Your contact center and the teams that staff it are ultimately responsible for delivering a consistent and positive customer experience. It’s your agents who are vital to your ongoing business success, with the power to create loyal customers or destroy the relationship with one wrong move.
The reality is that 68% of customers would consider leaving a business after a bad experience with a call center, according to Call Miner. And that’s just one reason why it’s essential for contact centers to continually look for ways to improve the customer experience. There needs to be a plan and software in place that makes it easy to perform effective quality assurance on every customer interaction.
That’s why, in recent years, speech analytics software and artificial intelligence have become popular buzzwords in contact centers, positioned to solve problems that will promote effective call center QA and holding out the promise of addressing three QA challenges:
Speech analytics is a technology-based approach that contact centers may use that combines speech recognition software with text analysis and pattern spotting to categorize interactions according to a set of custom rules.
Speech analytics software reviews every customer conversation automatically once it has been recorded, and translates it into machine-readable text. Speech analytics vendors offer contact center managers the ability to mine unstructured voice data to identify interaction trends. This can then theoretically be used to understand and interpret the text and determine the performance and accuracy of the agent(s) interactions with customers.
This technology can conduct an in-depth search based on phonetics and theoretically even has the ability to detect emotions. It can monitor call trends such as hold times, silent patches, or if agents are talking over the caller.
The key force driving speech analytics is the need for more efficient call center agent performance measurements. According to an Opus Research Survey, 72% of companies believe that speech analytics can improve the customer experience, 68% see it as a cost-saving mechanism, and 52% believe it can improve revenue.
But is the adoption of speech analytics a good thing? Does speech analytics really deliver? The main challenges in deploying a speech analytics project are:
If the solution is on-premises, it is easier to deliver the audio or text from your recording platform(s) to be processed. However, cloud SA offerings require large amounts of audio/text data to be transferred to a remote site for processing, which typically introduces a delay and may raise IT security concerns.
Tailoring the engine to categorize and assess the interactions for your industry and more specifically your organization, will require significant input from your vendor in the form of professional services and close cooperation by your in-house team to configure the engine with what it needs to accurately categorize and assess your client/agent interactions.
Mapping the analyzed interactions to a scoring framework can be challenging depending on the complexity of your scoring matrix. Furthermore, the level of granularity you are trying to achieve will determine the amount of real evaluator time needed to complete the scoring, add coaching tips or apply root cause analysis.
It’s a predictive technology that relies on algorithms and thousands of samples to identify trends and make predictions on what someone will and won’t do. The problem is that voice characteristics can vary quite a bit, so unless you have a large enough sample size, the analysis may not be accurate.
There’s a lot of potential, depending on how many parameters are measured—the Voicesense algorithm for instance measures over 200—but while the technology has now been around for more than 15 years, it still isn’t close to 100% accurate. It cannot replace the judgments of real human listening and scoring a call. It also cannot provide coaching/training based on the data. That is up to your team.
So if there is one caution, it’s that scoring agents who listen to call samples and score agents using a QA scorecard based system, cannot be replaced with speech analytics. Organizations need both the funding and sophisticated technology to record and translate recordings to text and then analyze calls over time.
While speech analytics and AI cannot and should not replace your current contact center QA system, it does have a place assuming you have the technology and the resources to implement it.
The benefit of speech analytics to improve call center QA, is that it can listen to 100% of the conversations and extract information about your customers’ preferences based on statements mentioned or actions taken. It can then take this information and make suggestions based on trends and account details, which help your team deliver a better customer experience. But in order for this process to be effective, it should be incorporated with your current QA process for monitoring and scoring calls.
The true value of speech analytics is the ability to track every call and not just a sample, to look for violations, risky language, and behaviors that lead to both success or failure. It will categorize these calls into positive, negative, or neutral interactions so that your evaluators can focus on specific categories of calls for further analysis. Combined with human insight into agent interaction via voice/call, chat, or email, you can create a sophisticated call center QA process that offers real and measurable feedback on improving the customer experience.
Much of the value provided by speech analytics comes from how it’s integrated, which is a three-phase process:
First, for speech analytics technology to work accurately, you have to perform categorization with machine learning. This step requires the software to categorize words, acoustics, and sentiments, by assigning meaning to unstructured voice communication. This is a critical step where the software automatically tags and analyzes thousands of communications to identify topics, patterns, tone of voice, and other variables. This will require the vendor and the in-house team to tune the output from the initial automated process.
The next step is call scoring. Based on the categorization and custom vocabulary you set up, speech analytics software can automate your call classification efforts. It offers real-time feedback for your agents after every call. During this phase, managers should review the auto-scoring to look for areas where agents can improve as a whole or individually, while also checking the software for accuracy. This can work well particularly in the greeting, IDV and closing phases of the call.
The final stage of speech analytics integration is QA system training in real-time. This needs the system output to map a scorecard that reflects your organization’s specific customer and process success criteria. Contact centers need to establish these criteria to ensure that the right scripts are in the system with the appropriate phrases or keywords flagged, to alert both management and the system. The goal is to set up the software to quickly identify when agents require training or more grounding in best practices, as well as to recognize factors, processes, and behaviors that lead to success.
Overall, speech analytics does add to the QA toolkit and with careful planning and deployment can become a valuable piece of the QA process in contact centers. Currently, it is an expensive solution and achieving a positive return on investment requires the vendor and client to work closely together over a period of time to reach goals and objectives. But, speech analytics should not be used alone. It currently cannot replace human observation and insight, and as such, should be integrated into your existing QA scoring and reporting process for the best results. With the current state of the technology, speech analytics can be very effective at identifying and categorizing calls, emails, and chat threads for further analysis and tagging. It remains to be seen what impact artificial intelligence (AI) may have going forward in helping the technology decipher the nature and outcome of complex customer interactions in greater detail and subtlety.
Deriv is a customer-focused fintech dedicated to offering accessible trading solutions to people all over the world.
Scorebuddy BI helps their QA Process.