On the Scene--Talk Good: Best Practices in Speech Deployment
Speech recognition technologies continue to evolve, but the market has seen sluggish adoption, partially due to hefty costs and complicated deployments. Here, a review of some industry experts' tips for organizations evaluating speech technologies, from the annual SpeechTEK International Exposition and Educational Conference in New York in August:
Collect and examine vertical-specific best practices to tap into examples of how companies within the same industry are deploying speech applications and capitalizing on their potential to cut costs.
Evaluate current performance.
Scrutinize the customer experience by taking into consideration essential contact center metrics to identify strong suits and areas for improvement. Develop an ROI model to validate deploying a speech application, suggested Elka Popova, industry manager of enterprise solutions and IP communications at Frost & Sullivan. "If you're...looking at self-service solutions, you should look at [elements like] the number of calls you receive per month, the automation rate, and the average queue time."
Set and prioritize sensible short- and long-term goals.
Too often, organizations jump into deployments without properly establishing objectives, without which gauging progress is virtually impossible. "Make sure [the goals] are realistic and achievable," said Daniel Hong, senior voice business analyst at Datamonitor.
Weigh vendors and their systems.
"One [reason] why adoption has been slow is, there is a high price tag attached to speech deployment, coming mostly from the professional services needed to deploy a speech solution," Popova said. A 2005 Frost & Sullivan end-user survey showed that on average, more than 60 percent of IVR interactions are touch tone, while about 34 percent are solely touch tone.
Companies should evaluate solutions based on price and performance, such as the ability to self-learn, accuracy, and languages supported. Vendor considerations should be based on points such assupport for multiple speech engines, technical support, professional service expertise, customer service, overall reputation, and viability. Investment protection should also be top of mind, Hong said. "You don't want to invest in a company that is not financially stable. If the solution fails, who's responsible? You have to really draw the lines within the service level agreement."
Take a test drive and then fully deploy.
Launch a pilot project prior to implementation to see if the established goals can be accomplished and to get a glimpse into areas that need to be tweaked before a complete implementation. But finishing an initial rollout does not translate into crossing the finish line. Engage in ongoing monitoring and maintenance and keep an eye on goal progress.
Get customer feedback.
"Just looking at raw call center [data,] you can't really gauge customer service from that, so you should really go to a third party to conduct surveys with the customers to see if they are satisfied and happy with the speech," Hong said. "Deploy small, and start getting [larger], but before you get [larger], make sure your operations are efficient, that your customers are accepting it, and make sure that internally, you have all of the support that you need."