Tim Rey, director of advanced analytics with Dow, offers a glimpse into that company's use of analytics and complex mathematics to examine multiple areas of its global supply chain.

The application of advanced analytics requires substantial resources drawn from multiple parts of the organization. "It's a balance of people, process, methods and technology," says Rey. Individuals must be highly trained in math, machine learning, forecasting, simulations and operations research, to name a few key areas.

At Dow, many of the people who participate in advanced analytics have a background in research and development, where they were already doing mathematics and modeling for manufacturing processes. Most possess advanced degrees, says Rey. Experts in Six Sigma and master black belts can be of great help in the effort.

While the concept of advanced analytics isn't new, it hasn't been applied to the business side of the organization until relatively recently. A handful of universities are beginning to certify students in that area. Dow is speeding the development of the discipline by bringing in graduate students and putting them to work on the analysis of business processes. The company has already done a wide range of work in areas such as fraud detection in auditing, strategy, portfolio optimization, forecasting and model, purchasing cost forecasting, and the use of purchasing decisions to minimize cost.

It's essential to get access to data as quickly as possible, Rey says. "Waiting six to nine months for a report to be available doesn't work." Commercially available software can help; companies don't need to build optimization algorithms from scratch. In any case, says Rey, one needn't wait for all corporate data sources to be structured before undertaking an analytical process.

There's always the risk of getting too complex in one's calculations. "All models are wrong," notes Rey. "Some models are useful." The trick lies in striking the right balance between theory and reality. And companies must always be aware of the "garbage in, garbage out" nature of data. "You have to be careful," says Rey. "Sometimes the quality of the data doesn't merit the complexity of the model."

To view this video interview in its entirety, click here.