When I was in Vegas for ITExpo, I participated on a Voice Analytics panel at the SmartVoice co-located conference.  Speech / Voice Analytics is really taking off.  Voice analytics is basically taking all the recording that is going on, and analyzing that information into more structured and reportable formats.  In the past, we called this “data mining”.  This is “mining” of the voice calls.  Dialogic gets involved in that because we have many contact center customers and we’re involved in recording of conversations, and have been doing that since the inception of these call centers years ago.  We recently did some feature requests where we recorded the binaural conversation – that is what each person heard, as opposed to a single stream of the conversation which is not necessarily what each person heard because of the talk-overs, etc. 

I felt like the whole voice analytics space was kind of languishing for a few years.  First of all, it got a bad rap from speech to text.  Even the day of the talk, I left a voice mail for one of my Dialogic colleagues.  Today, if you leave a voice mail in Dialogic, you get not only an attachment with the voice mail that you can play back from your computer, but you also get a transcribed version of the voice mail.  The transcribed version was comical – really not much in it that I actually said. So the whole voice space still has kind of a bad rap from that as this has been a tough problem to solve.  But Siri has energized the space, as well as the ability to talk to your car.  So we are beginning to overcome that.  But also the cloud has energized this space because a company might want to try voice analytics but might not want to have a large up-front investment. The cloud might work better for them in this case.  Also, the cloud has sparked some innovation/specialization in this space.

Let’s get back to the actual data mining.  The contact center providers or these voice analytic specialist companies can then analyze all this information.  Use cases fall into two major categories – helping provide better customer service and reducing agent costs.  For instance, let’s say that many customers are like me – I’ll try and do self-service on the web if possible, but then call only if I have a question.   I’ve been “trained” that if I call, it will take longer, so I try not to call.  If the contact center can find out if there are specific trends about why many people are calling, then maybe there is some fix on the web they can do.  This would help in both providing better customer service and also in enabling faster calls to happen, so maybe fewer agents could be required.  And in that case, presto, there is cost savings.

You may ask “don’t they already know why the people are calling” and the answer may surprise you.  If there are so many calls, then the contact center would have anecdotal info about that, but probably not actual data.  If they can understand the actual reason, then more specific and concrete actions can be taken in the self-help web area.

You can take this a step further.  Agent training can happen faster.  Script adherence can be monitored.  Transfer/Holds/Callbacks can be reduced.  Costs can be further reduced and customer service can get better.  And take this even further.  You can potentially get into real time analytics.  In real time, you can maybe do “emotion detection” and if a caller is getting increasingly frustrated, then maybe pass them automatically onto a more experienced agent, or start to coach the agent in real time.  There are a lot of possibilities here.

Ultimately, the data that has been recorded can be “mined”.  This space is not languishing any longer and in my opinion, will continue to have interesting innovations.  Intrigued? There is interesting market research out there.