Marketers who don't understand the importance of advertising data and the mounds that online ads and social signals collect might as well hang up their hats and look for other work.
Attempting to educate advertisers and marketers, Turn, a software and services company that provides platforms for managing data in digital ad campaigns, published this week the white paper, "Cheetah: A High Performance, Custom Data Warehouse on Top of MapReduce" at the 36th annual International Conference of Very Large Data Bases (VLDB).
Cheetah, built from scratch in Java running on MapReduce, Google's programming model, can efficiently process, sometimes in seconds, any portion of the Petabyte of data it stores. Advanced analytics like machine learning and statistical modeling lay on top of the framework to automate the process of mining data from campaigns. The aim is to reduce the cost for clients, according to Dominic Bennett, vice president of engineering and products at Turn.
Bennett says Turn engineers wrote the processes and built the technology that allows advertisers to sort through the mound of data companies can expect to collect from ad campaigns.
Turn designed Cheetah to augment its advertising applications, as well as simplify and customize ad campaign optimization.
The company's engineers also developed methods to find variables that correlate with behavior. "When you provide the correct tool sets, even when systems are very complex, people typically not familiar with technology become evangelists," Bennett says. "We all know behavioral targeting and data works, but how do you show it to make people understand?"
The white paper details the background of the work that went into creating the process and technology, and why it's important to have a platform in place that can process the mounds of data tech companies expect. It also provides an overview of Turn's data warehouse system, Cheetah's schema design and query language, query processing and optimization, and the steps required to integrate Cheetah into user program.