Improving Quality Assurance with Speech Analytics
Quality assurance (QA) is a mission-critical business function that identifies contact center and enterprise trends and provides insights into how well each agent is performing. For the past 30 years, QA has been performed in contact centers around the world in basically the same way. Managers search through volumes of recordings to find the 10 to 20 percent of calls that require attention, either because they are really bad or really good. If a call is bad, the QA specialist or supervisor must document the situation by completing a QA evaluation form, and then arrange a coaching session with the agent. If the call is really good, the agent should be commended and, ideally, recognized.
Traditional QA is a labor-intensive function, based on an often random selection of calls, supported by technology but never truly automated. QA evaluations are frequently not performed at all or left until the last minute and then rushed through, delivering little benefit for the department, agent, or customer.
Contact center leaders are aware of the weaknesses of their current QA programs, but recognize that although the process is imperfect, QA adds considerable value to their department. Even a limited amount of information about trends, systems issues, bad policies, and poorly performing agents is better than none. However, the challenge has always been to convince front-line supervisors, who are already too busy, to dedicate the time to listen to calls in order to find the ones that need their attention, and then to perform evaluations and actually coach their agents. Exit studies at outsourcers often find that the number-one reason agents resign is because they do not receive timely or effective coaching. This issue can be addressed with an automated QA process that can monitor 100 percent of calls, categorize them, and identify emerging trends and patterns. An automated call identification process provides a closed-loop mechanism to deliver appropriate coaching to agents on a consistent and timely basis. The concept of automated quality assurance is called precision monitoring, and it represents the future of QA.
What Is Precision Monitoring?
Precision monitoring is a process that is enabled by the introduction of speech analytics into contact centers. Managers start the process by setting up special QA searches in their speech analytics application (it is a best practice to set up multiple searches to address various aspects of QA). Recommended searches include:
- Greetings: Define all of the appropriate ways agents should perform a greeting, and capture interactions where this does not happen. Alternatively, set up a search that includes all of the words and phrases that agents should not say.
- Closings: Same approach as for the greeting.
- Verification: Define words and phrases that indicate that caller identification and verification is being done correctly.
- Customer satisfaction: Identify words and phrases that denote customer satisfaction. Allow for a broad range, as there are many ways that customers express their satisfaction.
- Customer dissatisfaction: Even when a customer satisfaction search is set up, it's also a good idea to capture situations where customers are dissatisfied. Keep in mind that a customer could be satisfied with an agent but dissatisfied with a product, policy, or process, or vice versa.
- Agent engagement: Though relatively tough to identify and measure, this is a great category to capture, as it can be a customer experience differentiator. Each organization has certain words and phrases that show agent interest, but the words vary based on the topic. For example, if a customer calls to report the death of a spouse, the agent's initial response should be, "I am sorry for your loss." Or if a customer calls and says, "I am really angry at your company for doing X, Y, or Z," the agent's first response should be "I am very sorry. What can I do to help fix (or improve) the situation?"
- Risk (script adherence): Create a series of searches to identify essential phrases that must be spoken as part of each conversation. These searches can be complicated, as they vary according to the nature of the call. The complexity is compounded because in most cases, it is not enough to say just one phrase or sentence. Agents often have to deliver a great deal of information to their customers, and they put their enterprise at risk if the message is not delivered clearly and completely.
These are just a few examples of the types of QA-oriented speech analytics searches that managers should set up. While the searches outlined here apply to most customer service situations, many others will need to be set up to capture the essential element calls. DMG recommends that businesses work with their speech analytics provider to identify and define appropriate searches to get started with precision monitoring. Establishing the searches for a new analytics-enabled QA process (precision monitoring) is an iterative process, and most organizations get better at capturing relevant and important calls as they fine-tune and enhance their searches.
Once targeted calls are identified through the speech analytics process, a supervisor still needs to listen to them to ensure that the system is working properly and has detected the calls that require intervention in the form of coaching or commendation. The difference is that, now, supervisors need to listen to only the 10 or 20 percent of calls that require their direct and often immediate action, instead of wasting hours reviewing all calls just to find the few that would most benefit from their intervention.
Once the calls are identified and verified, the QA evaluation process can be partially automated. The speech analytics application sends the verified calls to the QA solution and automatically fills in many of the fields in the QA evaluation form. The completed fields include the call metadata, agent name and indicative information, customer name, address, account number (if allowed by the organization's Payment Card Industry policy), date and time of call, and purpose of call (as identified by wrap-up or speech analytics). Precision monitoring also rates the agent based on the predefined searches. However, QA specialists and supervisors can override these automated evaluations and provide comments, if there is a need.
The value and contributions of quality assurance programs have been proven over the years. The challenge has been to find time for supervisors to perform this manual and often tedious task. As supervisors have been given additional responsibilities, QA has fallen to the bottom of their priority list. Speech analytics–enabled precision monitoring changes the dynamics and greatly increases the value of quality assurance. Using automation to review 100 percent of calls, it gives managers deep insights into call trends and customer needs. Speech analytics frees supervisors from the tedious aspects of QA, and provides them with the information they need to coach agents, instead of wasting much of their valuable time listening to calls.
Donna Fluss (email@example.com) is founder and president of DMG Consulting, a leading provider of contact center and analytics research, market analysis, and consulting.
Increased targeting and precision make an upgrade worthwhile.