Customer: A large online home festival by a high-end property portal that caters to a global market
The customer was looking to build a highly accurate product recommendation engine to suggest properties to visiting users. They also wanted these recommendations to be time-triggered. The challenge was, almost 95% of users were guest users and hence the prediction engine has to be built using very little user-specific data.
Powerup team built 2 engines, 1. A product recommendation engine that tracks user click-stream data and recommends users with right properties, 2. A prediction engine that predicts the time a user is going to spend on a web-page.
Apache Spark, Socket.io, BigQuery, Riak, Fluentd, Prediction.io.