What do you do when you’re tasked with collecting and analyzing data for 320 billion retail products and the 400 million households they’re being sold to, in real-time? Without a super-powered database machine and advanced analytics, there might be some panic involved. dunnhumby tracks and analyzes retail loyalty card data from households around the world. Insights are gleaned from that data to help retail and consumer packaged goods companies understand their customers on a granular level. That understanding then feeds into strategic, personalized marketing and communications for their tens of millions of customers, as well as new product development and other innovations.
dunnhumby’s business model revolves around individuals, helping companies fully understand their most loyal customers and how to provide value by knowing what matters most on a personalized level. We add customer filters to our analytics to uncover valuable insights: what customers want, where they want it, and how much they want to pay for it. While this makes the technological requirements much more complex, dunnhumby doesn’t look at an aggregate level of data and extrapolate. Granular data helps to provide insight based on actual individual interactions and behaviors.
Being able to work with that data - all one billion rows, across 33 solutions, processed weekly - in real-time is crucial. So dunnhumby turned to Oracle Exadata Database Machine and Oracle Advanced Analytics with the hope that our ability to do advanced customer analytics on an increasingly vast data set would benefit from a platform that would grow exponentially, supporting expanded performance, scale, and complexity requirements.
Exadata Enables Insight from All Data, Not Just a Sampling
Previously limited to analysis based on sampling, dunnhumby searched for a solution that would allow for a wider lens through which to view 400 million households. Oracle Exadata provides a huge supply of processor cores, in our case, making available as much as 2,560 degrees of parallelism (DOP), effectively granting us both the ability to execute dozens of simultaneous processes and the power to choose and allocate resources to the most CPU-intensive computations. As a result, we can now provide insight based on the totality of our data rather than sampling, offering a truly comprehensive and personalized view of today’s customer behavior. Our most processor-intensive, count-distinct analytics have become rather simple since the transition to Oracle Exadata. When a national food retailer, for example, delivers 10 million quarterly rewards mailers, each one is completely unique to the individual customer based on purchase behavior. Delivering these highly personalized mailers in a timely manner is only possible executing with efficiency on a bigger, better platform.
We are also able to widen our scope and look at a much broader range of categories. Oracle Exadata’s 2TB of RAM per database node enable dunnhumby to do more of these grouping, sorting, and counting functions at much faster speed. Where we would previously have been limited to tracking data in just a few key categories chosen in advance, patterns can now be identified and followed in any number of categories allowing the data to call attention to itself. This provides broader insights to more fully answer clients’ questions, as well as to answer questions that they didn’t even know to ask, opening up greater possibilities.
Exadata's Capacity Enables Longer Periods of Historical Data
Since our philosophy revolves around customer loyalty, providing clients with a long-term, historical perspective is crucial. With Oracle Exadata’s capacity for 105TB of triple-mirrored data, we now understand and can analyze much longer periods of historical data than before. A year’s worth of data might generate some helpful analysis when evaluating a customer rewards program, but five years of data allows a client to completely understand the effects of their CRM strategy and see how it has changed and engaged their most loyal customers.