Centerfirst has performed hundreds of thousands of monitors for our healthcare and bio-pharma clients. We are constantly looking for ways to bring greater accuracy, insight, and efficiency to our clients. Recently, we tested the latest voice to text automation tools and came away with some opinions on where they may apply and where some progress still needs to be made.
Definition – Automated quality monitoring takes voice recordings from calls in the contact center and translates them to text for analysis. Automated monitoring works on any text but this blog focuses on the voice to text application.
In order to test the voice to text software, we recorded fictitious calls using common medical terms with brand name products. The fictitious calls used scripts which were read by three different individuals to compare the results.
- The word “medication” was accurately identified on every fictitious call, but the phrase “medication by pill” was missed on each call where it was used in the script.
- “…medication by coan”
- “…medication back to you”
- “…medication by appeal”
- None of the brand name products were captured accurately, and the same script, read by a different person, had different results across the calls.
- Common words used in the discussion were captured with high accuracy.
- “treatment”, “patient”, “liquid”, “alternate”
Limitations of testing: Not all voice to text applications were tested, a small number of recorded calls were used, and the automated software was not assessed for non-call activity.
What are the current shortcomings of the automated monitoring software for bio-pharma:
- We have very low, to no, confidence that automated voice to text monitoring would provide assurance that adverse events or product quality complaints are being handled completely and accurately.
- Although word choice can affect the customer experience, an agent’s tone is the number one factor influencing the customer experience, and the voice to text software falls short on tone assessment.
What do we see as the best uses of automated monitoring in the contact center?
- Translation of simple words is good and getting better. There is some data from Microsoft that show translation error rate of under 10%, which is down from 15% in 2004. This will continue to improve. Automated monitoring can monitor 100% of your recorded calls. By monitoring all of the calls, the software may be able to detect certain signals that the traditional sampling might not discover.
- Automated monitoring can analyze all of your text data. If the center generates a lot of text data from email, chat, social media, and other sources, combining the call data with the other sources for analysis can provide a broad view of the customer interface.
We will continue to assess automated monitoring and use it in the places where it makes sense for our healthcare clients.