

Key highlights
Category
Digital advertising
Location
Headquarters: Chicago, IL
Solution highlights
- Modern Data Platform: Cloudera Enterprise
- Workloads: Operational Database, Data Science and Engineering
- Components: Apache Spark, Apache Kafka, Apache HBase
- Database: Greenplum
Applications supported
- Real-time communication personalization and targeted outreach
Data sources
- Log files from 150 billion transaction events per day
Impact
- New features and product innovations drive smarter decisions
Big data scale
- 500-node analytics cluster
With Cloudera, Conversant processes data from 150 billion daily transactions and makes that data immediately actionable for machine learning and analytics.
Conversant is the leader in personalized digital marketing, performing more than a trillion real-time decisions daily about what content to put in front of 160 million people, across all major ad exchanges.
Challenge
Conversant sees tens of billions of transactions every day, and must respond to them in milliseconds to remain competitive. With 30 percent organic annual growth, Conversant needed an infrastructure that could keep pace with the speed and scale of its business.
“We see the entire Internet because we're interfaced with all major exchanges,” said Patrick Jaromin, director of software engineering, ad tech at Conversant. “We needed a platform that could handle the volumes we're seeing on a regular basis. Things like relational databases weren't working. I had what I call a ‘lost year’ when I couldn’t accomplish goals because we spent so much time trying to work with a platform that wasn't performing. It wasn't reliable.”
Solution
Conversant partnered with Cloudera to create an environment to support better understanding of customers and products through real-time processing, analytics and machine learning.
One Cloudera cluster facilitates analytics and machine learning across the entire enterprise. In addition, Conversant maintains 100-plus node clusters that function as operational databases at each of its data centers for stream processing and storing information for 150 billion daily transactional events. These clusters maintain real-time online profiles and deliver query response in milliseconds.
Implementation
After struggling with the performance of legacy relational databases for its operational workloads, Conversant invested in a persistent big-data platform, but it was unstable and suffered frequent outages—impacting Coversant’s core business.
In 24 hours, Conversant migrated 60 billion consumer profiles into a Cloudera cluster.
Results
Without dedicating time and resources to keep the data management platform stable and performing, Conversant’s team can invest its energy into product innovations, introducing new features and making smarter, faster decisions.
“The faster we retrieve the data, the faster we process, the more processing we can do,” said Jaromin. “Because we're not suffering from frequent outages of the platform, I can focus on driving our platform forward and that's exactly what we’re doing.”