With more than $300 million in annual revenues and 1,600 employees, Merkle is a leading customer relationship marketing agency for such clients as Dell, Geico, DirecTV, and Chase. The company uses ParAccel's big data analytic platform to gain a 360-degree view of consumers so that Merkle's clients can work toward real-time marketing and measure campaign effectiveness more precisely.
Merkle has a long history of curating consumer information to provide "data as a service" for marketing purposes. Historically, this was achieved through monthly batch processing of huge flat files.
In order to evolve from a marketing database company to a customer relationship marketing firm, Merkle needed to reconcile and integrate big data sources. Digital consumer information such as IP addresses, cookies, and emails had to be merged with traditional offline information such as name, address, and phone number. Ultimately, clients needed deeper marketing engagement, such as emails and banners customized for individual consumers.
Reaching for real time
To meet those goals, Merkle needed processing to move from monthly batch jobs to near-real time, with the ability to integrate all interactions and achieve a 360-degree view of each consumer's behavior. To achieve these goals, Merkle builds data warehouses for big data analytics, some of which are deployed at its clients' sites and some of which are hosted by Merkle.
Merkle's challenges in selecting appropriate technology included the cost of a big data analytics environment, predictable high performance and scalability, and specialized requirements unmet by current analytics offerings.
Ultimately, Merkle chose the ParAccel Analytic Platform first for its efficient, scalable architecture as a massively parallel processing (MPP) columnar analytic database. "Merkle originally selected ParAccel because of winning execution speed and price performance," comments Peter Rogers, VP of Technology at Merkle.
For real-time analytics of structured big data, MPP columnar analytic databases have become a common choice. Columnar storage means that the relational database houses data in columns rather than in rows, yielding faster queries than in transactional systems. In addition, the data is compressed to further increase querying efficiency. Finally, due to the nature of MPP, you can "scale out" in linear fashion by simply adding more commodity hardware.