Speech Technology Magazine

 

Advanced Analytics Offer Greater Precision

The latest technology goes beyond simple audio mining for words and phrases.
By Phillip Britt - Posted May 1, 2007
Page1 of 1
Bookmark and Share

Continental Airlines call center in Tampa, Fla., takes a combination of calls for both domestic and international travel. The airline then performs a call mix survey report each month to determine the breakdown of calls coming into the reservation there was a sale made or not, whether it was a ticketing question or a policy/procedure question, etc., for a total of 30 categories in all to determine the success of the call center.

Though these call analysis reports have been done for several years, it wasn't until recently that Continental could listen to a large volume of calls without having to invest countless hours to accomplish the feat. Previously, it was a manual process that permitted Continental call center supervisors to listen to about 1,000 to 1,500 calls per month, only 2 percent to 3 percent of the total number of calls actually received. Any analytics were shaky because assumptions were based on a small sample of actual calls.

However, once Continental Airlines added speech analytics capability, the airline enjoyed several benefits. It was able to increase the number of calls analyzed to 10,000, or about 20 percent of the total, leading to much more reliable reports. Those reports are now generated automatically; before they had to be completed by hand.

The cost of analyzing calls has dropped as well. Continental has eliminated three full-time equivalent positions as a result of the speech analytics solution, eQuality CallMiner, jointly developed by Witness Systems and CallMiner.

Speech analytics tools like the eQuality CallMiner and many others available are starting to gain traction, analysts and technology providers agree, because the technology goes beyond simple audio mining to locate occurrences of spoken words or phrases in speech material. The speech analytics of today goes a step further in that it can include an analysis not only of specific words and phrases, but also of voice levels or other indicators of a callers mood or anxiety level.

Several Benefits
Firms that record voice conversations, most notably contact centers, are turning to speech-enabled data mining to deliver more comprehensive analytics to improve performance, help reduce the costs of operating the centers, and, in some instances, to help transition call centers from cost centers to profit centers through better marketing campaigns, cross-sells, and upsells.

DMG Consulting reported that speech analytics has already provided quantifiable benefits for early adopters, with systems paying for themselves in as little as nine months to a year. The analyst firm expected contact center speech analytics applications to grow by 120 percent in 2006 and to at least double in 2007 and again in 2008.

However, this growth comes on top of a very small base. DMG explains that the market has grown from 25 implementations in 2004 to 603 in 2006. That averages out to a 391 percent annual growth.

Research and development investments are driving investments in speech analytics technology, functionality, usability, and pricing, DMG added, while increasing competition is reducing the costs for contact centers and other adopters.

Industry research firm Gartner agreed. In its report, Unravel the Complexities of Call Center Speech Analytics, Gartner blamed the lack of earlier adoption of speech analytics technology on solution immaturity, competing technologies, investment costs, vendor viability, and a limited customer base. It did predict, though, that as the value propositions solidify, speech analytics would become a key part of customer service analytics portfolios.

The speech analytics technology itself has been available for several years, but was limited primarily to the government market until recently, according to several speech technology providers. The analytics applications are just now starting to get traction in the private sector as companies look for better ways to serve customers, reduce churn, and improve marketing campaigns. Fraud reduction is another use.

Speech analytics augments what companies can do with other analytic techniques because it enables more precise and more comprehensive collection of data from phone calls. Some systems also include analytics capabilities for other contact centers communications, such as the Web, or work with other systems that do. Some also integrate with various customer relationship management (CRM) systems.

For the last year and a half, companies have increasingly been bringing in speech analytics, says Cliff LaCoursiere, senior vice president of business development for CallMiner. Its gone from where we had to go out to find customers to now prospects are coming directly to us to see what we have to offer. Companies are becoming much more demanding. Categorizing people by other analytic tools is no longer good enough. The potential for integrating [speech analytics] with CRM technology is tremendous. I think that were at the beginning of a steep growth curve right now.

Call centers and other customers are seeking speech analytic solutions, LaCoursiere says, because categorizing calls by [more traditional] speech tools is not accurate enough.

Some call centers have been recording telephone conversations for several years to provide training for agents and to improve customer services, he continues. Recordings, particularly for financial services firms, help confirm transactions in the event of a dispute.

Whereas older speech technology tools might record conversations for later review by a contact center supervisor, speech analytics enables a company to capture speech in such a way that the details of the conversations can be mined for critical caller information, he says. One of the reasons that Continental Airlines can analyze so many more calls, for example, is that the calls are immediately flagged for specific words, phrases, or even caller stress. The technology converts speech to text to search for the necessary words and phrases.

Caller stress is recognized and flagged by the CallMiner technology when the callers tone or level of voice goes outside of normal level. Algorithms determine if the speech tone or speed changes outside of the norm, as some people talk much more slowly or quickly when they are angry, for example.

CallMiner customers can help determine their own definitions of the norm. Last summer, CallMiner launched a development toolkit that helps companies to customize their speech analytics applications. Called SpeechTools, the software lets users tweak the CallMiner Analytics Suite by simply tuning languages, terminology, or acoustic environments. For example, if a company launches a new product with a new name that is not in CallMiners dictionary, SpeechTools will let users add that product name and analyze customer responses to it immediately.

By mining calls for these semantics and caller satisfaction factors, a contact center can determine when a customer is becoming displeased enough with the level of service that company intervention is needed to prevent him from going to a competitor, says LaCoursiere, who adds that reducing customer churn is one of the major reasons that call centers are interested in this technology.

Steve Rutledge, vice president of product marketing for Genesys Communications Laboratories, agrees: Customer service still has a long way to go. It comes down to knowing what customers expect.

As a the central piece of the customer service puzzle, customers expect quick solutions when they call, and are becoming increasingly dissatisfied with IVR systems with lengthy menus. Adding analytic capabilities to the speech-enabled IVRs will help companies raise that customer service to a still higher level, Rutledge maintains.

Speech analytics tools better determine what customers expect because the technology is more telling than a call center agent or supervisor simply checking off a customers annoyance level on a report. LaCoursiere points out that the call center agent or supervisor who uses pull-down menus on a report completed on a computer screen is usually rushed, and will often indicate the first pull-down item, whether or not it correctly reflects the callers level of satisfaction.
Rutledge adds that companies can use the speech analytics information to augment and check on their CRM systems as well. He believes that the speech data can help determine if the CRM systems analysis of a call (why we think they called) actually coincides with why a customer did call. From that information, companies can tweak capabilities, staffing, and technologies to improve customer service.

Beyond customer service, speech analytics can also help a firm determine the effectiveness of its marketing campaigns and operations, says David Pennington, director of product management for Envision Telephony. For an outbound telemarketing campaign, for example, analysis of different questions or words by agents can help determine what syntax is the most effective, he says.

Yet speech analytics is still only a tool. By itself, it doesn't correct a companies issues, he argues. You can have all of the information in the world, but if you don't take action on that information, its worthless.

Customer service still has a long way to go, Rutledge agrees.

Speech analytics will become a standard solution for contact centers that are recording customer calls, many experts predict. Companies will use speech analytics to examine root cause analysis from the voice of the customer to drive enterprise performance.
They will also become easier to use and more accessible to key decision-makers within the organization, Pennington predicts, noting that this will improve the quality of information available, lead to more effective decision-making, and reduce time-to-decision, thereby improving enterprise responsiveness and customer experiences.

But the real power of speech analytics will be its ability to analyze and predict customer trends. By mining and analyzing hundreds, or even thousands, of calls, speech analytics technology automatically identifies the important information that helps managers predict what customers will buy, when they will buy, and how they will buy. By predicting this information managers can improve agent effectiveness, contact center performance, and enterprise performance. This will allow you to deliver legendary customer experiences that improve customer satisfaction and retention, Pennington says.
But even with the recent interest in speech analytics, theres still plenty of room for growth, says Mary , technology product marketing manager for SAS. Only 25 percent of call centers have some speech analytics technology in place. Were just at the beginning.

Page1 of 1