Random Acts of Data Science

For most businesses, analytic discovery used to be the domain of rogue statisticians who would scrape together data,  software,  and a few servers hidden under their desks.   These secret scientists would produce all manner of insights, graphs and predictions that would find their way into board rooms by way of PowerPoint and a delighted VP sponsor.

Unfortunately,  this is where the story often ended.  Although analytics would uncover fascinating opportunities and hidden dangers, businesses often failed to turn these discoveries into sustainable actions that would lead to the promised outcomes.

While many businesses are still trying to figure out how to turn analytic discoveries into profit,  the world of big data and analytics is growing at exponential speed.  This growth is creating increased competitive pressure and heightened urgency to figure out how to harness even more data and increase the number of actionable analytic discoveries.

To solve this problem,  today's leading businesses have figured out that they need to move analytic discovery from their hidden labs into mainstream business processes.  This requires transformation in three primary areas; discovery platforms,  discovery processes, and discovery lifecycle management.

Discovery Platforms

First, donate all those "shadow IT" servers to your local elementary school and have IT provision a scalable, high performance,  and secure analytic discovery platform.  This will dramatically increase the amount of data which can be mined and the number of analytic discovery business problems which can be tackled.  But,  perhaps most importantly,  moving analytic discovery onto a legitimate IT hosted platform allows for the security,  compliance and support of potentially mission critical insights.

Discovery Processes

Second,  platforms alone don't solve problems.  Processes need to be modernized to take advantage of all of this technology at scale.   Traditional IT processes,  policies,  and procedures are designed for the development of rigid and highly predictable (supportable) business applications.   The loosely defined exploratory nature of analytic discovery requires (1) a much more flexible approach to data ingestion,  transformation,  and decommission (2) ability to manage dramatic peaks and valleys of analytic workload,  and (3) greater support for self service tools which enable data scientists with capabilities traditionally held tightly by IT.

Discovery Lifecycle Management

Finally,  analytic discovery needs to transform from random acts of data science to a more well governed and managed business process.   This includes (1) improvements in how potential projects are evaluated and selected, (2) prior planning and commitment to testing/validation,  and (3) transition planning between business and IT on how successful discoveries will make it into production.

Today,  more data scientists are trading their lab coats for business causal attire as analytic discovery is moving from back office science projects to a more highly integrated and valuable part of today's data driven enterprise.

Ken is the Director of Analytics for Hewlett Packard Enterprise Services. He has over 25 years of experience delivering Business Intelligence and Analytic Solutions. Follow Ken on LinkedIn  Twitter