lens can be and should be turned internally — toward contact center employees and background processes. This is especially true for work-at-home environments, where managers can’t observe agent performance first-hand. Monitoring home-based agents with speech analytics during the pandemic could be particularly enlightening. Here’s how you can use speech technology to ensure your agents are doing their best for your customers.

1. Gauge employee comfort level – Employees may not give completely honest answers on surveys or in one-to-ones. You may, however, find clues about how they’re doing or the impact they’re having on the brand in the words they use during customer interactions.
2. Measure Internet connection quality – Modern speech analytics tools can separate different speakers. Using a string of queries, we can capture customer-spoken phrases like “your voice is muffled,” “I can barely hear you,” “you’re cutting in and out,” “too much background noise,” etc. to monitor the quality of work-at-home agents’ Internet connections. For one client program, 6.0% of agents experienced sound-quality issues and, after working with them, the sound-quality issue level dropped to 3.1%.
3. Detect fraud, privacy, and security issues – If keeping customer data safe is a concern, consider speech analytics to monitor agent data handling. For one financial services client, HGS created a series of queries to segment high-risk transactions and then, within those, to flag the disclosure of CVV/CV2 numbers on credit cards. Using those flagged calls, the client can rule out any mischief or connect the card back to a particular agent(s) if any card misuse occurs in the future. Speech analytics is also used to establish the prevalence and frequency of CVV/CV2 calls; if, for example, we discover that 16% of calls involve CVV/CV2 numbers and agents deviate significantly from that average number, we can investigate.
4. Monitor non-talk time and dead air – Long, agent-initiated periods of dead air may indicate problems. Use speech analytics to detect and then identify the reasons for those uncomfortable agent pauses. For example, for one client, we have configured speech analytics (a) to flag every instance of non-talk time that comprises more than 40% of the total call duration and (b) to flag agents that have multiple instances of this. In the early days of the pandemic, we discovered a pattern — that newly hired home-based agents were struggling to respond to customers with disputes.
5. Flag deceptive, undesirable sales tactics – When agents get commission for making a sale, unfortunately, there may be some temptation to engage in customer deception or fraud. To prevent this, consider queries for queues where agents extend offers. A speech analytics tool can reveal agents who use false phrases like “it’s the last day of the sale,” “this offer is only for a select few customers,” or “available for a limited amount of time.” Then have your quality assurance specialists and team leaders coach, monitor, and/or discipline appropriately.
6. Ensure justified transfers and escalations – Transferring customers is not only inconvenient – it’s costly. You can use speech analytics queries to find out why agents are transferring customers. For one client, using speech analytics, HGS identified the transferred calls, flagged the relevant agents, and then honed in on the transfer drivers with more keyword searches and agent input. We discovered that customers were contacting our agents as a last resort because they couldn’t get through on a line not within scope.
7. Comply with policies – Many transactions require customer confirmations (e.g., reprint an invoice, change an address). Using speech analytics, managers can confirm agents are confirming the spelling, dates, etc.
8. Address unresolved issues – First-call resolution is a goal for many contact centers. Searching for phrases like “I called yesterday,” “this is my third time calling,” “my wife called earlier,” etc. enable you to understand why; to determine whether scripts, processes, or websites need to change; and to decide how to change them. The queries can sometimes even confirm that nothing needs to change! For one client, we discovered that repetitive calls were inevitable due to the average advanced age of the customer population.  

Speech-to-text and speech analytics are incredibly powerful tools. With the right setup, you can synthesize data from not only calls, but also chats, emails, social media, direct messages, and SMS/texts. Long gone are the days of assigning dedicated human resources just to listen to calls, as are the days of making decisions based on a small retrospective call sample. With speech analytics, you can monitor 100% of the calls, automatically, practically in real-time to unearth new insights about your customers and your team. For additional speech analytics inspiration, check out our one of our latest financial services case studies.