Speech analytics is inarguably one of the most valuable tools that businesses can deploy to understand the changing requirements of their customers and monitor the day-to-day performance of their staff. High-volume contact centers record literally thousands of hours of daily audio conversations. The right speech analytics solution gives companies access to an untapped treasure trove of available customer interaction data. They can use this critical information to proactively make vital business improvements.
But for a speech analytics solution to be of any bottom-line value, it must be able to process large volumes of recorded data at high speeds. It must also offer a high level of accuracy and ease of use. The cost of implementing speech analytics, which can run as high as $1,000,000 or more, has been a barrier that has kept SMBs and many enterprise organizations from advancing beyond the investigative stage.
Despite pricing obstacles, the market for speech analytics stands at well over $500 million and is projected to almost triple to $1.6 billion by 2020 according to a MarketsandMarkets estimate. Now, a next-generation solution is emerging that could ultimately shift the balance. Yactraq, an audio mining and speech analytics company, employs advanced technology to deliver business intelligence using recorded phone calls as input. Their platform can be used to quickly and accurately search through tens of millions of hours of call center recordings and provide actionable insights and intelligence which can be obtained on a more cost-effective basis than anything currently available.
How is Yactraq different from existing phonetics-based solutions? “Our forte is customizing vocabularies,” said Jeh Daruvala, Founder and CEO of Yactraq Online. “We provide omnichannel BI through OmniTraq, a suite of features powered by our proprietary CoreTraq platform. It can generate and use massive customized vocabularies which we help companies to adapt to meet their unique needs. We can build vocabularies specific to a client and vertical consisting of tens of thousands of words in only a matter of days. Most legacy speech detection products cannot scale beyond 500 words, whereas we can ultimately scale up to 300,000 plus words.”
Major speech analytics solution providers focus on the enterprise market, mainly dealing with contact centers of 500 seats or more. Others place their emphasis on operations of 100+ seats. “Our sweet spot is between 10 and 100 seats,” noted Daravula. ”With annual subscriptions in the range of $25,000, any 15-seat contact center can now afford a high-quality audio mining system.”
CoreTraq is fully developed and proven having processed over 4 million minutes of customer data in OEM capacities. Yactraq has a pending patent for automating the generation of language models in speech recognition systems. This process includes techniques to automatically build speech recognition language models using the linguistic data within a Natural Language Understanding (NLU) engine. The company entered a long-term partnership with invention advocates Intellectual Ventures in 2015 to develop a proactive intellectual property strategy through its IP-for-Defense program.
Yactraq has a beta in progress with CanaDream, a leader in the Canadian RV rental and sales industry. The client was reportedly impressed with CoreTraq’s accuracy which ranges from 80% to 99%. Yactraq was also named a 2016 Gartner ‘Cool Vendor’ in Smart Machines, and was among the first 20 companies showcased in IBM Watson’s Cognitive Gallery.
While Yactraq does not include a call recording function, it is agnostic and can be integrated with most existing solutions. Daravula believes the platform can be a game changer in quality assurance functions. “Agent evaluation in customer service management is a critical business insight for contact center operations,” he said. “But when it comes to QA, many customer service organizations are still in the Stone Age. At present, call recordings are analyzed at great cost by a few human quality analysts who usually cannot get to more than a 2% to 4% random sample of the data, which means that the vast majority will never be heard. We offer organizations the opportunity to pull all the business information out of this ‘Dark Data. If four managers are spending 10 hours a week within a company trying to find the right recordings, they can now find what they need in a fraction of the time: it’s like getting back 40 man hours per week.”