How to Transition from Human Call Scoring to Auto-Call Scoring in 5 Easy Steps

by Emily Nave


Do you record massive amounts of phone calls? Do you monitor agents or evaluate calls to learn more about your customers? Do you currently use humans to score something like 5% of your total call volume? Yes? Yes? Yes?

Then can VoiceBase Predictive Insights change your (work) life? YES. YES. YES.

If you’re still reading this, then you’re ready to replace human call scoring with machine scoring, to quickly and accurately classify 100% of your calls, in one scorecard, at a fraction of the cost. It’s just a matter of when and where. 

I know it’s hard, it must seem like such a scary task, replacing your current system with a new one. Your mind must be running… first of all, what will this cost me?? (No upfront cost, no set up fees) Well, what kind of hardware am I going to need to set up?? (None) And how accurate is this going to be, right now I’m using HUMANS. (Very accurate, in many cases more accurate and more consistent than humans). What? No. How?

With the combination of highly accurate speech to text, intelligent keyword/topic extraction, and predictive insights training models, automatic event detection in spoken content is not only possible, but works really well.

Typically these phone calls are recorded at 8 kHz, and super compressed in order to be stored efficiently. This creates a pretty low quality of recordings, so low that when other variables are in play, such as background noises, car phones, and accents, humans even have difficulty discerning what was said, let alone who said it. 

Luckily VoiceBase has a unique method, which has become resilient to these negative factors affecting the call recording quality. By taking a ‘big data’ approach and analyzing 10s of thousands of columns of data to determine the exact pattern that defines what a HOT LEAD sounds like for YOUR COMPANY, getting a few words wrong doesn’t have as big of an effect on the results. And the results are customized for your business, since a hot lead for you sounds a lot different than a hot lead for an automotive dealer, property management or pharmacists.

Ready to get started and set up a test to see results on your own content? It’s as easy as 1,2,3! (..4…&….5)

Step 1: Request API Access

First things first, create a VoiceBase account here, to begin uploading files to test our transcription accuracy. 

Step 2: Contact us to set up a trial

VoiceBase’s Account Managers and Sales Engineers work with each customer to define the unique events they want to spot such as; hot leads, rude agents, close to cancelling, upset customers, or sensitive information like PCI, PII and SSN.

Step 3: Send us files

The customer then sends VoiceBase 2,000+ pre-tagged calls, half with the indicator and half without, to begin the machine learning process. (Or raw calls, which VoiceBase can get tagged). At this point VoiceBase will provision an API Key for you and your team to get started on an integration. 

Step 4: Review predictive model

Once VoiceBase has built and deployed a predictive model for each specific event, the customer begins sending live calls and receives back the prediction results in JSON format, which include a confidence score. Together this can trigger the automation of business processes. 

Step 5: API integration & sending content

Now that your predictive models are complete it’s time to finish that integration! Our development teams work together to complete sending recordings immediately from your platform to ours. Soon 100% of your calls will be automatically transcribed and scored and the real work begins now that you’re collecting that rich, new data!

What can you detect?

  • Was an order placed?
  • Was an appointment made?
  • Is this a Hot lead or Non-prospect?
  • Is this account about to cancel?
  • Is this a 1st time caller?
  • Positive/negative comments 

Happy Analyzing!

More From the Voice analytics blog

Predictive Analytics for Strategic Insights

Predictive Analytics for Strategic Insights

Predictive analytics is an advanced form of data mining that leverages machine learning to identify patterns in voice recordings, intuit a speaker’s intent, and predict a future outcome — be it a sale, account cancellation, or one of many customized “X” signals your clients might request.

read more