Top Menu

Speech Analytics: The Next Big Thing for Contact Centers 

Speech Analytics: The Next Big Thing for Contact Centers

Speech Analytics: The Next Big Thing for Contact Centers

4/5/2006
By Donna Fluss
ICCM Newsline

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

Contact center managers are constantly confronted with the challenge of reducing their budgets and absorbing an increasing volume of transactions without additional resources. At the same time, they are looking for ways to improve the customer experience, enhance service quality, reduce costs and improve productivity.

Speech analytics has the potential to accomplish all of these goals. Perhaps even more compelling, speech analytics applications can identify and quantify the impact of both contact center and non-contact center operations on an enterprise’s customers. This empowers contact center managers with a great deal of new information.

Managers are generally blamed for poor performance when transaction volumes (calls, emails and chat sessions) skyrocket, through no fault of their own. Speech analytics allows them to identify and explain why transaction volumes are growing. They can then work with the appropriate organizations—marketing, sales or operations—to address the underlying issue. Speech analytics functions as an early warning system, providing the tools to rapidly and unambiguously identify trends and issues so that the enterprise and individual managers can take corrective action much sooner than was possible in the past.

Speech analytics applications have been in development for many years and are already being used by governments around the world to identify and analyze security risks. Since 1994, a few speech analytics vendors have been trying to enter the large commercial sector. It has taken quite some time and significant enhancements in application functionality for such applications to be effective in the commercial arena, which precludes custom programming. Contact centers are a natural fit for speech analytics, as many already either record 100% of calls to meet Federal Trade Commission (FTC) guidelines or capture a random percentage of calls for quality assurance.

Even though commercial contact center applications are new—about two years old and in their first technology generation—some vendors have already released updates in response to needs identified from implementations. If speech analytics follows the typical development cycle, the market can expect second-generation solutions with major technical and implementation breakthroughs about 2008.

How Speech Analytics Works

There are two approaches to recognizing speech: large vocabulary continuous speech recognition (LVCSR) and phonetic-based search. LVCSR engines do a speech-to-text conversion of audio files. Phonetic-based applications separate words into phonemes, the smallest components of a spoken language.

Speech analytics engines vary greatly and are quite complex. Although the underlying approach may begin similarly in many of these solutions—either a speech-to-text conversion or breakdown to phonemes—the process following this step distinguishes one application from another. Most vendors will say that the initial speech-to-text conversion is not the most important part of their application, so much so that many applications that start with LVCSR nevertheless claim not to be LVCSR engines. As the accuracy of speaker-independent speech-to-text conversion is thought to be somewhere in the range of 20% to 40%, it’s essential for a speech analytics application that begins with this process to use complementary technologies and applications to enhance its results.

It’s challenging to select an application because of the differences in how vendors have applied the technology. A clear differentiator between applications is the user interface. Some vendors have invested in making it easy for customers to use their applications, while others still use older interfaces that require customers to be more technically knowledgeable. This is an area that has seen innovation and enhancement in the last two years. Although the user interface is a differentiator that impacts system usability, be careful not to base your purchase decision on this one criterion.

Implementation Requires Resources

There are 13 speech analytics offerings available today, and two are expected to be released this year. Some 250 contact centers have implementations and most are in their early stages.

Speech analytics applications do not work “out of the box.” Despite claims to the contrary, it takes a great deal of resources to make these applications work effectively. In fact, most of the vendors in this market make significant revenue from consulting services. (Of interest, a vendor may be willing to negotiate the cost of the application, but not its consulting services. Take this as a warning.)

To realize the greatest benefit from a speech analytics application, an organization needs to invest its own resources in the implementation, as well as ongoing support and maintenance of the application. Even if the vendor performs the initial implementation, the enterprise needs to be positioned to enhance and tweak it on a regular basis.

Speech analytics applications that have user-friendly interfaces can be managed by business analysts who have minimal technical knowledge. Once trained on the application, users should be able to make modifications and realize benefits, without ongoing support from the vendor. Other applications require a great deal of hands-on support from the vendor.

, ,