As IntegriChain looks to the future evolution of its product development to build a collaborative, data driven healthcare channel, new big data technologies are constantly analyzed and evaluated to create a modern data architecture. One of IntegriChain’s upcoming offerings revolves around a great deal of predictive analytics, including daily forecasting of sales – not just at the distribution center but in some cases down at a store level – to help healthcare manufacturers manage their downstream channel more effectively.

Predictive analytics tend to be computationally much more complex and expensive than IntegriChain’s traditional workloads. IntegriChain’s existing technology stack involving SQL and various programming languages tends to require significant development effort to:

  • Implement statistical, forecasting and machine learning methods properly
  • Manage and partition the jobs associated with running millions of forecasting scenarios across a cluster of multiple servers (which sometimes fail)

To date, predictive analytics have largely been either a manual process or a primitive one for healthcare manufacturers and, frankly, for IntegriChain. However we believe that predictive analytics are a core operational component of a mature data pipeline, worth doing with excellence at-scale.

In late 2013, the open source R programming language, which has one of the most robust libraries of statistical and forecasting operations, was extended to support and run its jobs natively on every node of a Hadoop© cluster (rather than on a separate node which would become a bottleneck eventually for a cloud provider like ourselves). We knew then the time was ripe to take advantage of this innovative combination in IntegriChain’s next-generation predictive analytic products.

Hortonworks Hadoop makes it possible for IntegriChain to support an increasingly varied set of data types, improving speed and flexibility while reducing cost. At the same time, Revolution Analytics’ Revolution R Enterprise drives informed decision-making powered by the most advanced big data analytics platform available today, enabling insight into trends, behaviors, predictions, and even outliers in enterprise data. With these technologies, IntegriChain now possesses almost unlimited scale.

By leveraging R and Hadoop, IntegriChain can direct the bulk of our efforts at helping our customers solve their problems and avoid cycles spent on managing the technology and reimplementing aspects of forecasting methodologies that that are either well-worn — or brand-new and cutting edge. With both technologies being open source and able to scale out to thousands of customers, we can leverage volume economics to meet our customers’ predictive analytic needs in a cost-effective way.

Look for details of IntegriChain’s new predictive analytics offering later in the year.

About the Author

Brandon Underwood

Brandon Underwood

Product Marketing Manager