How CallRail built CallScore: Using AI to make better marketing decisions
Nearly three years ago, CallRail’s Conversation Intelligence (CI) team started a journey into uncharted territory. The challenge: Use machine learning and AI to create a tool that makes classifying the quality of a phone call completely automated, saving valuable time for our users.
At CallRail, our CI team is dedicated to giving businesses of all sizes access to AI-based services — technology that can help you optimize marketing efforts, save valuable time in analyzing quality of communication with customers, and gain a larger understanding of your customer journey.
We’re proud to say that we emerged out of those years with a strong offering of features unmatched by our competitors in terms of ease-of-use, cost, and accuracy. And it was the very process of building CallScore that gave birth to other powerful CI features like Transcriptions, Automation Rules and Call Highlights.
CallScore: Using AI to deliver hassle-free call analytics
CallScore set out to solve a huge concern for our users: Many were spending countless hours reviewing every call that came into their business, or even hiring analysts to do this around-the-clock. We knew we could build a solution that could help save valuable time by automatically predicting the quality of a customer. At the same time, we knew we needed to build something that works for our customers out-of-the-box, with as little hassle as possible.
To start, we built a dataset of hundreds of thousands of phone calls, from every type of industry and customer. And to make sense of it all, we hired a dedicated team to analyze phone calls and organize the resulting data. Since we began building CallScore in 2016, we have created a massive human-scored dataset, while also experimenting with countless methods of natural language processing, text analysis, and classification.
After building countless prototypes, we finally came up with the right architecture and an even more accurate scoring algorithm, resulting in the CallScore engine our customers use today.
CallScore is powered by a machine learning model that both transcribes a call and extracts ‘events’ from the audio. Once a call is completed, our system transcribes it and we pass that transcription — along with other call metadata — to our machine learning model. Within a millisecond, the algorithm analyzes the conversation, classifies it, and tells the user if the person is a qualified lead. CallScore can save you time and money, and help optimize your marketing by determining which campaigns are driving your highest-quality leads.
But our work doesn’t stop there — we’re constantly looking for ways to improve our algorithms and accuracy as we continue to analyze more and more conversations. (As of this article, we have automatically classified over 5 million phone conversations!). CallScore is a great foundation for the many other CI features you can use to streamline and optimize your marketing process.
Adding CI features to a solid foundation
When reviewing a phone call, it can be very difficult to determine the most relevant parts of the conversation. That’s where Call Transcriptions and Automation Rules come in — while our team was building and refining CallScore, we knew there was another gap to fill that could help save even more time by analyzing the quality of a conversation.
And so, using machine learning and natural language processing (NLP), we created Transcriptions to automatically transcribe every phone call that comes into your business, giving you an at-a-glance look at those conversations without having to listen to thousands of phone calls.
Much like how Apple’s Siri can determine what you are asking her in real-time, Transcriptions can do the same with your phone calls. When a call ends, the CallRail's engine analyzes the audio and generates a transcription in return. It’s a pretty amazing to consider that Transcriptions can analyze audio signals, gather possible words that were spoken, and weigh their context — all within minutes of the call ending!
Furthermore, Call Highlights works in tandem with Transcriptions by automatically finding relevant words within the conversation, giving you an at-a-glance look at each call without having to click the play button. Are you looking to take action when a specific word or phrase is found in a conversation? Automation Rules is the ticket.
With this feature, conversations can be tagged, marked as a qualified lead, or given a value based on the specific key term sets you configure, giving you even more control over the automation process. (Check out how one company was able to save hundreds of hours by using these features!)
CallRail’s Conversation Intelligence features are here to stay, and our team is working around-the-clock to continue to enhance and improve the service, while also delivering new features that all support our central mission: Using AI to help you save time and optimize your marketing efforts.
We are excited for the new features that are coming down the pipeline in the near future, so stay tuned for more Conversation Intelligence updates in the coming months!