March 2003

Speech Recognition: Making Training Match Reality Through Simulations

By Jeff R. Carpenter, Sivox Technologies, Inc.

Every day, each of us is inundated with sales calls, brochures, e-mail blasts and articles that all promise to cure the ills of agent training with dramatic results. After sifting through the piles, you're probably left with the same questions most of us are left with: What will really help me recruit, train and retain valuable employees?

Over the past few years, technology has dramatically changed the way in which contact center agents are prepared to handle customer calls. Yet this creates its own dilemma. While training departments have been in a perpetual race to keep up with these advancements, they often have little or no experience with implementing the various technologies. As a result, they wind up falling short of the return on investment (ROI) promises. Then the next best thing arrives on the scene, promising to train their agents better, faster and cheaper ' only this time they don't need to 'pick any two.' Eventually you start wondering, 'How did I get myself into this?' and then start counting the days to retirement. For those of us who are still counting, and trying to make sense of it all, we need to take a moment and review the past to get a clear vision of the future.

Applying Past Theories To The Present
It is important to note that there have been no incredibly dramatic new advancements in adult learning theory over the past few years. Most of the new theorists have built upon the discoveries of past theorists, adding a sprinkle of techno-babble to make it sound new and exciting. The challenge facing most training organizations is how to apply the theories of the past with the learning environments of the present and near future. While learning hasn't changed, the environment and methods by which you present the learning have changed dramatically. Still, the goals have always remained the same: to more quickly assimilate the employee to the work environment and to decrease the time to optimal performance.

The last 10 to 20 years have seen some interesting 'advancements' in the training of contact center agents. From interactive video tapes to laser discs to the advent of computer-based training (CBT), these advancements were all considered revolutionary during their time. Even though they may bring back painful memories, it was the start of providing a better multisensory learning environment. The evolutionary path for training and technology has been to create an environment that allows the users not just to regurgitate knowledge, but instead apply that knowledge in a realistic work environment. In the 1960s, educators and theorists, revolutionary at the time, wanted to create a learning environment where, 'You learn them (concepts) because your environment is organized in such a way that it permits or encourages or insists that you learn them.' (Teaching as a Subversive Activity, Postman and Weingartner).

Applying Learning To Real Situations
Something that continues to be the brass ring for most training organizations was a concept originally meant for grade school children. There have been many theories since then, ranging from self-paced training to distance learning, but the underpinning theory remains the same. To prove learning has taken place, you must be able to apply that learning in a setting that is as close to the actual work environment as possible. So how do you create that environment?

As technology has advanced, it has now become more possible, and economical, to provide agents with the ability to practice in a realistic environment without negatively affecting the current or potential customer base. Simulation and speech recognition have entered the scene, and promise to be the next advancement in agent training. Speech recognition allows agents to practice with the tool they use most every day: their voice. Simulation gives the training organization the ability to allow agents to practice in simulated production systems to ensure service order accuracy.

This becomes more critical as organizations use the contact center as their sole source of interacting with their customers. An article in the October 2002 issue of Fast Company stated that, ' 'call centers are one of the country's fastest growing industries: three percent of all Americans now work for a call center, and Gartner Research predicts that the industry could nearly double in size by 2010.' These agents are often the only 'face' of the organization with which customers will ever interact. Each customer contact is your chance to win or lose that customer's mind share and your market share. The often forgotten and maligned agents are your company's sales force, customer relationship management team and public relations group all in one. For most companies, the only place to distinguish themselves from the competition is on the contact center floor.

Types Of Simulations 
Much has been debated about how to incorporate the different styles of learning into agent preparation. In a landmark study, the National Training Laboratory found that agents retain 90 percent of what they have learned when they are able to work with a simulated real-world experience (see Figure 1). This study suggests that agents learn best when they are able to integrate and incorporate the material into the context in which they will use and apply that knowledge. With that in mind, many learners are still left holding the bag when it comes to applying those concepts in the real work environment. Not being given the ability to practice those newly gained skills in a safe environment prior to interfacing with the customer causes agents to revert back to known, comfortable behaviors. This reversion is contrary to what is needed to move the organization in a positive direction.

Figure 1

Organizations are currently using different types of simulation to better prepare their agents for the complex customer interactions. Simulation allows the organization to give agents the ability to demonstrate mastery, enhance skills or identify potential areas of opportunity. Every organization is using some form of simulation, from role-playing to immersion environments. Each has its benefits and weaknesses, but which one gets better results?

Role-Playing
Role-playing allows users to practice learned customer and system interactions and generate real-time feedback from all participants, making this form of evaluation and practice valuable in the development cycle of a learner. The drawback to this method of simulation is that the role-playing is only as strong as its weakest participant, or as effective as the level of seriousness the learner places on this activity. Additionally, much of the potential benefit may be lost due to an organization's inability to control that environment, to identify and verify that learning objectives are being met, and that appropriate remediation, feedback and coaching are provided at the point of failure. Finally, only one of the partners in the role play is practicing actual skills at any given point, an inefficient use of time that delays getting both agents up to speed.

Voice Detection
Voice detection combined with system simulation has all of the strengths of a standard role-play, while enhancing the ability to define learning outcomes. Voice detection listens for breaks in silence and records the agent's response to predefined customer recordings. This type of simulation allows the learners to be presented with customer situations that must be resolved by the agent. Voice detection allows the agents to record and review their audio later for voice quality attributes separate from the simulation. The drawback to voice detection is similar to standard role-playing; in fact, it can be considered a step back in training the agent. What you gain in the ability to have predefined learning outcomes and review of audio for voice quality is greatly overshadowed by the simulations' inability to provide immediate feedback and coaching at the point of failure, prior to any unwanted behaviors taking hold. 

Speech Recognition
Speech recognition technology, when combined with system simulation, is the most promising of the new trends in simulation. Speech recognition listens to the agent's response to a customer request and identifies whether the appropriate words or phrases are spoken. If the agent does not communicate appropriately, the system provides immediate feedback based on predefined organizational or regulatory requirements. This becomes a very powerful tool for many large organizations that want to ensure a consistent customer interaction or are required by federal or state regulators to use specific verbiage. 

Many fear that preselecting words or phrases will either take away the agent's ability to creatively problem solve or cause the agent to become robotic with the customer. Those fears can be put to rest by selecting the most important concepts for each dialog with the customer. There are many ways to say the right thing when interacting with the customer, but each of them should contain the same overall concepts. The drawback of speech recognition is the inability to judge voice quality during the simulation. This drawback can be overcome by recording the simulated call and allowing the learner and coach to review it. They can then make appropriate adjustments regarding voice quality and professionalism. This also allows learners to gain a deeper understanding of their own voice quality behaviors by performing self-analysis.

Much has been debated as to whether speech recognition is ready for primetime applications within the learning environment. Speech recognition has seen marked advancements in the past few years, allowing it to become a more viable solution for contact centers. With speaker-independent recognition rate percentages in the mid- to upper 90s, vocabularies that are in excess of one million words and multiple language support, it becomes a valuable tool in agent screening, preparation and remediation. With that being said, the e-learning industry is still trying to find ways to integrate it into their current product suites with limited success. Others are trying to downplay its importance in the evolution of simulation-based training. 

Giving Agents Immediate Feedback
The goal of all simulation-based training is to create an environment that is as close to the real work environment as possible, blurring the line between training and the production floor. This cannot be achieved without interpreting, comprehending and reacting to the agent's verbal and nonverbal inputs just as a real customer does every day. Allowing agents to see what happens when they give the customer incorrect information, are rude, hide behind policy, place customers on hold for too long or are nervous (just to name a few) helps agents gain the experience necessary to be successful immediately out of training. By using a simulated environment, agents will have already faced situations that usually would take months of customer calls, coaching and remediation to learn how to handle 'all before they leave the training environment. Agents would no longer be expected to memorize abstract concepts and then formulate their own conclusion of how to apply those concepts. Simulation training also allows learners to demonstrate that they know how to perform a required task resulting in the required outcomes from performing that task. This only becomes possible when the agents' entire environment is simulated and reacting to their inputs.

Next Steps For Simulations
The next big step for simulation-based training will be to allow branching through the use of agent verbal responses, emotional detection and randomized events. This will allow any speech-driven system to react to the agent's verbal inputs. Is the agent unsure of the answer they are giving the customer? The virtual customer becomes more inquisitive or confused by the response. Hiding behind policy and procedures? The customer becomes annoyed and asks for a manager. This will allow the agent to immediately see the impact of their decisions. All of this would not be possible without the advancements in speech recognition technologies. 

Agent training has come a long way over the last few years, and it can be difficult to keep up with all the advancements. One fact has remained constant: the closer you can simulate the contact center production environment, the more productive your agents will be. The industry will move toward simulations that mirror reality to the point that agents will be unable to discern whether they are talking to a real customer or a simulated one. As speech recognition continues to mature and be accepted as a viable technology, a whole host of new tools will be developed to allow nontechnical users to develop and deploy these powerful learning events. Speech recognition is about to change the way agents are trained, coached and prepared to handle customer contacts. 

Jeff R. Carpenter is vice president of Client Services for Sivox Technologies, Inc., developers of SIVOX RealCall, which creates simulations using speech recognition technology and emulation of data systems to immerse agents in an interactive e-learning experience. He is an expert in adult learning, having designed contact center training programs for large organizations, and also teaches courses in the subject at Roosevelt University in Chicago. In addition he is a member and past vice president of the Chicago chapter of the International Society for Performance Improvement as well as the American Society for Training and Development.

[ Return To The March 2003 Table Of Contents ]

By Mike Hoff, Access Direct

Every day, thousands of sales are generated by telemarketing sales representatives (TSRs) on behalf of their client organizations. Any analysis of the effectiveness of these telemarketing campaigns must include a look at the back-end retention figures. In other words, what percentage of the sales fall out and are then charged back against revenue already booked? And, what impact does this sales fallout have on a client's telemarketing cost per sale and campaign ROI? Ensuring front-end quality is the only way to prevent significant back-end fallout, control cost per sale and maximize one's telemarketing budget. The question is, how do you build in quality up front when constructing a telemarketing campaign? Any company working with a telemarketing vendor should ask what processes and procedures are in place to ensure strong sales results. Two critical areas to analyze are TSR training/coaching/motivation and sales verification.

The first step to ensuring upfront quality is to provide TSRs with the appropriate training, coaching and motivation to produce complete and accurate sales.

Training: Key to the success of any telemarketing campaign is how well the people on the phones have been trained. All TSRs need classroom as well as hands-on training with classroom time devoted to understanding policies, procedures, systems and industry regulations, in addition to sales techniques and the specific product knowledge required for the calling campaign. Training should not stop once TSRs move to the floor. Ongoing observation of actual calls is the only way trainers can judge the effectiveness of the training efforts and help mentor TSRs on to success.

Coaching: Regular quality assurance monitoring of all TSRs needs to be an integral part of any successful campaign. Assigning a monitor score is useless if there is no system in place to address and correct specific quality-related issues with the TSR. The QA department must have a way to communicate the specific concern and the TSR in question to the call floor supervisor, who can then coach and re-monitor the TSR for the remainder of the shift. Daily monitoring and coaching of all TSRs is the only way to prevent a quality problem from escalating to the point that it derails an entire campaign.

Motivation: Motivating TSRs monetarily to produce quality sales is an effective means of building quality into a campaign. When TSRs know that part of their compensation is based on net sales, rather than on gross sales, they are more apt to send sales through that are accurate and with solid end-user buy-in. Sales that are rejected at the verification stage for any reason will then cause the TSRs to loose their bonus for that sale. This policy can dramatically increase the sales pass rate because it holds the TSRs personally accountable for their sales quality. A higher pass rate translates into less re-work and ultimately stronger sales results. 

Once the tools are in place to enable the TSRs to deliver quality sales, a sales verification team needs to take over and ensure that only quality sales are being delivered to the client. When questionable sales are identified, then a strong callback/customer service team takes over to save the sale.

One hundred percent sales verification is an added expense some users of telemarketing services would rather not incur. Yet, when sales results are sent to the client without the benefit of a quality review, high back-end sales fallout can result. A good verification department identifies any misrepresentations, corrects any data entry errors and pulls out of queue any questionable sales prior to sales data transmission. They should also produce error reports, identifying specific quality issues by TSRs that are sent to the call centers prior to the next day's shift. This 'Pass Rate' report is posted in each call center so the TSRs can review their sales verification statistics. The failures can then be addressed by the call center operations team and individual coaching to TSRs on specific issues can begin.

A strong callback/customer service department is essential to saving those sales that have been rejected during the verification process. In cases where the offer or product was not completely communicated or a piece of information was missed, the customer is contacted and the missing information is obtained. Timing of the callback is critical ' ideally within the first 48 hours after initial contact ' to maximize sales recovery efforts. 

What are the costs associated with high sales fallout? Lost revenue and opportunity can add up quickly for the client whose telemarketing campaigns do not have quality built in from the beginning.

Take 1,000 sales with a 10 percent fallout: that translates into 100 sales. Historically, only 50 of those potential sales will be reached within the 48-hour window of opportunity. Of those 50, typically about half will be unrecoverable due to 'buyer's remorse,' leaving only 25 saved sales. This equates to a loss of about 75 sales in a single day, or 7.5 percent. With a focus on quality up-front, that fallout number can be cut to 5 percent, and by applying the same formula as above results in an ultimate sales loss of only 3.8 percent.

High sales fallout costs the vendor as well in terms of time lost to re-work. At the call center level, supervisors can easily spend an extra half hour at the start of each shift to address and correct performance and quality issues that translate into lost time. The verification/callback team also loses productivity due to re-work. Referring to the above example of 1,000 sales, when the verification team is faced with 10 percent fallout, the time they spend dealing with quality issues will obviously be double what they spend on campaigns with only 5 percent fallout. This additional personnel cost will often be passed through to the client in one way or another. 

Building quality into every step of a telemarketing campaign is the only way for a client to achieve the high quality and solid sales results they expect from their telemarketing vendor. Making this kind of investment up front and avoiding costly re-work delivers maximum ROI for their telemarketing budget.

Mike Hoff is senior manager, Policies and Procedures at Access Direct and has been with the company since the doors opened in 1995. He has developed many of the quality processes used today. In addition to quality assurance, Hoff oversees verification, internal reporting, auditing and policies and procedures. He has spent the majority of his business career in the telemarketing industry, having started on the phones, and worked in a variety of operations positions.

[ Return To The March 2003 Table Of Contents ]