March 2010

Speech Analytics: The New Frontier of Call Center Performance Improvement

Alan Mattei & Pablo de los Santos

Call centers represent one of the greatest missed opportunities in the retail and service industries. Over and again, companies launch major call center initiatives in areas such as problem resolution and retention, cross-sell and customer acquisition, only to see disappointing results and inadequate explanations as to why.

It is not for lack of trying. Realizing the need to improve the individual success of telephone representatives, call centers have adopted a suite of tactics that include training, coaching, incentives and various kinds of computer-based sales tools. High employee turnover in call centers only reinforces the need for these basic measures.

Yet even as the urgency for performance improvement rises, there is little clear guidance on the measures that will be most effective. Management remains stuck in a murky world of anecdotes, intuition and corporate mythology. Beyond simple productivity metrics, there is little analytical feedback on the conversational success factors that drive superior results in sales, service and customer satisfaction.

Most call centers have little systematic information about caller profiles; can’t pinpoint the techniques that are most effective in closing sales and resolving problems; and don’t compile or analyze information about the nature of customer objections or indications of satisfaction. This is a lost opportunity for the parent company as well, given that the call center is a potential source of rich insight into effective customer interaction across the enterprise.

To be sure, there have been significant advances in speech recognition software in recent years, which is a technological achievement unto itself. Yet tabulations of such data have proved of limited value in the call center, providing better descriptions of conversational highlights but not as much insight as hoped into the factors that cause various outcomes.

This picture is changing with the introduction of "speech analytics," which statistically examine call center transcripts to identify the verbal patterns most closely associated with key outcomes, both successes and failures. When married with a feedback system for call center reps, speech analytics provide a basis for continuous improvement in conversational techniques.

Increasingly in modern research-based marketing, companies compile and analyze the results from various customer campaigns; identify the techniques that work best; and incorporate findings in a new wave of initiatives that again can be analyzed. Through speech analytics, this test-and-learn approach can be replicated in the call center. In turn, analytically distilled insights become a knowledge bank for the overall staff, helping to lift the productivity of average representatives even though they may not be among the ranks of the innately gifted superstars in sales and service.

Pressure Cooker

While call centers long have been characterized by operational intensity, it is safe to say that the performance challenge has greatly intensified in recent years.

One reason is that much of the traditional workload of handling information requests and facilitating transactions has been replaced by a much more complicated set of responsibilities. In an era of complex products and services, shrinking in-store service and skyrocketing online purchase volume, customers increasingly turn to the call center for personalized support on major transactions and major issues.

In U.S. financial services call centers, including those of banks and credit card companies, for example, the proportion of customer contacts devoted to information requests fell by more than half, to less than 20%, between 2000 and 2006, according to Novantas research. Meanwhile, the contact proportion devoted to account service and problem-solving roughly doubled, to more than 55% of total volume.

The removal of basic servicing to pure online delivery primarily leaves complex customer requests in the hands of live representatives. When pertaining to sales, these requests are higher in value and better lend themselves to cross-sell and up-sell than the previous transaction mix. When pertaining to service and problem resolution, these requests are higher in risk and carry much higher potential for customer dissatisfaction and defection. As result, the manner in which each call is handled has a much greater impact on customer lifetime relationship value.

A second driver of rising call center intensity is the growing need for tailored responsiveness. Recent research by Novantas and Informa Research Services, for example, revealed dramatic differences in the attitudes and behaviors of checking account customers. Some people are highly receptive to fee-based services that provide more account information and control. Other account holders are resolute economizers but may be interested in savings-related offerings. Still others overspend their accounts habitually and use the call center sort of like an emergency payments facility.

When applied across tens of thousands of calls, these patterns routinely defeat management attempts at generic coaching for call center representatives. The sheer spontaneity of human interaction assures a constant stream of call center experimentation as individual representatives converse with customers.

But absent an ability to glean insights into what does and does not work, performance improvement defaults to familiar techniques — promoting standards for courtesy, empathy, accuracy and quick responsiveness; disseminating valuable tips and anecdotes culled from the crème de la crème of call center representatives; and episodic call monitoring and coaching.

A third driver of call center intensity is what might be termed the "cross-roads effect" of the first two factors. With call centers carrying a rising responsibility to handle complex transactions and also deliver segment-based responsiveness, they naturally get dragged right into the middle of emerging issues in various industries.

In U.S. financial services, for example, new laws and regulations are having a profound effect on the credit card business and the checking business. In a transition era when various services, fees, terms and conditions are changing rapidly, call centers are being flooded with customer inquiries about their accounts. They also are expected to play a prominent role in proactive outreach to account holders, both to inform customers about changes in products and services and to help in selling new generations of offerings that will help the industry to rebuild disrupted revenue streams.

While these examples stem from U.S. financial services, the core performance issues resonate across industries and across the globe. Generally speaking, the call center industry is well past the tipping point where operational excellence is the defining factor in performance. The new dynamic is all about excellence in customer interaction, a much more nuanced challenge that requires detailed knowledge of conversational dynamics between call center representatives and customers.

Points of Entry

One of the difficulties in improving customer interaction is establishing clear priorities. Because the call center handles so many important responsibilities in a fast-paced environment — problem resolution; sales; information and referrals; transaction fulfillment — distinctions begin to blur and management almost inevitably gravitates to sweeping initiatives, typically under the banners of service and productivity.

The problem is compounded by the prevalent reliance on shop-floor performance metrics. Efficiency campaigns, for example, typically focus on narrowing call times, reducing headcount and speeding transaction processing, with few adjustments for revenue considerations such as sales conversion ratios on inbound calls, balance retention achieved by averting account closure, and lead generation.

From this perspective, the success of technology-based performance initiatives largely hinges on the organizing management principles. There’s a conspicuous risk of generating too much information with too little basis for specific action.

In setting up projects with speech analytics, there are two main approaches that provide a focused point of entry. One is to focus on a high-visibility "burning issue" that is already known to management, and use the knowledge gleaned in initial problem-solving as a basis to investigate and improve overall customer interaction.

By contrast, a diagnostic approach begins by identifying exceptional highs and lows across a variety of performance dimensions, such as sales conversion ratios, customer complaints, call-handling time and crisis-mode account retention.

After modeling the lift from repeating certain successes more broadly or from minimizing poor outcomes, management can then select a short list of priority items for improvement, starting with a methodical investigation of the root causes of performance dispersion.

Burning Issues. In more than a few instances, companies already know the burning issues in their call centers that require immediate attention. Again drawing on U.S. financial services, for example, we can immediately name three situations where the need to improve customer interaction is clear:

• In the checking industry, changing laws and regulations have overturned the practices that banks have used to handle account overdrafts, introducing a new set of dynamics in serving customers and also disrupting an important revenue stream. Call centers will play a major role in working through this crisis, in areas such as fielding customer inquiries, handling overdrafts in a new manner while keeping customers satisfied, and also presenting new products and services that will provide additional checking account value to customers and also help to rebuild fee revenues.
• In the credit card industry, call centers will play a prominent role in direct communications with customers about legislative- and regulatory-driven changes in terms, conditions and fees on accounts. Additionally, following the deep U.S. recession, call centers are a focal point in nuanced collections activities, and they also will play a major role in rising industry competition to acquire and retain the most valuable accounts.
• In the insurance industry, call centers provide an essential touch point for online policy sales. But market changes and internal performance issues are beginning to throw auto insurance call centers into defensive mode. Mega-dollar ad campaigns are losing some of their former punch in attracting customer inquiries. Pricing is razor thin in an increasingly crowded online market. Meanwhile, phone representatives are failing to convert more than two-thirds of qualified customer inquiries into completed policy originations. The pressure is compounded by the recession-driven slump in overall demand, with drivers buying fewer new cars and also cutting back the extent of coverage.

Diagnostic Approach. Often call center managers need to evaluate the overall functioning of their units. Examples include:

• Where the call center has been unable to improve customer satisfaction;
• Where there has been a change in management;
• Where two units are being combined;
• Where performance has hit a plateau;
• Where staff turnover has hurt performance;
• Where the performance skew among representatives has become intolerable;
• Where there is difficulty in meeting corporate targets for sales revenue growth;
• Where there has been in a shift in the unit’s responsibilities and/or assigned product set;
• Where competitor actions have shifted customer perceptions and attitudes; or
• Where there is a need to monitor and improve offshore operations.

In such situations, management can begin by analytically identifying "pressure points" where basic results are out of kilter. Large numbers of callers may be getting placed on hold in one type of activity, for example, while an excess of call transfers is plaguing another area. In other instances, a certain activity may be generating the largest proportion of customer complaints, while yet another area is experiencing a high volume of repeat calls from the same customers.

From this short list of pressure points, management can then establish specific priorities for call center performance improvement.

Voice of the Customer

A key goal of speech analytics is comprehensive listening, such that patterns of customer verbal responses are methodically gleaned from thousands of conversations. Once management has established the category of calls to be analyzed, then the race is on to identify and statistically correlate conversational phrases with key outcomes:

• Looking at the pressure point of sales conversion, what are the key phrases that customers use in indicating their receptivity or objections to a certain product, and what verbal patterns best explain the difference between the most and least successful call center representatives in closing transactions?
• Looking at the pressure point of call handling time, what are the major causes of delay (verifying identification, for example), what groups of customers are most prone to this type of problem, and what actions on the part of call center representatives are most and least effective for each of the major groups?
• Looking at the pressure point of customer retention, what are the early indicators of customer propensity to defect, and how can those insights be married with proactive strategies that will lessen the need for crisis-mode retention? What concessions are most and least effective in persuading people to maintain their accounts, and how do these patterns vary by customer group and product?

With this type of in-depth analysis in hand, management is positioned to effectively apply traditional tools such as training, technology and performance incentives, as well as the advanced tools permitted by speech analytics, including call guides and enhanced monitoring.

For each major type of call, it is possible to develop succinct guides that identify the probing questions that are most effective in improving outcomes, the most effective retention techniques, and the most effective techniques to control handling time. It is also possible to identify the situations where segment-specific responses are warranted, and the particular types of responses that best resonate for the segment in question.

Speech analytics also permit a level of all-inclusive call monitoring that is conducive to staff performance measurement and performance improvement programs based on continuous feedback, experimentation, and rapid-cycle testing.

Within the call center, each major type of activity has a different group of top-performing representatives. From an aptitude perspective, it is helpful to be able to pinpoint the types of activities where various individual representatives are most productive. Also there will be a different set of conversational success factors for each major type of activity, and speech analytics can help to identify the practices of top-tier representatives that will be broadly helpful to the rest of the staff. A further nuance is tracking the degree to which representatives are following the scripts and techniques that have been recommended to them.

Among large organizations, call centers carry heavy workloads that range up to millions of interactions daily. Until recently, however, management has been unable to analyze even 1% of call center volume. Through speech analytics, by contrast, management is now positioned to not only capture but analyze 100% of call volume.

Such progress will provide new performance metrics that set the stage for continuous performance improvement within individual call center units. And from a broad perspective, speech analytics promise to transform the call center into a larger corporate asset that will help in improving customer interaction across the enterprise.

Alan Mattei is a Partner in the New York Office of Novantas LLC, a management consultancy, and Pablo de los Santos is a Director in the Madrid office of Novantas España.