Using AI in Natural Language Processing for B2C, Call Centers

PRODUCT ANALYSIS: How Marchex brings actionable intelligence to business interactions with an AI-powered platform that leverages natural language processing in real time.

Marchex

Seattle-based Marchex is looking to vastly improve how businesses interact with customers, increasing lead conversion, improving the customer experience and establishing repeat business. These are goals that are not lost on those trying to grow businesses in today’s highly competitive market.

Bringing Additional Value to Business to Customer Interactions

Many organizations have discovered that interacting with customers (or potential customers) is not as easy as it seems. There are a plethora of challenges when it comes to maximizing the value of customer to business interactions. Those challenges are further complicated by the lack of relevant metrics to initiate change for the better. Simply put, you can’t manage something if you can’t measure it.

For the majority of organizations that seek to sell to and service customers, most interactions start with the age-old technology of a phone call. A typical customer interaction scenario is when a customer (or potential customer) calls an organization to purchase a product or a service. How a business handles phone call interactions has a significant impact on the customer experience–and ultimately buying decisions.

Every interaction has the potential for multiple outcomes, each defined by a level of success or failure. Proper orchestration of those interactions requires measurement and management to ensure the desired outcome. This is something most businesses have failed to encapsulate into their customer communications initiatives.

That is where Marchex comes into play with the Marchex Call Analytics platform. Its speech analytics technology analyzes business-to-consumer interactions in real time using advanced NLP (natural language processing) coupled with machine learning and artificial intelligence. Marchex is able to generate multiple signals from customer interaction to determine intent, sentiment, success and numerous other outcomes that can improve results in marketing, sales and service departments. The latter includes reaching new customers, converting more sales and retaining more customers to increase their lifetime value.

The Intersection of Machine Learning and Natural Language Processing

Acronyms such as ML, NLP and AI have been thrown about in the analytics arena for some time, yet the foundations for those systems are anything but equal. For example, NLP systems have made great strides when coupled with AI-powered assistants. Technologies such as Apple’s Siri, Amazon’s Alexa and Microsoft’s Cortana have now readily made it possible for machines and humans to interact and are getting better all the time. IBM’s Watson has also become a master in the NLP game, able to analyze both the written and spoken word to bring forth value in captured data.

That said, intelligent assistants used in the call center space need to have a different focus on NLP than human-to-machine interaction. AI, ML and NLP must intersect in such a way that the technologies can understand the conversation between two or more humans and must be able to garner meaning from those conversations in real time. That real time, conversational NLP is where Marchex has put its greatest AI efforts. Marchex has tuned its system to perform in the real world, where conversations take place in real time and AI-powered systems are able to complement human efforts rather than replace humans.

Recent in-house, independently verified testing demonstrated that Marchex is able to exceed what other NLP vendors are doing in conversational real-time NLP. Using an extensive library of consumer-to-business conversational calls, the test measured the metric of WER (word error rate). For comparison's sake, humans can achieve a WER of about 5 percent. The same data set was used to test Amazon Transcribe API, Google Cloud Speech to Text API (phone and video), Microsoft Azure Cognitive Service API, and IBM Watson. Marchex’s solution scored favorably across the board, achieving an overall WER of 8.4 percent, which is more than 35 percent more accurate than either IBM Watson’s or Microsoft Azure’s respective solutions. Amazon’s tool reached an overall WER of 9.5 percent while Google Speech (phone) had a 9.7 percent overall WER.

Accuracy matters when dealing with real-time consumer-to-business interactions, which is Marchex’s primary domain, which most likely tilts the scales in Marchex’s favor over other NLP solutions. That said, the other vendors mentioned here are focused on a different domain, such as human-to-machine processing and deep data analytics based upon recorded or written information and may offer superior performance in their own respective domains.

A Closer Look at the Marchex Platform

Marchex Call Analytics brings previously unforeseen value to the data that surrounds speech.

Marchex takes a multi-pronged approach to resolving the issues associated with human-to-human interactions in the enterprise business space. The company’s primary domain fits into the call center ideology, where customers and representatives interact. Those interactions can range from basic sales inquiries, customer service requests, product information and many other customer to business interactions. Regardless of the type of interaction, the critical measurement of success is still strongly linked to customer satisfaction.

Marchex has created a cloud-based platform that gathers verbal communications in real time to build relevance into the data and bring additional value to interactions. Primary analysis is accomplished using an advanced conversational Large-Vocabulary Continuous Speech Recognition (LVCSR) engine, which performs speech-to-text conversion in real time, while also creating the metadata needed for advanced analytics. That information is then presented on browser-based dashboards that illustrate all of the critical metrics needed for businesses to take the necessary actions to improve interactions with customers. All interaction data and associated analytics processing takes place in Marchex’s own secure data center and is enabled by the Marchex Call Analytics platform.

Marchex’s cloud-based ideology eliminates many of the setup and scaling concerns that plague businesses today that are trying to survive the throws of digital transformation, while also introducing advanced analytics capabilities into their scope of IT functions. Marchex is also on a path to analyze more than speech, with APIs in the works to garner data from SMS, texting, chat and other interaction technologies used when businesses and customers communicate.

The idea here is to present managers with the actionable data to make decisions to grow the business while improving relationships with customers. Near-immediate improvements can be envisioned and deployed using Marchex’s reporting systems and dashboards, which aggregate data into relevant sentiments to target areas of potential improvement. What’s more, Marchex presents that information using browser-based customizable consoles that integrators can further customize for specific use cases.

Hands on With Marchex Call Analytics

Marchex Call Analytics is a cloud-based solution, meaning customers can start using the platform without the use of additional hardware or onsite servers. This simplifies the process to integrate with call centers or businesses with distributed retail locations. After the platform integration process is completed, end users are granted access to the analytics platform via a web browser-based dashboard. The platform records and organizes call data and brings forth additional insight into interactions--an important capability.

Marchex brings value to the equation by analyzing conversational data and presenting it to users with easy-to-understand dashboards. Marchex thus enables a browser to become the avenue of insight into call and communications effectiveness. While the technology that makes that possible is complex, that complexity is completely hidden from the end user. That means using all the service has to offer does not require modifying infrastructure, adding processing power, reconfiguring storage, worrying about scale, hiring data analysts or even implementing extensive training programs.

Onboarding into the system is straightforward: A hierarchy is set up on the system that breaks down a business’s operational matrix into the most logical elements for analytics, and a master business account is set up, with satellite locations associated with it. That allows data analytics to be sliced and diced based on any criteria needed by the end user.

The majority of businesses have internally defined how voice communications should take place between a customer and an agent. For example, agents are provided with scripts and/or training to educate them on how to interact with a customer. Key elements, such as greetings, introductions and so forth are introduced to the agent so that they may optimize the interaction with the customer. Those scripts are ingested into the speech analytics system, so that businesses can measure agent performance and track how successful interactions are.

Other preparatory steps may include defining campaigns, where a particular call-in number is associated with an advertising campaign. Also, regional information, such as operational hours, geographic locations, keywords and many other elements can be included to fine tune the analytics to provide the most pertinent information with an interaction.

Behind the scenes, calls are recorded, metadata collected and speech is analyzed to create graphical representations of cumulative interactions, with the ability to drill down further into specific time frames, individual calls, as well as campaigns. That is exactly where the various dashboards come into play.

Overview Dashboard

Dashboard1

The Overview Dashboard brings forth the most critical information needed by call center managers in an easy to understand format. Managers can set date ranges to look at aggregated activity and quickly spot trends or indication of trouble. For example, calls that are classified as lost opportunities are displayed graphically, allowing managers to quickly spot where they may be having an issue with how calls are being handled.

Dashboard2

Additional information is available, such as the total number of calls broken down by region or location, with additional metadata exposed in the table, giving managers a quick peek at how defined call center expectations are being met.

Lost Opportunities Dashboard

Dashboard3

One of the most critical aspects of operating a sales call center is the concept of lost opportunities. Simply put, a lost opportunity indicates that something may not have gone as planned on a sales call. Marchex’s Lost Opportunities Dashboard exposes all of the metadata behind those calls identified as a lost opportunity and presents it in an easy to understand, customizable table. Managers can filter the results based upon different criteria, such as date ranges, locations and even costs.

The Lost Opportunities Dashboard can serve multiple purposes, such as helping businesses determine insights around revenue goals. For example, a business may be looking to calculate the ROI of an advertising campaign, where by associating that campaign with a given agent, call center, location, or even a dedicated phone number. Those calls can then be analyzed and potentially reveal the success or failures of that given campaign. With Marchex Call Analytics, the overall impact of the campaign can be measured and if need be, dissected, to find out where additional revenue could have been generated.

Elements including the number of calls, the length of calls, failure rates, call abandonment and others are readily displayed on the dashboard, allowing call center managers to drill down into additional metadata elements. That gives additional insight into the reasons why a call became a lost opportunity. Reasons such as “customer hangs up during hold,” “no answer,” “bounce to voicemail,” “lost connections” or even “agent call handling performance” all are aggregated and presented in such a way that business managers can quickly determine what is happening.

High-Intent Dashboard

Dashboard4

As the name implies, the High Intent Dashboard visualizes those callers with a high intent of interacting with the organization. For call centers focused on the sales side of the equation, high intent indicates the likelihood of closing a deal. In other words, high intent is a direct measurement of a successful call.

The visual representations of successful a calls proves to be very useful for measuring campaign performance, agent performance and customer interactions. The dashboard offers critical metrics such as call volume, call timing, and the overall success rate of those calls. Information which is correlated to revenue, giving call center managers a dollars and cents view of success.

Agent Script-Tracking Dashboard

Dashboard5

Successful customer-to-business calls require that agents stay on message. In other words, most organizations coach their call center agents and provide scripts to keep all messaging uniform. For many businesses, numerous hours have gone into perfecting their customer facing messaging, and making sure that messaging is consistent has proven to be a major challenge. With the Agent Script Tracking Dashboard, Marchex has automated the tedious process of validating messaging and visually demonstrates when and where agents stay on message, as opposed to going off on tangents that may be harmful for the business.

For example, a business may have a general rule that every interaction starts with a proper introduction. The speech analytics engine of the Marchex platform can detect if that happens or not. Another example may come in the form of making sure prices are not discussed during an initial call. Here, Marchex is able to detect the words surrounding pricing information and report on it.

Ultimately, script tracking delivers aggregated data that can be filtered to give users insight on how well agents are performing and whether or not a call passes defined criteria. That criteria can be defined by the user and can be representative of keywords, phrases, scripts, and other signaling data, allowing complete customization of what is tracked.

Transcription Search Dashboard

Dashboard6

One of Marchex’s biggest strengths comes from its ability to record calls and store them for future reference, while also creating the associated metadata and transcribing the call. Calls are recorded, analyzed, transcribed, indexed and processed by the NLP system. That gives call center managers the unique ability to search through all of the calls using fuzzy logic and correlate the analysis with the actual calls.

It is a capability that creates numerous discovery scenarios, where call center managers can find calls based upon a keyword or phrase and then listen to the original audio associated with the call, while also viewing the relevant metadata, such as caller ID, time frames, call length and so on. Advanced filtering allows managers to narrow the search and base results on specific regions, agents, campaigns, and so forth.

Conclusions

The Marchex Call Analytics platform offers capabilities once only dreamed of in business organizations that interact with customers. The product’s accuracy and speed fuels new interaction ideologies and sets the stage for interactive NLP that can bring forth success during customer conversations. The company is creating APIs that ease integration with CRM systems and other tools that have become all too familiar to phone-based sales people, help desk engineers and many others who need to interact with customers several times a day.

Aggregation and analysis of call data is only a small part of the overall picture; as the technology advances, businesses will be able to create custom solutions that bring helpful AI into the mix that can enhance conversions, speed resolutions and interactively train sales agents to be more effective in their communications.