Nov 2015

The Age of Speech Analytics Is Close at Hand

posted by Bumicom private_bank

The “Johnny-come-lately” in contact centers is poised to become “Johnny-on-the-spot”

By Leonard Klie
Oct 2015
CRM Magazine

In the world of contact center software, speech analytics is still very much a Johnny-come-lately. While call recording is practically ubiquitous, speech analytics has often been seen as a nice add-on that might uncover something interesting from time to time. Its true value has often been lost in a jumble of marketing buzzwords, confusing technologies with not-so-subtle differences, and grand expectations that were nearly impossible to meet early on.

“Contact centers have not been quick to adopt [analytics],” observes Donna Fluss, founder and president of DMG Consulting. “We have the technology, but it’s only slowly being rolled out.”

The latest research from both DMG Consulting and ContactBabel puts the number of contact centers currently using interaction analytics at just slightly less than 20 percent. The good news for the industry is that of the more than 80 percent of contact centers that do not have analytics in place, 25 percent have plans for implementation in the not-too-distant future, according to ContactBabel’s research.

“Analytics in general are starting to gain traction across a wide swath of the contact center market,” says Larry Skowronek, senior vice president of product at Nexidia, a provider of the technology. “Call recording is essential to the contact center, and speech analytics is moving in that direction.”

Analyst firm MicroMarkets earlier this year estimated the North American speech analytics market at $233.2 million in 2014, but the firm expects it to reach $614.1 million in 2019, at a compounded annual growth rate of 21.4 percent.

Among other reasons, MicroMarkets credits the rising number of U.S. contact center seats for the growth.

The U.S. contact center industry added 20,499 jobs between April 1 and June 30, following the previous quarter’s gain of 7,965 new jobs, according to jobs4america. Through 2014, the industry saw the creation of 50,000 new jobs in the United States.

Other drivers for the growth in contact center analytics include new implementation areas, rising demand for cloud analytics and risk management solutions, and the need for real-time and predictive capabilities. A growing emphasis on customer service is also at the heart of the adoption curve.

“The older generations put up with poor customer service,” Fluss says. “Younger consumers will not stand for it. They expect and demand a better level of service” that demands immediacy.

With those elements in place, analytics is finally starting, after more than a decade, to take its rightful place at the contact center table, and vendors of contact center analytics have met this growing interest with investments in research and development. And one of the biggest investment areas is real-time (or as close to real-time as possible) analytics.

That’s the case with Verint Systems, where Alain Stephan, global vice president of customer analytics, notes that “speed to insight is something that [the company] is constantly working on, and not just with our speech analytics, but with text and desktop analytics too.”

Verint has already gone through some pilot programs of its real-time analytics solution and is now moving those customers into deployment, according to Stephan. Many other vendors are at the same stage.

In fact, Nexidia’s Skowronek expects real-time speech analytics to become mainstream within the next five years, though it could happen a lot sooner. “The kind of interest that we’re seeing from customers and prospects tells me that it could really pick up steam in the next year or two,” he says.

This year, Fluss adds, is shaping up to be “the pivotal year” for real-time analytics in the contact center. “Companies are now starting to see the need to reduce customer effort, and that can best be done with real-time analytics.”

“For some businesses, real-time [analytics] is an important and growing part of the armory they have to improve their efficiency and effectiveness,” Steve Morrell, founder and principal analyst at ContactBabel, wrote in his company’s Inner Circle Guide to Customer Contact Analytics.


Real-time analytics is essentially a merging of several technologies. It starts with speech analytics capable of scanning for predefined keywords and phrases; instances of talk-over; changes in pitch, tone, or talking speed; and even sentiments or context, while the call is happening. Then, some sort of artificial intelligence or decisioning engine determines what activity to take once a predefined marker has been identified. The application might, for example, alert an agent, supervisor, or retention team member; offer guidance to agents; escalate calls to second-tier support agents; or even trigger back-office processes. Then there needs to be some sort of reporting to close the loop.

To be most effective, real-time speech analytics should be able to not just look at the current call but also consider data from previous calls to spot overall patterns and trends, allowing the business to change course as needs dictate, Stephan explains.

“It’s about bringing context into the conversation to deliver the next-best-action recommendation,” he says. “Real-time analytics goes beyond just an analysis of the interaction, beyond the individual customer or call to the wider business. It starts with the interaction, but you need to link it to something much broader. Then you can be more nimble as you make decisions.”

“Real-time analytics allows the business to take systematic action while the call is still in progress,” Skowronek adds. That could mean preventing churn or ensuring that agents are complying with government regulations and company standards.

Under the current federal Fair Debt Collection Practices Act, for example, collection agencies making outbound calls to debtors now have to advise them that they’re calling about an outstanding debt and that any information obtained could be used for that purpose. This warning is referred to as the “mini Miranda” because it’s similar to the Miranda rights that law enforcement must issue to suspects prior to starting an interrogation. Real-time speech analytics can be used for quality assurance purposes to ensure that agents are issuing the appropriate advisories.

“There’s a lot of interest in [real-time analytics] from a risk and compliance perspective,” Stephan says.

It’s not just the financial services industry that stands to gain from real-time analytics, though. Retailers and subscription-based service providers—businesses where customer churn is a real concern—could also benefit from real-time capabilities. Other key target verticals include utilities, government agencies, emergency responders, and healthcare—all businesses where the ability to respond to situations quickly is imperative. The size of the firm doesn’t matter either, experts agree.

“If you know what’s happening in real time, you can alter the outcome in real time,” Fluss says.

Among other benefits, real-time analytics will allow companies to identify language indicative of a disgruntled customer and respond with relevant retention offers, all within the same call. That is something that could potentially have huge financial benefits for all businesses.

So, too, could the ability to identify caller intent and move frequent questions out of the contact center into automated self-service, says Nick Gyles, chief technology officer at WDS, a contact center outsourcing firm acquired by Xerox in 2012.

Real-time analytics can also improve call efficiency and efficacy by providing real-time guidance to only those agents who need the extra help resulting in less training overall. Errors or missing information that lead to costly repeat interactions can be flagged for immediate action.

“It’s about being much more responsive to the customer and still having much more control over agents’ behavior,” Gyles says.

As an added benefit, “with real-time [analytics], you can take the appropriate actions for each individual caller,” Fluss says. “You can customize each interaction, and when you can provide a more customized experience, it’s better for everyone.

“It’s not just about getting fancy, innovative new technology. It’s about improving productivity, increasing agent satisfaction, and reducing operating costs,” Fluss continues. “Real-time [analytics] is an important step in providing better customer service.”


But for all its potential, real-time analytics in the contact center is still extremely rare. Of the 20 percent of contact centers that have deployed analytics, most (77 percent) use historical, post-call analytics; real-time or near-real-time speech analytics is used by just 38 percent, ContactBabel reported in its 2015 U.S. Contact Center Decision-Makers’ Guide.

Post-call solutions that analyze recorded conversations are more mature by far than real-time applications, which only started to enter the market in 2011, according to Fluss.

Part of the reason for the slow uptick in real-time analytics, according to Jim Davies, an analyst at Gartner, has been a lack of organizational understanding and preparation.

And there’s this: Achieving full real-time capabilities is not easy. Typically, a slight few-second delay is necessary for the application to collect and analyze a large enough snippet of the conversation to identify the triggers that require action.

Gyles says true real-time capabilities are still at least 12 to 18 months away—which does little to help the current situation. “The industry is not at the level we need right now,” he says. “We can use the technology for keyword spotting, but we still have to drill down on our own. [Current solutions] do not give us the level of real-time insights we need.”

That was the motivation for WDS to develop its own Agent IQ system. Agent IQ provides a real-time view of what is happening in the 175 contact centers WDS operates around the world. Agent IQ can interpret natural language, associate relevant root causes, and provide real-time guidance. Automated call logging captures data. Real-time, evidence-based self-learning draws on agent usage to optimize results against the latest call drivers. The self-learning system works with live chat, voice, and text interactions.

But even with Agent IQ, as with most other solutions, capabilities are limited. “We need to analyze calls with more detail,” Gyles states. “There’s a lot of work to do to get us there.”

Gyles is certainly not alone in his frustration.

“In most cases, agents are receiving a very limited amount of real-time data, which is wrong,” Fluss says. “The contact center is a real-time operation, yet, when it comes to information, a lot of it is not real time; it’s historical.”

That’s not to say that post-call analytics should be phased out. Just the opposite: Post-call speech analytics is invaluable for identifying root causes and emerging trends from the recorded audio that most contact centers gather.

“You need post-call and real-time analytics together. Real-time is most effective when it’s coupled with post-call, and vice versa,” Skowronek explains.

Furthermore, experts explicitly warn against starting down the path toward real-time speech analytics without first collecting post-call information.

“Get a baseline understanding of what is happening [across the business] with post-call analytics first instead of starting with real-time analytics right away,” Stephan suggests.

Skowronek agrees. “You can’t do real-time analytics if you don’t have post-call analytics to shape the types of things you need to look for and take action on,” he says. “Post-call analytics are needed to point out the things that you should be focusing on in real time.”


Like most major CRM technology introductions, greater adoption of real-time analytics will require a major attitude shift, not to mention a serious financial commitment, Fluss points out.

“Real-time analytics have not yet taken off because the core systems that most companies have in place weren’t designed for it. There are definitely architectural issues, and so it can be expensive to bring new technologies on board,” Fluss says.

“Companies need to rethink their entire service strategies,” she continues. “They have to change the mind-sets, processes, and company cultures, which probably means retraining everyone.”

Despite technological advances, real-time analytics will require significant processing power. And the current software can be expensive and require extensive and ongoing tuning to remain effective. Yet for all that, making the switch to real-time analytics “is difficult, but not impossible,” Fluss says.

And for businesses with cloud-based contact centers, it might be a little easier. “With real-time analytics, you need to tap into the voice stream, and with an IP communications infrastructure, that is easier to do,” Skowronek says.


Skowronek and others caution against jumping into real-time analytics without being mature in your understanding of analytics in general first.

“It’s important not to get carried away with overdoing real-time analytics,” Morrell warns. “There is a danger that businesses can get too enthusiastic and set alert thresholds far too low, resulting in agents being so constantly bombarded with cross-selling and upselling offers or warnings about customer sentiment or their own communication style that it becomes a distraction rather than a help.”

Agents already are likely to have several applications running and several screens open at once, and the last thing they need is more windows popping up all the time.

“You want the ability to look at things in context so you’re not reacting needlessly,” Stephan says. “You want to be able to manage and react to patterns and larger trends rather than the outliers.”

A tempered approach can also help ease the transition for agents and give them a measure of control, Gyles says. “You want to guide, not dictate to, the agents,” he says. “They still like some autonomy in the process.”

And even when a supervisor needs to intervene, the action shouldn’t be jarring for the customer. “You want it to be seamless from the customer’s perspective. You want it to augment the customer experience,” Verint’s Stephan suggests. “Operationally, you don’t want calls to come to a screeching halt unless they really need to.”

Developing a concrete business plan ahead of time is also critical. “Fit the solution to a specific business need,” Stephan says.

Skowronek, likewise, advocates forming clear goals, such as increasing compliance or reducing churn. “You don’t want to deploy [real-time analytics] just because you can. You want to have a real use case in mind.”

Wider business changes will probably be necessary as well. “To benefit from real-time operations, you need to have a real-time [company] culture and a real-time organization,” Stephan contends. “Culturally, the organization has to be dynamic or the return on investment will not be there.”

And then the organization needs to empower agents and supervisors to do something with the information the analytics software uncovers. “The agent needs to be able to use the data to deliver the best service to the customer,” Stephan adds.

The technology still has changes to undergo. “Real-time monitoring and alert functionality is available now, but significant efforts still need to be made to reduce the amount of time required to process a trigger or event, cutting the reaction time from a handful of seconds to getting sub-second response times,” Morrell points out.

Vendors would also do well to make their solutions simpler to use, he adds.

Money will be an issue, and experts caution that even with the best solutions, adoption will continue to lag if only large organizations can afford them and the staff needed to maintain them. Innovative solutions that allow businesses of all sizes to analyze more interactions and share the insight with more people across the company will be key to the technology’s future.

Speech analytics will also need to move beyond isolated channels and perform a more outcome-based analysis that informs improvements around best practices, behavior modeling, and customer preferences. Analytics teams will need to become cross-functional; capabilities will need to reach outside the contact center to enhance the entire business.

Taking contact center analytics to the next level will require work, but the effort will not be wasted. Experts suggest that by keeping customers at the heart of any rollout of real-time analytics and embracing the abundance of insight that such solutions can deliver, companies will be able to accomplish great things.

Senior News Editor Leonard Klie can be reached at lklie@infotoday.com.