Four months ago we announced the launch of our Quality Check feature, designed to save our clients time and money by using machine learning to detect potential low-quality Leads and automatically submitting them for review.
For our initial launch of Quality Check, we chose to focus initially on auto-detecting 2 types of calls:
Together, these 2 call types represent 43% of all low-quality calls in our platform.
“It's been really exciting to watch Quality Check find those leads that we want to credit back to our clients with lightning fast speed. We love this human enabled smart technology because it means we get to ensure our clients have quality leads by harnessing this powerful technology but also ensuring that humans verify it's working correctly!”
Alycia Cameron, Quality Assurance Manager
So what has happened since launching Quality Check? Let's take a closer look at what impact it has had for our clients since its launch.
Within our system, spam calls are calls that serve a pre-recorded message trying to sell something. Your business and personal phones probably get these types of calls daily. Unfortunately we’re not immune to the same predatory, annoying practices. Fortunately, Quality Check is doing its job, making it easier for our clients to deal with spam.
Just looking at the last 30 days, Quality Check has been able to automatically flag almost half of all calls discovered to be spam within our platform.
Quality Check is only as good as its ability to accurately determine the nature of a call, and fortunately our engineers have built a pretty darn smart machine (that’s getting smart, but more on that later). 80% of the time Quality Check automatically suspects a call is spam it ends up being correct.
The largest type of low quality calls we see within our platform are considered to be ‘wrong number’ calls. These can be wrong numbers for a variety of reasons, but ultimately they are calls where the caller is either trying to get in touch with another company or a person that is not associated with the client who received the call.
These types of calls represent more of a challenge for machines, because the patterns within the calls are more wide-ranging. As such, our algorithms will continue to need more time to learn how best to detect these types of calls, but already Quality Check is automatically flagging 38% of wrong number calls within our platform.
Despite the challenges in detecting these types of calls, Quality Check has an accuracy rating of 88% when it comes to wrong number calls within our platform.
To date, Quality Check has automatically flagged almost 4,000 calls within our platform. These calls totaled over 72 hours in duration, meaning our Clients saved that much time having to listen to calls in order to determine whether to dispute them.
Best of all Quality Check, like any well-built machine-learning system is…well, learning. That means it’s getting smarter at knowing when to flag a call as either spam or a wrong number, and it’s getting more accurate when it does.
Spam Catch Rate First Month After Launch Versus Most Recent Month
| First Month | Most Recent Month |
Spam Prediction Accuracy First Month After Launch Versus Most Recent Month
| First Month | Most Recent Month |
Wrong Number Catch Rate First Month After Launch Versus Most Recent Month
| First Month | Most Recent Month |
Wrong Number Prediction Accuracy First Month After Launch Versus Most Recent Month
| First Month | Most Recent Month |
We're incredibly excited about Quality Check's immediate impact for our clients. Even more exciting, we believe these efforts barely scratch the surface in terms of our ability to help our clients more easily manage their Service Direct accounts. So stay tuned as we roll out more features in the near future!