Key Takeway

Online behavioral data can be used to predict customer lifetime value and churn, but current models may require costly manual feature engineering and can accommodate only shorter data sequences. A model combining automated log processing with several representational learning frameworks can improve predictions while reducing the time and cost required.