Speech Analytics is the future of quality monitoring for Call Centers, and Callzilla is jumping right in! We have partnered with Observe.AI to automate tedious quality assurance tasks and receive faster actionable insights on a broader spectrum. Observe.AI performs a deep analysis of each audio recording, and gives us human-like insight on the result of that call. Our QA Monitors are able to quickly identify pain points based on the system’s analysis, and focus their time on providing Agent feedback and improving processes. We’re going to give you a run down on how the process works.
While our Quality Assurance Team was only able to listen to 10% of our customer interactions manually, our Speech Analytics tool is able to monitor and analyze 100% of our recordings 3x per day. That in itself has been a huge process improvement because our sample sizes are now all inclusive. Also, instead of our Monitors spending time grading all types of calls, they are able to focus on the pieces that require improvement by pulling out what we call “moments”.
The tool will analyze a call for us, but we input moments to tell the software exactly what it is we’d like to analyze. Each moment is made up of multiple key words that can be triggered based on if the key words are present in the conversation or not. When the tool recognizes a moment, it flags the call and also redlines that portion of the audio so the live QA Analyst knows exactly where to look.
The Speech Analytics system comes with standard moments, such as Customer Satisfaction, but you can also add custom moments with your own key words. As a Call Center with multiple clients, it’s important for us to have a customized measurement system. While one client may focus on Satisfaction moments, another may need reporting on specific script adherence. Observe.AI’s platform makes it easy to analyze general metrics across all programs, and also to customize the analysis of each program’s experience.
You’re able to preview a moment within a few seconds using historical transcripts, to verify how effective your key words will be before publishing to your dashboard. When adding new moments, you’ll see that data can be pulled from past recordings as well as your future calls. This is another perk of using Speech Analytics, because changing anything on your quality scorecard would only apply to future calls if you were monitoring manually.
That being said, it’s important to think carefully about the key words you’re choosing for each moment. The tool is only as good as what we ask of it! With Observe.AI, you’re able to assign a “super user” to lead the approval process for each moment. Any Supervisor, Operations Manager, or Quality Analyst with a system user can request to create a moment, but the super user (who is the Quality and Training Manager in our case) must review and approve before it’s applied.
Moments can be categorized as positive, negative, or neutral in the system. If sorted as a positive moment, the goal for that metric will be to reach a higher percent and the more key words the system finds, the better. For example, script compliance would be a positive moment. We may set the goal to be 90% compliance, so we are looking for those key words from the script to be present in 90% or more of our calls.
A negative moment is just the opposite; we are searching for those key words with the hopes of finding minimal alerts. For example, Customer Satisfaction is actually categorized as a negative metric. The key words we input are terms of frustration, and the system can also measure negative tone. The goal for Customer Satisfaction % is to be as low as possible, because if you have, let’s say, 35% Customer Satisfaction it means that the system recognized customer frustration on 35% of your calls. We find it easier to track negative customer sentiment than positive customer sentiment, because the negative tone and keywords stand out more in conversation.
Neutral moments are just for tracking information, and not necessarily agent performance or customer satisfaction. If the client wants to know how many people are asking about their new promotional offer, we can add a neutral moment that’ll let us know how many times those key words are mentioned during each call.
Your positive, negative, and neutral moments are then placed in four measurement categories within your dashboard: Customer Satisfaction, Process Adherence, Compliance, and custom moments. We covered customer satisfaction above, and process adherence would be any moment relating to how the call was handled by the agent. For example, we’re looking to see if the opening of the call was handled correctly, if the agent offered the survey, or if they cover shipping on the call. Compliance moments are looking for the important aspects of quality for your clients, like offering empathy or the lack of dead air during the call. You’re not tied to these three categories, you can create any custom moments outside the box.
Now that we’ve created moments, what do we do with them? You can find all of the moments you created in percentage form on your Observe.AI scoreboard. The scoreboard shows your general KPIs across all programs, and you can also break it down by program or moment for any date range. This dashboard is useful for general performance tracking, and we give this data to our clients during weekly business reviews.
Keep in mind when looking at this overview that these metrics are not 100% indicative of agent performance, but more of the program as a whole. As an example, we measure Negative Customer Sentiment. If the score is high, it means that the customer is frustrated but doesn’t indicate the source of their frustration. This is where our live Quality Analysts come in; they would listen to these calls to find that the source of frustration may be policy related, system errors, or maybe agent provoked. In either case, the Quality Analyst is able to give feedback on how that call or process could be improved.
We’ve always used Quality Assurance to improve agent processes and performance, as all call centers do. What we’ve found with Speech Analytics is that it gives us clear picture of the whole customer experience, and we’ve been able to make general process improvements as well. We launched Speech Analytics in July 2020, and the first metric we focused on was improving average handle time. With this new system, we were able to quickly identify patterns behind the calls exceeding handle time goals, which were mostly related to agent’s product knowledge and efficient call handling. By August, our AHT overall decreased by 2%. AHT has always been a metric that we keep an eye on, but we saw great improvement when the system does the analysis for us, and leaves our QA and Training teams to focus their efforts on giving the feedback and creating action items toward improvement.
The software transcribes the call recordings into written text, and that’s how it’s able to identify and analyze key words. This is also useful for our Analysts, who can review transcripts and manually search for key words without having to listen to full call recordings. If you don’t want to create a full moment, anyone with access to the system can search a key word within the transcripts section. You can also search the transcripts by moments or categories, so it’s easy to filter your search when looking for a specific customer experience. The speech to text feature is not 100% accurate at this point, but Observe.AI does have a machine-learning component that continually improves this function.
QA Process and Calibration
As mentioned earlier, the Speech Analytics system will flag calls where designated moments are recognized, and then redline that specific segment where the moment was triggered. Our Quality Analysts then go in and either listen to the call recording or read the transcript, and input their feedback directly into Observe.AI’s platform. There is a scorecard for each call that the Analyst manually evaluates, and also a space to input action items for the training team if coaching is necessary. During this process, the analyst can also calibrate with the system, ensuring the system analysis is functioning the way it’s designed to be, or updating moments if not.
It’s so convenient to have the Speech Analytics and QA scorecards all in one place! Observe.AI also stores call recordings, and Analysts are able to download them as normal or in 1x speed. The scorecards and feedback are organized in a way that you can search for evaluations per agent, per team, and per time period. We’ve found this to be a quick and easy system for evaluating improvement in our agents!
After the initial set up, Observe.AI’s integration with new programs is actually quick and easy. All we need is for our client to agree to place all call recordings in a safe ftp where we can import them on a daily basis into Observe.AI. Our platform is also PCI compliant, so there is no cause for concern with your customer’s personal information. Based on our experience with Speech Analytics so far, this small process update is well worth the knowledge and analysis gained!