Case Studies

CallRail Amplifies Conversation Intelligence Leadership with AssemblyAI’s Speech-to-Text API

With AssemblyAI, CallRail has seen call transcription accuracy improve by up to 23% and has doubled the number of customers using its product.

CallRail Amplifies Conversation Intelligence Leadership with AssemblyAI’s Speech-to-Text API

The Challenge

CallRail is a leading SaaS solution for small-to-midsize businesses and marketing agencies that provides marketing analytics products, including call tracking, form tracking, and Conversation Intelligence alongside a centralized business communications solution. CallRail helps their customers understand which marketing campaigns are driving the best converting conversations with leads along with real-time data to improve customer service, so that they can efficiently invest their budget and close more business.

In order to scale these best-in-class marketing analytics features, CallRail was looking to partner with a modern Speech-to-Text API that would improve Conversation Intelligence transcription accuracy and quality.

The company’s needs, however, went beyond basic call transcription. To lead in the field of Conversation Intelligence, CallRail was also looking to boost its PII Redaction features, aggregate reporting accuracy, and ways of categorizing and qualifying data.

“For many of our customers, knowing that a conversation happened is the first step,” explains Michael Cantrell, former Director of Product at CallRail. “Being able to know the contents of that conversation at a glance can allow them to truly unlock the knowledge of their business and how their customers view them. It can help our customers better close deals by speaking in the same language their customers are using and refine their marketing efforts.”

The Solution

CallRail’s search led them to AssemblyAI, a Speech-to-Text API with high accuracy, an extensive feature list, and demonstrated success in the telephony industry.

CallRail’s initial implementation focused on two main AssemblyAI features: Automatic Transcript Highlights and PII Redaction.

AssemblyAI’s Automatic Transcript Highlights feature automatically detects key phrases and words in the transcription text. This feature enabled CallRail to automatically surface and define key content in their phone calls to their customers -- such as specific customer requests, commonly asked questions, and frequently used keywords and phrases. It also provides CallRail users with a better understanding of the context of a phone call “at a glance.”

AssemblyAI’s PII Redaction Feature automatically detects and removes sensitive data found in transcript text. This could include social security numbers, credit card numbers, personal addresses, and more.

Together, these features enable CallRail to deliver a powerful Conversation Intelligence solution to its customers.

CallRail was equally thrilled to find a partner in AssemblyAI, whose team of developers and deep learning engineers worked closely with the company to optimize the API for CallRail’s needs.

“We were able to work directly with AssemblyAI on our rollout and had a relatively simple cutover,” explains Cantrell. “The ongoing support was and is still excellent. AssemblyAI has felt like a true partner all the way through.”

The Results

With AssemblyAI, CallRail has seen call transcription accuracy improve by up to 23% and has doubled the number of customers using its product.

The feedback loop between the two companies has also been a huge positive. For example, when the COVID-19 pandemic started appearing in conversations, the new term wasn’t in AssemblyAI’s model vocabulary. CallRail passed this information along to AssemblyAI, whose team was able to add it and increase overall transcription accuracy with a quick turnaround. The advanced automatic punctuation and casing produced by AssemblyAI’s API has also made a significant impact on the readability of CallRail’s transcripts.

As a deep learning company, AssemblyAI is committed to continuous improvement and innovation. Its dedicated in-house research team pushes accuracy updates to production weekly, including a recent update to its architecture to a Recurrent Neural Network Transducer (RNNT) model, increasing accuracy, to industry-best levels.

”This approach gives us the peace of mind that our product will always be providing great transcriptions. And we can utilize newer features to bring even more powerful analytics to our customers,” says Cantrell.

This includes plans to integrate AssemblyAI’s new Topic Detection and Content Safety Detection features into the CallRail platform in the near future, providing additional confidence for customers in their ability to improve customer experience and call conversion rates with the solution.