Predictive Analytics Milestone: WordSentry's PROPHESYTM Predicts Customer Churn in Real Time

INDIANAPOLIS - Oct. 25, 2016 - PRLog -- WordSentry, a provider of communications analytics, has made a major breakthrough in predictive speech analytics, creating a model that predicts when a customer is about to cancel a service or product, during a support conversation, analyzing text at a higher level of comprehension than when considering custom key terms.

Solution to a vast problem

Annual attrition in the US is projected at over $680 billion, with the cost of replacing an existing customer several times that of maintaining an existing one. With customer support being a critical component of retention, the predictive capabilities of PROPHESY during these conversations – whether voice, email, chat, or text – will dramatically reduce attrition and avoid unnecessary retention offers, while identifying at risk customers and effective agents.

The technology was developed by a team of scientists, incorporating deep learning, leading-edge artificial intelligence and natural language processing, and was trained and validated on millions of actual conversations, all of which were scored as churning or non-churning.  The breakthrough, resulting from incorporation of such a vast trove of training data, allows the model to transcend metadata requirements, drives predictive accuracy up to 90%, and delivers analysis in milliseconds.

"It would require 16,000 to 20,000 years for a customer support agent to gain the experience and knowledge of PROPHESY," Dr. Johan Bollen, lead theorist on the team, explains, "allowing the model to warn of a client possibly attriting before the agent has even contemplated an appropriate response."

Why a milestone

Existing churn recognition technologies are brittle and inefficient; they require the tagging of metadata to individual products or services and the development of custom lexicons, including generic words indicative of churn (I.e. "cancel," "quit," "frustrated," etc.).  Resulting analyses then access custom algorithms to look for these words and/or phrases in a conversation, compare to how the same were previously used, and attempt to predict the outcome. The development of each model is expensive, time-consuming, and only marginally accurate, with results delivered at the end of the conversation – too late to prevent cancellation.

"The predictive capabilities of PROPHESY could be a game changer in this industry," says Scott Bakken, a Principal at http://www.maintrax.com, a leading speech analytics services company. "Having worked with many of the leading communications analytics providers, it's refreshing to see the sophistication of the analytics, even though clients can access it through a simple, easy-to-use interface."

About WordSentry (visit http://www.wordsentry.com)

Every day, customers and clients communicate with providers in multiple ways, via phone, chat, social media, and email. In the absence of face to face inflective cues, participants often misinterpret emotion, intent and motivation. WordSentry provides speech analytics models, identifying the intent of communicators and outcome of conversations.