Insuring a Smoother System

Say one division of a major health insurance provider integrates its claims processing operations with those of another division within the same company. As a result, each customer receives a new identification number with an alphanumeric prefix, and some benefits packages change. Now imagine that not all healthcare providers in the area are aware of the changes and fail to adjust their records accordingly, causing customers to receive statements that no longer make sense. One could take a wild guess as to why the health insurance company’s contact center would receive a major spike in call volumes.

This was the scenario in which Blue Cross Blue Shield of Northeastern Pennsylvania (BCBSNEPA) found itself last year. While most in the organization knew what the problems were, customer service managers couldn’t begin to address them without first gaining the support of corporate leadership. For that, they would need very detailed data.

"We needed the data to back up what we were thinking so we could get the support to take action and correct things," says Bob McDonald, BCBSNEPA’s director of service operations. "We needed to give [ourselves] more meat and reiterate what we were doing and put more emphasis on [the problems]."

The company had a system in place to track why customers were calling, but it was decidedly low-tech. The system could categorize a call as related to a benefits question, for example, but didn’t provide McDonald’s team with much context beyond that. To dig deeper, agents worked on a spreadsheet, placing ticks into subcategories. McDonald knew this was not the most efficient route to take; there had to be a way to glean context from an otherwise open-ended spray of content. Taking a 180-degree turn from its previous data analysis methods, BCBSNEPA looked for a speech analytics solution that would not only categorize calls, but would also provide the additional  context needed.

Rather than traditional word spotting done through speech recognition, speech analytics catches utterances, categorizes calls grouped by subject, gauges a customer’s emotions throughout the call, and then contextualizes the call by compiling words in a way that creates a more detailed picture of the caller’s intent. A company can then use this information to change its contact center infrastructure, train agents, or optimize call paths. McDonald wanted to validate his theories as to why call volumes spiked, to "see if there was something we were missing, and see if we could improve upon those situations."

BCBSNEPA eventually selected Nexidia’s Enterprise Speech Intelligence solution. Though speed was a definite deciding factor in BCBSNEPA’s choice of vendor and solution, one of Nexidia’s most appealing features was its phonetics-based approach to collecting data. This method strings words together based on phonemes, in stark contrast to Large Vocabulary Conversational Speech Recognition (LVCSR), which finds key words and phrases using a preloaded dictionary.

McDonald says BCBSNEPA chose a phonetics-based solution because it suited the medical field particularly well. "We like the fact that there’s no dictionary in our product," he states. "We have [calls with] a lot of jargon and medical terms, doctors’ names, and all those types of words that would be very difficult to maintain in a dictionary. We’d have to add words constantly; every other month there’s a new prescription out there."

The company also liked the fact that Nexidia’s analytics engine could process 100 percent of its call volume. And it offered the solution via an on-demand model, an attractive option since many analytics solutions can be expensive to implement. It also gives companies like BCBSNEPA the chance to learn from the pros.

Of course, before BCBSNEPA could test the system, it had to let Nexidia take the reins. No matter how simple an analytics interface may seem, being able to mine the data for meaning is something that must be learned over time. To start, McDonald gave Nexidia recordings of calls collected over a two-month period. During the three months that followed, Nexidia employees analyzed those calls and generated reports based on the findings. Throughout the process, McDonald worked with Nexidia to help optimize the results, primarily because Nexidia had to be coached on the healthcare vertical.

"The people that did the [initial] analysis did a great job considering they’re not in the health insurance industry," McDonald says. "They spent a lot of time with me on the phone, asking what I was looking for, what my challenges were, how the organization is designed, what the industry was like. They went into a lot of detail to understand my business. Within the first week they already had a good concept of what I did."

Getting to the Cause
Two weeks after receiving the initial data, Nexidia filed a preliminary report that not only identified the cause of BCBSNEPA’s spike in call volume, but also the reason so many calls were being misdirected. As one would expect, the change in benefits led to a spike in that part of the call center’s volume, but call volumes also increased for other areas not related to claims. Ten percent to 15 percent of calls from customers were misdirects that then had to be rerouted, costing agents up to two minutes of handling time, disrupting call flows, and increasing hold times. Many of those callers were dialing into the wrong number on purpose. Before the analytics solution was implemented, McDonald’s team had no idea why customers would do that. Nexidia found that they were trying to circumvent clogged lines by calling different service numbers, in hopes of having  their questions answered faster.

"We helped [agents] by rerouting the call flows a bit so the people that were calling to circumvent the system weren’t getting prioritized; they were being treated like everyone else," McDonald explains. "We were helping script the reps so they could handle situations [in which the customer called the wrong number] faster, and spend that time moving on to the next phone call. That was a big eye-opener."

Nexidia also uncovered a similarity between the name of BCBSNEPA’s new service and another section’s line, further adding to the confusion.

Additional findings only solidified what McDonald and his team already knew: Healthcare providers were improperly processing new customer IDs with the three-character alphanumeric prefixes. Using that data, McDonald pushed for BCBSNEPA field agents to remind healthcare staff of how to properly identify clients.

Today, BCBSNEPA is working on using the system to improve contact center processes, as well as activate an outbound calling system that will contact customers whose initial calls to the contact center were unresolved. This information can be taken from data produced by the analytics system.

"We will be using it to help us proactively contact members who maybe weren’t too happy with the phone call from the previous or same day," McDonald says. "I can run a query and make an outbound phone call; it would make a big impact on member loyalty."

The company also hopes to achieve first-call resolution by further training agents. But for now, McDonald says it wants to focus on just three ways to use the data and "not go overboard."

Currently, Nexidia still hosts BCBSNEPA’s analytics through the on-demand business model, but McDonald says his company will bring the solution in-house this summer. He also believes his pitch to higher-ups that Nexidia’s solution would produce a return on investment within three years is extremely likely. "It easily justifies itself within three years," he says. "I’m sure you could do an ROI for Nexidia within a year if you needed to."

When the Nexidia solution is placed on-premises, McDonald’s team will compile daily reports based on the previous day’s calls. These reports will include the types of calls the contact center received, if there was a spike in call volume, and why there was a spike.

For now, McDonald is focused primarily on how to improve overall contact center interactions to better the customer experience. "[Analytics] is going to help us in coaching our individual reps. We’ll be able to identify situations with individuals to improve average handling time," he states. "We’re really excited about it."