Using RudderStack To Power Your Machine Learning Models

Here at RudderStack we talk a lot about data, analytics, technical tools and occasionally football and cricket. We are self-proclaimed data nerds, and nothing gets our juices flowing quite like the topic of machine learning. For us, machine learning combines the critical sciences of engineering, programming and analytics with the creative (some might argue dark) arts of intuition, experience, language and behavioral psychology. But building a system to leverage and integrate Artificial Intelligence or Machine Learning is hard and can sometimes feel overwhelming from both a programming standpoint as well as a data collection one.

It should come as no surprise that we have a number of experts in many of these contributing fields and as a result have been able to prototype a handful of machine learning models as well as support a number of our clients in architecting their own. One of our goals at RudderStack is to considerably reduce the barrier to entry for teams who want to apply machine learning within their organizations. Here are a few examples of how you might use RudderStack to unlock the power of ML with your own data.

Enriching Zendesk Using Google’s Sentiment Analysis AI

The team here at RudderStack built a sample project outlining how to use Google’s Natural Language API tool to enrich inbound tickets from Zendesk using Google’s Sentiment Analysis methods. The Google Sentiment API allows you to access ready-to-use, pre-trained natural language processing models to perform effortless sentiment analysis for your organization. You can leverage that API in real-time with RudderStack’s Transformations to enrich inbound track and identify calls. We will have a deep-dive walk-through soon, so if you would like a sneak peak on the step-by-step tutorial we are putting together, please send us a slack (join here) or hit us up on Twitter!

Predicting Customer Churn for Wynn Casinos Gaming App

We recently wrote a blog post on how RudderStack helped Torpedo Labs predict player churn for their Wynn Casinos app. Torpedo uses Rudderstack to capture player event data and forward it to BigQuery. Within BigQuery, a standard linear regression model was trained and tested using data in the Tracks table and customized event scoring. The net result was a major reduction in churn and an increase in daily revenue and player engagement.

LoveHolidays Property Suggestions

LoveHolidays is an online travel booking site in the UK using RudderStack to stream user event data from its website and mobile app to Google BigQuery. They have built a machine learning algorithm in Python to predict which vacation properties will best suit an individual traveler based on his or her prior search history. The model gets retrained every 24 hours, the results of which are fed into a Redis database via RudderStack Reverse ETL. When the user returns to the site or app, suggestions of available properties can be instantly promoted and/or refined, delivering a more relevant user experience that requires less direct searching. For Loveholidays, the result is a higher booking per user search and a lower search per user ratio.

If you or your team would like to know more about how RudderStack can support your AI/ML needs or would like to see a demo of RudderStack in action, subscribe to our product newsletter below or connect with us on Slack to be sure you get the inside track on what we are cooking up.