When Domino’s Pizza wanted to move away from its one size fits all couponing strategy, the chain analyzed data from more than 10,000 locations and tied sales to redemptions at the individual shopper level. The result? The pizza chain began issuing personalized coupons for its customers’ favorite pizzas.
While being widely discussed but not widely understood by many marketers, big data is already hard at work in the world of marketing and customer experience, noted Martin Kihn, research director, Gartner for Marketing Leaders.
Big data basics
Big data is not a single technology or a shortlist of vendors. It’s a loose collection of evolving tools, techniques and talent. Gartner defines big data as “high volume, velocity and/or variety of information assets that demand new, innovative forms of processing for enhanced decision making, business insights or process optimization.” In practice, big data can be divided into three categories: storage, processing and analytics.
Enterprise data is traditionally stored in relational databases which are structured in tables that can join with other tables in a carefully defined way. Big data strains this approach because there is too much data to fit easily into big enterprise databases and many uses require faster processing and analysis. Big data storage differs significantly from relational databases because it stores data that has not been mapped to a particular format or structure. By not being contained to such structure, the data is available much more rapidly for use.
Processing big data means collecting and moving it into storage or other systems in an organized way. Big data needs to be distributed across a number of different hardware locations and is generally not in a predefined format so it requires its own approach to processing.
Batch processing is working with data that sits in a constellation of database clusters which are spread across hundreds or thousands of different pieces of hardware. There are a number of frameworks to execute batch processing of data including MapReduce and Spark. Real-time processing works on data that is “in motion,” potentially at or near the point of data capture. Think of a marketer being able to process behavior data from a website visitor in the moment to serve that same visitor relevant ads, promotions or content throughout his or her site visit.