Machine learning comes to call analytics

New Invoca capability promises language pattern recognition based on the entire call vs. isolated keyword spotting.

Chat with MarTechBot

Ai Head Devfc4

The term “AI” is everywhere these days. And it’s often difficult to separate the marcomm from the actual martech.

Invoca CEO Gregg Johnson assures me that the company’s new Invoca Signal AI is the latter. He says that the company is bringing machine learning to call analysis for deeper customer insights.

In contrast to many marketers’ perceptions, there are more, not fewer calls being generated today via mobile devices. Johnson told me that there are roughly 100 billion calls being made annually, yielding potentially valuable data for marketers. That data goes beyond basic close rates and includes details about customer experience, product attributes and the effectiveness of various marketing channels.

Just as location data is increasingly important for analyzing physical-store visitation, call analytics offers another category of offline attribution.

Inovca’s Signal AI analyzes calls in real time and expands upon what previously was a practice of simple keyword spotting. For example, call analytics tools have historically been trained to recognize isolated keywords in call transcripts, such as “order,” “credit card” or “price quote,” depending on the marketer’s objective. Inferences about call quality or conversions were derived from the presence (or absence) of those words and phrases.

Inovca’s new methodology uses natural language processing to analyze the entire conversation. Johnson explained that the company has out-of-the-box algorithms, tuned by vertical, with a set of classifiers (phrases) that can be customized and adapted over time. For larger customers, the process can be entirely customized.

These predictive classifiers are based on analyses of millions of historical calls in verticals such as automotive, insurance, financial services, healthcare and more. Johnson says Inovca is now using sophisticated language pattern recognition to understand the content of calls.

“The system relies on speech to text initially,” explained Johnson. However, as it learns over time, he said, “The algorithms can determine the effectiveness of the calls and outcomes.”

While it may be evolutionary vs. revolutionary, Invoca’s Johnson argues the company has built the next generation of call analytics, which will give marketers a better sense of which calls are effective and why. In addition, it can provide marketers with visibility on what channels are driving higher-quality calls. Marketers can also do retargeting based on call content, if they fail to close.

To that end, Invoca says that its analytics platform is integrated with 30 other martech tools and companies (including CRM and bid management), which then enable “actionable data from live conversations to improve marketing spend and the customer experience.”

Often, new technology solutions are more sophisticated than marketers themselves. Johnson told me that about half of Invoca’s customers need the preset call classifiers, while the other 50 percent have data sufficient to build their own classifiers from scratch.


Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.


About the author

Greg Sterling
Contributor
Greg Sterling is a Contributing Editor to Search Engine Land, a member of the programming team for SMX events and the VP, Market Insights at Uberall.

Get the must-read newsletter for marketers.