Customer Case Study: LoyaltyOne - Databricks

Customer Case Study

LoyaltyOne, Inc. is a global provider of loyalty marketing and programs to enterprises in the retail and financial services industry. AIR MILES is their flagship product and Canada's largest loyalty program that serves over 11 million households.

Vertical Use Case

1:1 Conversational Marketing – A recommendation engine that provides optimized offers that enable partner retailers to deliver the right offer at the right time to motivate customer behavior.

Technical Use Case

  • Data Ingest and ETL
  • Data Warehousing
  • Machine Learning

The Challenges

  • Vast amounts of data – millions of transactions from dozens of retailers, 100+ partners, 500 million emails/year, 1200 campaigns/year, and 11 million households served.
  • Lack of Spark expertise – Struggled to make Spark accessible to a large and diverse analytics team that had a range of skills and needs.
  • Slow processing performance – Netezza data warehouse was very slow and limited their ability to do more complex analytics on their historical data:
    • A simple select statement would take up to 4 hours to complete because they must analyze the entire historical data set each time.
    • Did not allow analysis of both historical and real-time data.
    • It was not a scalable solution for data science and productionization.

The Solution

Databricks provided LoyaltyOne with a unified analytics platform that simplifies and accelerates ETL and empowers their data science organization to collaborate via interactive notebooks to build, train and deploy machine learning models.

  • Unified data engineering and data science to ensure an efficient analytics pipeline from ingest to production of ML models.
  • Collaborative workspace and interactive notebooks improved cross-team collaboration, allowing them to greatly accelerate model prototyping for new features.
  • Simplified infrastructure management and reduced operational costs through automated cluster management and cost management features such as autoscaling and spot instances.

 

The Results

  • Simplified infrastructure management – They don’t have to waste time provisioning clusters. Self-service cluster management with auto-scale/auto-termination of clusters helped reduce costs.
  • Improved collaboration – Notebooks make it much easier to share work. The interactive nature of the workspace enabled them to support multiple user types across the organization.
  • Strong partnership – Databricks provides the level of support they need to speed time to market.
  • As a result, they were able to increase offer response rate by 2x with a 97% improvement in speed.

Databricks has provided us with the support and technology to modernize our architecture, enabling us to do data science at massive scale.

Bradley Kent, AVP, Program Analytics at LoyaltyOne