How Eastman Chemical Organizes To Gain Value from Big Data Analytics

It’s hard to miss the hype around analytics, says Mike Paulonis, Data Science Manager at Eastman Chemical Company, as he observes the world applying analytics and data science to create better sports teams, predict election outcomes, stop fraud, minimize customer churn, and seemingly revolutionize business. 

Nevertheless, he suggests, “if your company is one of the few that has great strategy and capability in this space, and is generating plenty of business value, this article is not for you. But if, instead, you are wondering how to take best advantage of the capabilities you have, how to accelerate capability-build in these emerging technologies, and how to manage and control retooling needed to convert know-how into insights and results, I suspect you’ll want to continue reading.”

Eastman Chemical Company has a culture of data-based decision-making rooted in a heritage of statistical analysis and business reporting. Our Applied Statistics group has been in existence for more than 40 years and we’ve been practicing a modern version of Business Intelligence for about 20 years. However, we have a culture of continuous improvement, and it became obvious to us that if we were to keep up with the rapidly changing field of analytics we would have to evolve beyond where we stood.

Adjustments we can make are primarily in structure and strategy, so I’ll describe what we’ve done in those areas at Eastman to get more value from analytics and big data.

Consolidate

Eastman’s Data Science organization (see illustration next page) has been in place for just over a year. We brought together groups that were generating insight from data in some way but were operating largely in their own silos.

We also created a Business Analytics group, but more about that later. The organizational merge will be followed by a physical move of most of the department into the same building on our large Kingsport, Tennessee campus after construction is completed in 2015.

The intent of bringing the groups together is to improve collaboration and synergy, whether that involves a quick consultation, a cross-functional project, or sharing project opportunities that would fit some other group better. We’re succeeding in this vision, having documented 22 significant collaborations in the first half of 2014. The capabilities each previously existing group brings to the organization are:

  • Business Intelligence: deepest knowledge of Eastman data of any group; owns the data warehouse and online analytical processing (OLAP) cubes; delivers primarily descriptive analytic applications such as interactive reports, dashboards and scorecards.
  • Applied Statistics: extensive knowledge of experimental design and regression techniques; creates predictive models primarily for technology and manufacturing clients from experimental or operating data; linked tightly into the technology side of innovation efforts.
  • Operations Research: experts in optimization, discrete-event simulation and Monte Carlo simulation; develops and deploys prescriptive model applications primarily for business and functional clients; applications may be for one-time use such as a capital investment or business strategy decision, or may be ongoing such as production scheduling or inventory management.
  • Library and Information Services: access to external data and information sources; steward of many sources of unstructured internal data; unstructured text analytics primarily of patents.

Create

For all the capability Eastman had in siloed groups before Data Science, there was still a gap in predictive analytics and higher-end descriptive analytics in business and functional areas. Applied Statistics occasionally engaged on a business project but had more than enough work just in technology and manufacturing. We sought to close this gap by creating a Business Analytics group.

Because we had no one with well-rounded capabilities in business analytics, we chose to launch the group with a diverse set of employees bringing parts of the eventual whole we would like. We started the group with five people whose capabilities included:

Person 1.   Statistics, R, scripting (group leader)

Person 2.   SQL, MDX, SQL Server Integration Services, business process     modeling

Person 3.   SQL, data warehousing, clickstream analytics

Person 4.   Statistics, six sigma, business process improvement, .NET       programming

Person 5.   Neural networks, expert systems, supply chain processes

Among these team members all the skills needed to implement a full analytic process such as CRISP-DM are present (http://en.wikipedia.org/wiki/Cross_Industry_Standard_Process_for_Data_Mining).

The components of technical skill are important, but possibly even more important are the right non-technical strengths. We tried to choose people who are voracious learners and exhibit incessant curiosity and persistence.  Everyone needs to pick up not only the skills present in the rest of the group but also those skills no one in the group started with. Examples of these include Python, ensemble regression methods, unstructured text analytics, and graph analytics.

Just one year in, we have made great progress in rounding out skill sets.  Most projects have included two or three Business Analytics team members so each can learn from the others in a practical setting.  One team member delivers a tutorial on a topic of his or her choice at monthly team meetings. The team dedicates time mid-day on Friday (and as much personal time as they choose) to participate in high-quality online courses in Business Analytics topics through services like Coursera, edX, and others.

Innovation Support

Eastman’s Data Science organization works across the entire company, engaging in projects or consultations with a committed partner and a valuable opportunity. Although Innovation Support is just a fraction of what we do, that support is the focus of this article.

The Table below lists examples of how each group within Data Science supports innovation. The work cited may be completed, in progress or upcoming.

 Data Science Support for Innovation

 

Group Innovation Support
Business Analytics Learn from CIMS how to utilize IBM Content Analyzer for unstructured text analytic queries on a corpus of Internet documents.  Support front-end innovation in their use of IBM Content Analyzer.
Business Analytics Perform unstructured text analytics on internal sales and marketing call reports to help uncover opportunities or gaps that can be addressed with innovation efforts.
Business Analytics Apply data mining and visualization to product sample requests to uncover unexpected application possibilities that can be pursued with innovation efforts.
Applied Statistics Design and analyze experiments to directly guide innovation projects to go/no-go decisions or the best directions to pursue.
Business Intelligence Develop competitive intelligence databases and reporting mechanisms to help guide selection of innovation opportunities.
Operations Research Perform Monte Carlo analysis of innovation project financial scenarios to help inform go/no-go decisions or focus work on variables with high financial leverage.
Library and Information Services Perform search and unstructured text analytics on external research literature, internal technical documents, and patents to guide innovation efforts.

Summing Up

There are, of course, many ways in which a company could bring data and analytic capabilities to bear on valuable opportunities. Whether it is for innovation or any other purpose, at Eastman we believe the collaboration, cross-training and one-stop shopping of a centralized Data Science organization is going to deliver the highest productivity and best outcomes.

At just beyond a year into our new organization, we’re seeing those benefits without any downsides that might be considered watch-outs for others who might wish to try this model.

Mike Paulonis,
Manager, Data Science; Eastman Chemical Company, Kingsport, Tennessee; paulonis@eastman.com

 

 

Comments are closed.