Real-World Results: How to Reduce Queries from 8 Hours to 8 Seconds

How a large manufacturer cut preaggregation to power real-time customer segmentation

A 2020 study found that 94% of customers are likely to purchase more from a firm with a good Customer Experience (CX). This is even more striking when you consider that 80% of customers are NOT willing to purchase more following a poor experience. Delivering a valuable CX is a critical requirement for all organizations to be successful. 

Under the hood, meaningful CX is often driven by a flexible, fast, and impactful approach to leveraging data. Segmenting target audiences based on measurable characteristics and dynamically creating customer cohorts is critical to any company’s success. 

A large multinational manufacturer identified an opportunity to enhance an existing CRM application that contains dynamic audience groups. The team set an objective to move from several hours to generate a customer list to as close to real time as possible. Their audience group includes 100s of millions of individual records and is a central repository for marketing teams to create related cohort lists. The engineering team recognized that decreasing the time it takes for marketers to build a segmented audience list could reduce the overall cycle time and increase their velocity to market. 

Marketing organizations that aim to rapidly deliver personalized ads must manage, analyze, and operationalize data from disparate sources. A major challenge for the customer was data preaggregation and slow runtime performance. Simple segment count queries in Apache Hive took 5-8 hours to execute. And with tables holding hundreds of attributes of varying complexity across hundreds of millions of records, preaggregation was time-consuming.    

With Molecula, the manufacturer was able to integrate FeatureBase into their existing application stack and saw immediate results:

  • Before: 8 hours to generate cohort lists for marketing activities
  • After: 8 seconds to generate intricate audience segments, consisting of several complex queries (often even hundreds at a time)

Adding FeatureBase has greatly improved query speed, helped to enhance overall CRM application performance, and dramatically reduced the preaggregation burden. Now the marketing team is able to identify and retarget users within seconds of a missed opportunity instead of hours or even days, transforming the previously impossible into the possible.


Watch our on-demand webinar: Data Aggregation is Unavoidable! (And Other Big Data Lies)


Start Free Trial