Fast, accurate transcription of calls with speech analytics can deliver insights quickly, helping you transform your business.
At SOCAP’s Symposium, we let members record commercial scripts and receive an analysis of their voice quality and performance. The idea was to put members in the hot seat to see what’s involved in making a quality voice recording, and then compare their results with the same script recorded by a veteran voice talent.
To make it enjoyably competitive, we created a mock voice analyzer that dispensed Simon Cowell type feedback along with voice ratings and suggested career choices. What started out as a fun activity, sparked serious discussions about voice analytics and its application to customer service and call center operations and training.
With such a high level of interest around analytics at the conference, I turned to our speech analytics partner, Voci Technologies, to uncover the underlying value of voice transcription and speech analytics, along with market trends and advice.
Analyzing voice calls is not a new concept. Companies generally recognize upwards of 70% of customer feedback is sitting in their call recordings—an untapped resource. And while speech-to-text (STT) technology has evolved during the last 10 years, it continues to have numerous limitations. Audio files have historically been large and laborious to transcribe to text, and call recording providers haven’t made it any easier. The low margins associated with call recorders has meant low quality output, with some providers holding the audio files hostage, charging exorbitant fees to extract the data.
This means that while most companies acknowledge the enormous potential of stored voice recordings, they simply sample audio to feed performance management or quality control operations of their call centers. The inability to cost efficiently transcribe large volumes of voice to text all but eliminated voice calls from most analytics platforms.
A Typical Call Center “The Old Way”
To highlight the limitations of old school speech analytics, let’s take a look at a typical call center. With the intention of quality assurance (QA) and working towards improved agent performance, a call center employs a certain number of agents to pull calls from recording archives. A typical goal is to review 10 calls per agent per month. The agents listen to all calls manually and makes notes against a quality checklist.
Each agent is scored for the entire month based on these 10 calls, but a full-time employee may handle 700 to 1,000 calls a month. The 10 monitored calls represent a 1% sample size of the agent’s work. The result is both faulty QA and performance management. Suppose the QA agent is having a bad day or is fatigued. Manual reviews of calls result in subjective notes and will be inconsistent across QA agents and other factors. For matters of compliance, with a 1% sample size, a company would have no way of knowing exactly how often agents are compliant with specific scripts, resulting in risk to the organization.
From the agent’s perspective, how accurate are the evaluations he or she receives? Statistically speaking, the sample size is not large enough to know either way. Agents recognize this and often resent the evaluations leading to decreased engagement across the board.
A New Way Through New Technology
New technology out of Carnegie Mellon University has eliminated the limitations noted above, disrupting the marketplace and leading to the contact center of the future. Fast, accurate transcription with open architecture delivers insights quickly in virtually any environment, regardless of the call recording or analytics components.
With the swing back toward focusing on the customer experience, companies are realizing the value of integrating customer feedback across channels to better understand how they can improve not only the CX, but internal efficiencies at the same time. While CX professionals start their journey focusing on the digital, they are quickly realizing voice calls offer the bulk of customer feedback. And marrying the voice feedback through transcription and integration, they can get a 360-degree view of their customers’ experience.
Unlike in the past, an STT engine with open architecture allows companies to process voice calls once and feed output to as many analytics systems and lines of business as needed. This lowers total cost of ownership of the entire ecosystem.
A Case Study With New Technology
Let’s look at an example of how new technology can enable remarkable results. Consider a medium sized B2C company. One call center sales agent repeatedly closed 20% to 30% more sales than the rest of the sales team. The sales manager tried to discern what he did to achieve such relatively high performance. The manager silently monitored the agent for multiple shifts, but nothing surfaced. The sales manager then had this star agent train other sales team members on his sales methods, but the team experienced no change in performance.
Three months transpired with no improvement. Multiple tactics were used, taking time and resources. Still, no improvements. And still, the star agent continued closing 20% to 30% more sales than the rest.
Around this time, the company ran a trial with speech-to-text and voice analytics to better understand how compliant their agents were across the call center population. As part of the trial, the analyst ran voice calls through the system, generating transcripts for all calls. With knowledge of the performance disparity within the sales team, the analyst ran an experiment. She separated the star agent calls from the rest of the sales team.
Next, she filtered the calls to group those that were successful versus unsuccessful. The analyst looked for patterns in the conversations and noticed in every successful call, the star agent mentioned the company was A+ rated with the Better Business Bureau (BBB). The general sales agents rarely mentioned the A+ rating. Aha! This insight was immediately reviewed with the head of sales.
Upon review, the head of sales thought this could be the secret sauce. They immediately arranged for a sales training session. The agents were instructed to mention on every call that the company was A+ rated by the BBB and sent them back into the field. One week later the sales stats were delivered and the numbers had not improved. The head of sales concluded the analytics were useless.
Befuddled, the analyst pulled that week’s calls and ran the same analysis. To her surprise, she found the agents were, in fact, not mentioning the A+ rating on calls. The training was ineffective in changing behavior.
The head of sales called the sales team back for a meeting and asked why they were not compliant with the request. The sales team responded with, “We don’t know how to work that into our pitch.” The team adjusted the training to include small-group role playing to get the agents comfortable and identify methods to work the statement into their calls. They were sent back into the field.
What were the results? Within weeks, the performance of the entire sales team jumped significantly. Analyzing 100% of the audio calls enabled more effective training and incremental sales growth.
Beyond Agent Training
While the case above illustrates a major benefit of full voice transcription and speech analytics, there are more use cases and benefits. The following highlights key areas most likely apply to your business. 1. Reduction of voice calls and therefore customer dissatisfaction and costs:
Self-service is the name of the game. Consumers want it, companies prefer it. Analyzing voice calls offers insight into issues with online tools and services and interactive voice response. Oftentimes when an issue is highlighted through analytics, a quick fix solves the problem. 2. Marketing experimentation:
Your company wants to A/B test messaging to drive traffic to the web or calls for new business. With an omnichannel approach, analytics can uncover which words or phrases are referenced when customers reach out, identifying the most effective messaging. 3. Internal processes:
Analyzing the amount of silence on a call can uncover internal process or training issues that are leading to longer call duration, which leads to increased customer dissatisfaction.
With so many use case areas, we could not capture them all. The important part of this topic is identifying what is most important to your company, building use cases that will affect your strategic initiatives and starting to analyze. Remember, if you are not capturing voice calls today, you are missing close to 70% of customer feedback, and in doing so, perhaps chasing opportunities that may not lead to greater CX and performance.
Thinking back to the booth activity, it’s easy to speak in generalities like we did with the fake analytics, but it’s subjective in nature and prone to error. However, using technology that can evaluate the outcome empirically takes the subjective decisions out of the equation, leaving facts and data which can be acted upon then measured for change.