The nature of a company’s relationship with its customers has changed. The voice of the customer has a greater impact today than in the days of one-way communication. The rise of blogs and other online sources, such as You Tube and Facebook, have given customers new forums to air frustrations and overall experiences. Technological growth is a blessing to organizations that are beginning to mine the voice of the customer in order to better understand key business pains and improve everyday business decisions, surface potential escalations - before they end up on a blog or impact hundreds or thousands of other customers.


Speech analytics provides real value from customer interactions. Taking a proactive approach that goes beyond what is happening in contact centers and other customer-facing departments to why it’s happening can help businesses get ahead of the issues.


Unique and Common Business Pains


Speech analytics technology can be very flexible and is often applied to many of the unique “pain points” organizations experience in specific lines of business. That said, there also are common drivers, including the opportunities and challenges that industries and sectors share.


Examples can include:

  • Determining why customers call,
  • Improving first contact resolution,
  • Increasing sales effectiveness,
  • Reducing customer churn,
  • Implementing focused quality and
  • Managing partner/vendor relations.

To help organizations solve these issues, it is critical to determine time-to-insight and value-of-insight, especially in today’s fast-moving and competitive environment. This is where speech analytics can make a great impact - transforming rich structured and unstructured data into actionable intelligence.


Speech Analytics - The Early Days


For many years, businesses have had meaningful interactions with customers through the contact center. The challenge was to effectively address and leverage structured and unstructured data – some studies suggest that more than 80 percent of enterprise-wide data is unstructured – by having call center managers select, listen and analyze, at random, less than one percent of calls, while looking for keywords. Sometimes they would find patterns leading them to root causes, but sometimes they would not. Often, conclusions were found using incomplete information and guesswork.


Early speech analytics solutions relied primarily on a process called word spotting, which uses a search function comprised of key words/phrases to sort through recorded customer interactions. For example, if a manager wanted to find out why customers were cancelling their subscriptions, he/she might enter the key word “cancel” into the search box and receive a list of calls that matched the term. However, these word sets were limited, and early solutions rarely produced accurate, meaningful results. Individuals might describe the same situation using different words or phrases, such as cancel, cease, stop sending or unsubscribe.


Maximizing Unstructured Data – Phonetics, Speech-to-Text and Complete Semantic Indexing


Today’s most robust solutions have moved beyond simply word spotting. They leverage advances in technology to categorize conversations and provide automated root-cause analysis to users who are investigating key business issues. Such solutions not only help find problematic calls, but they explain why those calls exist. Three levels of speech analytics can be deployed to achieve this level of sophistication.

  1. Phonetics. The first step of processing speech data is a phonetics stage. Phonetics break audio into a string of sounds. (Phonemes are the smallest sound component parts of a language.) Some solutions only provide a phonetics index that matches queries against a string of sounds and returns audio files that match the “sound” query. The output of this initial stage alone can provide basic audio search capabilities, but it has limited accuracy and no linguistic context. Phonetic-only solutions do not leverage the conversational context to distinguish between similar sounding words, such as like “cancel” and “can sell.”
  2. Speech-to-text or LVCSR. To improve upon the limitations of phonetics, some solutions also use a speech-to-text process to put actual words behind the initial string of sounds. This process is known as large vocabulary continuous speech recognition (LVCSR). LVCSR translates phonemes into searchable text. These solutions are available with a full-language model of tens of thousands of predefined industry-specific terms that can analyze entire communications and the context in which words are spoken. This increases the accuracy of each transcribed word within specific contexts, and it is especially pertinent with regard to homonyms and words that have a similar sound. For example, clothes versus close or whether versus weather are words whose different meanings can be significant in pinpointing trends and cause/effect relationships.
  3. Complete semantic indexing. Although transcriptions can improve the accuracy or results and add context, they do not automatically surface all the potential insight hidden in mass volumes of customer calls. In order to do this, you need a complete semantic index that combines all the output of the phonetics stage, the transcriptions of calls with relevant statistical and metadata, into a single index that can automatically surface how customers express themselves and why. A complete semantic index increases call-categorization accuracy and can also automatically surface the root cause of customer interactions, completely unguided by the category creator.


How to Use Speech Analytics


Business users that leverage speech analytics typically use the technology to categorize interactions, drill down, replay key calls, derive root-cause analysis and combine with traditional data mining capabilities to extract actionable intelligence.


Categorizing for Faster Intelligence


Categorizing calls – such as billing issues, product feedback or repeat calls – based on call content provides quantitative information (such as an increase in customer complaints about late shipments) as well as qualitative information (such as a set of calls in which customers state that they are switching vendors because of a complex product feature).


Categorization automates analytic outputs that would otherwise require a staff of trained analysts to individually listen to hundreds of thousands of calls. Armed with reports that display category-related information and analysis, business users can identify the success of a new product or service offering that was recently launched. A category report may reveal a high level of customer complaint calls, and playing back even a sample of these contacts may elaborate on what customers would like to see changed.


Part of the challenge of any speech analytics solution is knowing in advance how customers and agents express themselves in order to accurately define effective categories. This again is where a complete semantic index can help eliminate the tedious guess work by automatically suggesting the exact expressions used in specific environments- including slang and unique expressions.


Drilling Down and Replaying Key Calls


By drilling down and replaying customer interaction recordings, mangers can uncover even more specific customer feedback, agent behavior and process issues. Rich audio/visual playback environments create a visual map of the conversation, automatically highlighting phrases of interest and using call visualization techniques focused on segments that include emotional expressions. Users can quickly navigate to the area of interest within the call by clicking anywhere within the synchronized call transcription or on any highlighted keyword or phrase. This saves valuable time reviewing the call and allows users to focus attention on the most actionable elements of the interaction.


Leveraging Automated Root Cause Analysis


A lot goes on in contact centers from day to day. Themes and call drivers emerge, but the key is to get beyond what is happening, so managers can hone in on why and take action. Solutions that leverage a complete semantic index help organizations accomplish just that by introducing root-cause analysis through a “tell-me-why” feature, for example.


With a complete semantic index, solutions can begin to reveal unknown issues and provide greater business value. For every call set – whether it’s a category of billing-related calls or calls from a specific day when call volume peaked – automated root causes can surface the key drivers of those interactions. A complete semantic index compares content of the subset of calls to overall call content, surfacing unique terms/phrases that appear within the call set. Some solutions cluster these terms into top root cause groups, prioritizing the subsets according to the volume of calls each subset represents. This can help identify unknown issues and focus users’ attention to call drivers that have the most impact on their businesses.


Many contact centers struggle with first contact resolution. Speech analytics can help by mining the root-cause of unresolved calls, identifying terms that appear more frequently in the unresolved calls than in the resolved calls.


For example, a Fortune 500 insurance provider with more than four million customers had first contact resolution rates of 60 percent – significantly short of expectations. While the company was aware of the issue, management lacked clear insight into why customer queries were not settled on the first call. The automated root cause identified terms indicating that back-office delays were causing issues with the mailing of claim forms and checks. When armed with reliable knowledge about the cause of repeat requests, management is able to modify policies and train agents to better understand them.


Users can apply root cause analytics capabilities to all emotionally charged calls, as well. Beyond getting a simple list of interactions that contain expressed customer dissatisfaction and emotion, root- cause analytics may find that emotional calls typically mention terms such as “did not show,” “technician,” “never called” and “waiting all day.” This can help point to recurring customer dissatisfaction and problems associated with technicians failing to show up on time, or at all, for installation appointments.


Addressing and fixing the why problems rather than the what problems enables companies to experience the benefits of improved customer satisfaction, streamlined processes and reduced business costs. The ability to uncover why is the greatest value that is delivered from a speech analytics solution, and organizations are increasingly leveraging these capabilities.


Integrating with Data Analytics


Some speech analytics solutions are integrated with data analytics to build relationships between spoken audio and call metadata to uncover hidden trends and identify scenarios that impact these call types. Such information pinpoints both positive and negative trends as well as areas of improvement.


A business may know that customers are defecting, but not understand why. A unified speech and data analytics solution can mine customer retention calls, as detected by speech analytics, to identify and suggest the root cause of customer churn. It may show for example, that midmonth calls mentioning competition are 33 percent more likely to be a customer closing their account. Working with independent applications, analysts may not otherwise uncover these significant correlations and trends.


Speech Analytics and Actionable Intelligence Yield Measurable Results


Speech analytics has been called one of the hottest technology areas in quality/performance solutions for contact centers and broader enterprises. It’s transformational – enabling executives to address broader quality and performance optimization of the entire customer service process. It’s redefining quality monitoring. Driving quality monitoring activity based on business issues, companies can attain more effective quality insight with fewer evaluations. It helps reduce call volumes and optimizes processes. And it can deliver a high ROI.


Sure, technology has created a two-way communication with customers that requires carefully managing their experiences, because it can come back to haunt an organization. At the same time, companies can now leverage powerful solutions - such as speech analytics - to instantly analyze millions of interactions for better insight and understanding into customers’ wants and needs. They enable businesses to use information better to meet expectations, drive product development, improve sales, impact marketing and create unique, competitive advantages - all based on what customers are saying.


Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access