Top Menu

The Facts about Speech Analytics 

The Facts about Speech Analytics

The Facts about Speech Analytics

By Donna Fluss

  Printer Friendly Format       View this document on the publisher’s website.

Speech analytics is an emerging application that is capturing the attention of contact center managers because of its ability to identify the root cause of customer issues. Speech analytics applications work by structuring customer conversations and creating a database of call reasons or phonemes that can be mined. Analysis of this information yields a detailed accounting of the reasons why customers call. This enables contact center managers and executives throughout the corporation to address the underlying problems that generate call volume. It also provides insights into customer needs and wants regarding new products and services.

There are a lot of questions and misconceptions about speech analytics applications, a number of which are addressed in this article.

Question: How is speech analytics different from speech-enabled IVR or speech recognition?

Answer: There are two parts to this answer, the technical side and the practical aspect. The technology behind speech recognition and speech analytics is similar. How they are used is very different. When using speech recognition, customers expect the application to be 100% accurate in recognizing every conversation. Speech analytics has more leeway, as it collects hundreds or thousands of conversations to identify trends.

Additionally, when evaluating speech recognition applications, the most important factor is how well the application interfaces with customers. Much of the underlying speech recognition software (speech engines) is very good today, and its accuracy is continuing to improve. What isn’t always as good is the voice user interface (VUI) – what the customer hears when he or she calls. Developing VUIs requires expertise. This skill is significantly different from designing touch-tone IVR applications. If a speech recognition application is difficult to use or doesn’t follow common sense, it is not designed well and is likely to fail. This can have a significant negative impact on a company because it means that technological shortcomings are disappointing and frustrating customers. Poorly designed speech recognition applications continue to be a serious problem for the industry, but this issue is totally independent of the many positive developments in speech analytics.

Question: What is the primary challenge in using speech analytics?

Answer: Speech analytics applications are necessary because companies do not know all of the reasons why customers call. The biggest challenge is building the taxonomies and databases that are used to identify the issues or root causes of calls. The more advanced speech analytics applications can identify a broad range of themes in customer conversations, even those that are not predefined in the application database.

Question: Are there differences between the speech analytics products in the market?

Answer: Yes, there are significant differences between the offerings, although all are new. The differences range from the underlying technology to the vendor’s ability to execute. There are three categories of speech analytics applications:

  1. Transcription-based products that attempt to transcribe 100% of conversations.
  2. Keyword/phrase-based solutions that identify key words or phrases within conversations.
  3. Phoneme-based offerings that break conversations into their underlying phonemes, which are the primary sounds that create words.

This is a new market with at most 200 implementations in contact centers around the world, as of November 2005, so there are still many unknowns about these products. What is well understood is that when used properly, these applications result in a rapid return on investment because they identify the root cause of calls and allow contact centers to resolve issues rapidly.

Question: Who are the competitors in the contact center speech analytics market?

Answer: The vendors fall into three primary categories:

  1. QM/recording vendors: etalk, Envision, Mercom, NICE, Verint, Voice Print, VoiceLog and Witness.
  2. Stand-alone speech analytics vendors: Aurix, CallMiner, Nexidia, Sonum and Utopy.
  3. Contact center vendors: SER (which offers Aurix).

With the exception of etalk, which has its own speech analytics application (inherited from its new parent company, Autonomy), all of the QM vendors either use an engine from one of the speech companies or resell one of the stand-alone solutions.

The Future of Speech Analytics

The speech analytics market is in its infancy, but will grow rapidly because of the benefits these solutions deliver to contact centers and enterprises. The reach of speech analytics is also expected to increase, as it is adopted in other operating areas where it would be advantageous to structure conversations, such as bank branches, government agencies and doctors’ offices. There is so much rich and valuable information that is shared with enterprises today that cannot be used because it isn’t structured. Speech analytics provides a vehicle for structuring, interpreting and using the rich store of information that remains hidden in customer conversations.

, ,