Speech analytics is a technology which is receiving a lot of attention in the contact centre, and quite rightly so. The potentials are enormous, but only for those who accept it as an enabler for a solution rather than a product which will deliver on its own. However, current costs mean that ROI for speech analytics is limited to medium–large contact centres.
The Need for Speech Analytics
Companies already recognise the need to listen to what’s going on in their contact centres. Voice recording, screen recording and quality management have been available for years. These solutions allow calls to be filtered on the usual meta-data, such as agent ID, date/time, call duration, CLI, etc. Voice and screen can be replayed in synchronisation to get a picture of how the call was handled and to enable coaching to improve future performance. Already in a mature state, these solutions continue to evolve and deliver good results and ROI.
But the more savvy contact centres are also questioning the content of their calls. The cost of communication is high, therefore unnecessary or inefficient communication should be reduced. Competition is high, so knowing how customers react to offers and service is important. If the worst happens and customers complain then the ability to analyse and rectify the causes will ensure customer satisfaction.
For companies with high call volumes speech analytics offers the ability to analyse the content of their calls efficiently and consistently.
Current solutions utilise three technologies to categorise calls.
Keyword Spotting is probably the most common and most cost-effective. It enables calls to be categorised automatically by looking for calls containing certain words or phrasing. These calls can then be manually evaluated, used for coaching, or used as part of root-cause analysis. The cost of setting up and configuring keyword spotting can be in the range of £10k’s, depending on the scale of the contact centre. Currently this should be considered the entry-level speech analytics and where most companies are likely to see a ROI.
Phonetic Indexing is the next level of sophistication. This requires more detailed processing to translate calls into a database of phonemes. Phonetic indexing is slower than keyword spotting, and uses a lot more disk space, but it does allow faster repeat searches once the phoneme translation has been completed.
Speech to Text Transcription is the most expensive and the most powerful. Calls are automatically transcribed into text files which can be analysed quickly and as many times as you like. It relies on a high level of accuracy which can only be delivered by training the system to recognise words in the specific contact centre domain. This training requires a lot of manual interventions, so the time and cost of implementing speech to text can be high, typically £100k+.
The Future of Speech Analytics
The speed of speech analytics results is one of the areas which are likely to improve. Current solutions are post-process, which means the results are available some time after the event. The delay depends on the resource available to analyse the calls. In the future this is likely to speed up to the point where near real-time processing is available.
In the future we’re likely to see the emergence of more sophisticated speech analytics which will allow age or gender spotting. Imagine being able to determine the age of your caller, in real time, and route them to customised services.
- Typical uses for Speech Analytics
- Speech Analytics in Customer Service
- Word Spotting vs. Phonetic Search vs Speech Recognition
- Mike Murlay at ASC