Real-Time Results
1 Sep, 2008
By: Tim KraskeyWhen a customer says, “Let me talk to your manager” or “Cancel my order,” your contact center needs to act fast before your business takes a hit.
Traditional recording tools can capture customer voice interactions with a contact center agent. Speech and content analytics tools, in turn, can then help uncover trends in the customer and agent interactions within those saved recordings. This insight allows a contact center to enhance its processes to improve overall business. However, the initial damage is done based on the customer contacts that have already occurred. What contact centers need is a way to remedy issues and enhance interactions with customers in the moment, as they occur.
An effective way to achieve this is with real-time speech and content analytics, which pinpoints telling speech phrases and then combines this data with other identified key content, to make improvements to agent-customer interactions in real-time. As a result of real-time analytics, improvements to processes within the contact center no longer are left to hindsight.
When applied correctly, real-time speech and content analytics is a powerful way to analyze unstructured data immediately. It provides valuable insight into contact center activities, identifying issues as they happen and preventing situations from escalating to those levels, for example, where a customer takes his or her business elsewhere.
Speech and content analytics is not a new concept, but virtually all previous applications have focused on post-call analysis — that is, analysis after the call is completed. Real-time implementations of speech and content analytics still remain fairly untapped, yet their potential to transform business is undeniable. By leveraging these real-time tools, an enterprise can secure an unparalleled view into contact center activities, including calls in progress — and ultimately improve contact center operations in line with overall business objectives, before it’s too late.
Basic Analytics Applications
The basic premise behind content and speech analytics is the ability to monitor conversations between customers and agents, and then analyze and evaluate these interactions in alignment with overall business goals. Content and speech analytics leverages factors like speech inflection, phonetics, periods of silence, and key words and phrases. Analytics also can identify specific elements such as inappropriate or foul language, mentions of competitors’ names and phrases like those common to a dissatisfied or satisfied customer. And it can evaluate metadata, or “data about data,” including telephony data as around customer hold time, to help present a clearer picture of the overall customer experience.
For example, a contact center may use speech and content analytics to understand the relationship between customer satisfaction, customer retention and a specific business service. Contact center managers can use information gleaned from speech analytics to make the necessary adjustments to business processes and train agents on how to better serve customers and, ultimately, help increase customer satisfaction.
However, the gap between when an issue occurs and when a contact center implements the necessary measures to prevent it from reoccurring can be glaringly large. For a business, red flag phrases like “cancel my order” indicate a larger problem, but such issues are difficult to resolve if the customer has long since hung up and the analysis is only starting to take place. This is where a real-time approach to content and speech analytics becomes key.
By applying speech and content analytics to processes and data in real-time, contact centers can detect issues and prevent them from escalating to larger problems, including lost sales or lost customers, or seize new business opportunities before they evaporate.
Analytics in Real-Time Action
Consider the insight real-time speech and content analytics can yield when applied to a call in progress:
• Is the agent listening to the caller or are the agent and caller talking over each other?
• Is the caller becoming frustrated or angry?
• Does the caller sound happy and satisfied?
• Is the caller showing signs of inclination toward buying more products or services?
• Is the caller considering contacting a competitor?
• Is the caller considering cancelling his or her account?
By answering questions such as these, real-time speech and content analytics can influence calls to effectively alter their direction and boost first call resolution, a key factor in customer satisfaction levels. The tool’s power to improve first call resolution is important because if a customer issue can be correctly identified during a call, an agent will be more likely to resolve the problem within the same call, eliminating the likelihood of additional calls. The more frequently this occurs, the more likely a contact center is to improve customer satisfaction — and, ultimately, sales.
Conversely, the real-time speech and content analytics also can leverage the sales potential of positive customer calls and seize opportunities to sell more to customers who demonstrate levels of satisfaction. Overall, real-time analytics tools can powerfully impact call results and improve contact center performance by saving sales that might otherwise have been lost, or identifying and taking action on opportunities to up-sell.
Consider an example scenario of real-time speech and content analytics at work. During a call, the tool might detect a noticeable shift in the pitch of a caller’s voice or phrases potentially indicative of dissatisfaction. The tool can go so far as to help determine if the dissatisfaction is associated with the caller’s interaction with the agent or with one of the enterprise’s products or services. In response to this information, the tool can then display messages to help coach the agent through the call.
If the caller’s dissatisfaction is in response to the agent’s behavior, the tool sends an alert message to the contact center supervisor’s computer, which in turn can evoke or generate a response from the supervisor. The supervisor can then take a number of actions to remedy the situation, such as monitoring the call, joining in on the call or silently coaching the agent through the call via instant messaging.
Real-Time Interaction Methodology
For optimal results, contact centers should apply a combination of both post-call and real-time speech and content analytics. By applying the tools post-call, contact centers can analyze business interactions and uncover areas where business process enhancements are needed. Applying the tools in real-time, however, allows contact centers to optimally train agents and reinforce behaviors to help make business process improvements that, ultimately, result in successful outcomes.
Beyond this basic premise, speech and content analytics is a highly customizable tool that contact centers can tailor to serve their specific business needs for meeting overall objectives. Plan for an evolutionary process that will first identify issues and opportunities, but continue to improve and refine search and analytics criteria to maximize overall business results. Consider the following steps when applying real-time speech and content analytics:
Step 1: Formulate Key Business Questions
Determine which key business questions speech and content analytics will help answer. This will provide a clear and defined purpose for the way in which the contact center applies analytics. For example, a contact center might seek to determine why it is selling more of one product than another. Other questions might seek to determine which competitor names callers mention when they talk about canceling their service.
Step 2: Select Evaluation Criteria
Using the key business questions, determine the evaluation criteria which analytics will use to answer the key business questions. The criteria will help filter calls to glean the most relevant insight from data.
To select evaluation criteria, review several past call recordings to identify potential indicators, like certain words or phrases, detected speech energy levels, competitor name references, average caller hold times or the average number of transfers within calls. Ultimately, indicators tie in to the call outcome, so keeping this in mind will help narrow the possibilities of what to include.
Step 3: Apply Criteria, Then Refine and Repeat
As with any implementation, testing is a natural part of the process. For speech and content analytics, this entails applying the selected evaluation criteria to live calls. Review the filtering results to determine how effective the criteria was in identifying information, noting that criteria applied in real-time can often yield significantly different results than post-call filtering. Consider if the selected criteria reliably detect problems and opportunities, and adjust accordingly.
Step 4: Continue Testing and Refining
While some criteria may seem obvious and work from the outset, others will take time to develop. As a result, it may also take time and continuous tweaking to uncover ways to improve contact center performance.
Solid and Flexible Foundation
A solid architecture is a key element to laying the groundwork for an effective real-time speech and content analytics system. First and foremost, the architecture must be flexible and configurable, complete with a rich set of features that can truly support many kinds of analysis, including speech energy analysis and word-spotting capabilities. The architecture must also be able to support the analysis of other call data, such as caller hold time, caller history and IVR selections. If the system supporting real-time speech and content analytics has several different features, it will be more useful for extracting information from the variety of processes and activities to provide a complete view of the customer’s experience within the call center.
Specifically, for a real-time speech and content analytics system to be most effective, it should be based in an “edge”-oriented framework. While centralized architecture supports many speech and content analytics systems, a system that incorporates “edge” intelligence offers several advantages over this traditional approach, most notably for its flexibility, scalability and ability to most effectively work in real-time.
Examine how speech and content analytics works in a centralized framework. The system copies audio streams from both the caller and the contact center agent over the network for recording and then analysis back at a centralized server location, or “server farm.” The disadvantages of this approach are numerous, however, and include excessive bandwidth and network delay — key issues that hinder the basic premise behind real-time operation.
Conversely, speech and content analytics based in an edge-oriented architecture relies on processing at each agent’s personal computer, where the audio streams first enter the contact center. As a result, contact centers reap several benefits, including:
• Improved Economy and Scalability. Using the excess processing capacity at each agent’s PC means the system doesn’t require as much centralized support — and thus reduces the number of centralized servers needed to support contact center operations. This edge processing also enables voice recording and analytics to scale with contact center size, automatically increasing the contact center’s capacity without requiring more centralized servers and allowing for multi-site applicability.
• Speedier Processing and Responsiveness. Edge architecture optimally facilitates real-time voice analysis by enabling it to occur essentially as it happens. It eliminates network delay and “backhaul traffic” between live conversations, the recording process and, ultimately, the opportunity to immediately seize opportunities or resolve issues.
• More Flexibility. Real-time speech and content analytics isn’t a “one-size-fits-all”-type of system and requires adjustments and alterations in order to adequately match up with a contact center’s needs. Edge-oriented architecture provides this needed system flexibility since processing takes place at the network edge and thus allows greater freedom for evaluation criteria customization down to the individual agent. This allows different teams or individuals to focus on specific business issues, which can vary across an enterprise.
Finally, to realize the full benefits of real-time speech and content analytics, the system should integrate with other contact center applications. This includes CTI systems and other applications used to facilitate contact center processes and provide indicators, such as supervisor notifications or alerts when issues arise. Ultimately, this integration allows for real-time views of not only agent-caller interactions, but it also ties together performance, revenue and customer satisfaction indicators.
Seizing Real-Time Opportunities for Success
“If I would have known then what I know now.” With today’s real-time analytics tools, this can be a phrase of the past. When used together, real-time and post-call speech and content analytics can powerfully impact business by enabling contact center managers to both learn from the past and influence the present — and, ultimately, positively impact the future.
