Big data startup applies deep learning to customer churn prediction

Big data startup Wise Athena has presented this week their novel approach to churn prediction based on deep learning technology. This way, the San Francisco based company becomes the first company to apply deep learning to customer churn prediction.

Churn, defined as the loss of customers to competitors, is currently one of the most pressing challenges for companies.

It is estimated that the telecommunications sector in particular loses $10 billion per year due to customer churn and, in general terms, customer acquisition is five to six times more expensive that customer retention. Being able to predict churn in advance has become a highly valuable insight in order to retain and increase a company's customer base.

Wise Athena is a pioneer in the application of deep learning to customer churn prediction.

Deep learning is a machine learning method capable of automatically extracting patterns across input data. This technology allows for the recognition of features it has never been trained on and has already been successfully applied to complex data abstractions like speech recognition or computer vision.

“The ability to combine the complexity of telecommunications data with recent innovations in deep learning will quicken and deepen the pace of telecommunications innovation. Siri like customer care will become commonplace. Discovering new uses for data will regularly occur.

Telecommunication Industry really needs a revolutionary application, or a whole new way of thinking about data, and we want to be a part of that” has declared Alfonso Vazquez, Wise Athena´s head of products.

Wise Athena´s large scale application of deep learning to real life customer datasets has shown that this technology can lead to significant increases in churn prediction accuracy while reducing the huge costs normally associated with the traditional methods consisting of timeconsuming feature engineering.

Download the latest Wise Athena WHITEPAPER: "Using Deep Learning to Predict Customer Churn in a Mobile Telecommunication Network" from: