Having longer periods of data to access also allows us to identify and create models of new product trends, an endeavor previously unproductive due to the insurmountable size and scope of information. We can now look at previous trends in customer behavior and use it to build models to predict the size and nature of the next trend. The final piece is to tie our findings together with years of granular behavioral data and provide insight as to why and how a particular trend makes sense for a specific brand.
With 22TB of flash cache, the system intuitively stores all of our data for the most efficient access. The internal Infiniband backbone transfers this data from the storage cells to the database nodes at 40GB per second, at least four times faster than standard Ethernet. With Oracle Exadata smart-scans, only the columns and rows required to satisfy user queries are sent from the storage to the database nodes – meaning less data is going through bigger pipes for ultimate performance. This means that the analytics and insights that we provide today are truly relevant today.
Oracle Exadata’s specs have driven our capabilities to the point where we are now building a new product database of performance metrics that will bring together past new product performance with new, time-aligned data. This will help to provide reliable benchmarks and guidelines for a new product launch, based on common product attributes. The database will also act as an early assessment tool, providing the ability to diagnose by metric why a product may not be meeting those benchmarks, and then to determine whether to steer the launch back on track or to abandon it. Conversely, this will also help to spot when a new product is a growing success, and provide the tools to push it further.
The Ease and Efficiency of an Engineered System Managed in One Place by One Team
Operating within a more stable, reliable platform is beneficial in order to smooth out internal wrinkles and focus on science. Standardization of menial tasks, such as summarizing real-time data, and potentially development of a subscription model that will automatically generate client reports are examples. Ensured data security and compliance within requirements for safeguarding personal protected information (PPI) will be conducive to our more sensitive projects without the burden of security concerns. Oracle Exadata provides both encryption on data at rest and seamless decryption for convenient internal access.
Perhaps most importantly, because Oracle Exadata is an engineered system whose entire infrastructure is managed in one place by one team, a key business benefit is concentration on utilizing technology, not managing it. The growing volume and unstructured nature of data requires a high level of focus on increasingly complex customer science and less on environmental concerns like managing disk space or repairing disk drives. Our data is triple-mirrored, meaning the entire technology stack can be patched and upgraded online with no downtime or significant effect on performance. Operations continue even at the scale of multiple concurrent queries, increasing productivity and allowing analysts to work uninterrupted by technological constraints. Having a reliable, proven technology platform managed by a single entity is significant for our business.
Although the results of the transition to Oracle Exadata boil down to simple words like bigger and faster, we see the success as anything but simple. Beyond changing today’s functionality, it has opened up new possibilities for the future, germinating ideas and excitement with retailers and consumer packaged goods companies, who are now working on new ways to capitalize on the potential offered by our increased capabilities, performance, and proficiency.
About the author:
As global chief information officer, dunnhumby, Yael Cosset provides leadership focused on building innovative, world class products and solutions including an expanded portfolio in the digital personalization area.