Spotlight PlasmaNet : Collecting Everything

PlasmaNet FreeLotto

Performing Real-time Targeting

One of OmniTI’s long-term clients, PlasmaNet, was experiencing explosive growth as the largest online lottery site in the world. They needed to track behavior of 80 million users and send tens of millions of emails, with highly personalized offers based upon each user’s previous actions: purchases, offers acted upon and geo-location. Because they were expanding worldwide, the had to provide multilingual experiences, as well, for 19 languages that had to be automatically detected, along with different business rules and user experiences for different countries.

PlasmaNet wanted to collect “everything.” This meant capturing views, clicks (pages, emails, offers), retaining historical data for the life of the user, (including ecomm transactions, responses to offers, behavioral trends). They had a multi-terabyte of data and were continuously growing.


OmniTI engineers set up a “live” database, highly optimized for write-heavy transactions, with data replicated and aggregated to offline storage in almost real-time, targeted to specific targeting use-cases and allowing for extensibility for new use-cases. We created heterogenous database replication and split data sets, used for user targeting between “live” and offline storages to use for different use-cases. By separating the storage we allowed for independent optimization offunctional uses. We optimized data sets for particular use cases for real-time selects/analysis and cached reusable sets to further increase data mining operations.


The system OmniTI created meant that the client could do real-time behavioral targeting, geo-targeting, demographic targeting and historical targeting, while crunching the numbers for business analytics at the same time. The resulting system allows for analytics/BI, and reporting on data sets and further analysis of behavioral trends, based upon segments/targets. We also setup business monitoring for better decision making, and an ability to further refine use-cases to optimize data sets needed.