Statistical Revolution: Data Analytics Team Optimizes Decisions Under Uncertainty
You've heard about "big data" – but the real value in data is in its use. Enter the small but powerful Data Analytics team, which is creating a statistical revolution for Pratt & Whitney.
An argument can be made that every decision at Pratt & Whitney is made under some level of uncertainty – some more than others.
That's why Erica McCullough, manager of the company's Statistics and Data Analytics group, and her team are using statistics to make the best decisions possible given that uncertainty. "Statistics is essentially a set of tools for using data to model risk and variation and, ultimately, optimize decisions," she said.
The eight-member team works with business units across the company providing decision support on projects ranging from robust component design and root-cause analysis, to Aftermarket forecasting and QN reduction. In all cases, the group's goal is to capture and account for uncertainty and variation.
"We work with our customers to capture the range or distribution of potential outcomes rather than a single, deterministic point estimate that, as we've seen in multiple cases, can drastically miss the mark," said Grant Reinman, senior fellow in Statistics and Design for Variation.
McCullough added that utilizing statistics becomes even more critical when making decisions with minimal data and lots of uncertainty, versus when you have a lot of data. "That's when it becomes even more important to look at the risks involved in your decision-making," she said.
While the group is part of the Engineering organization, it is actively working to get the entire company "thinking statistically" – whether engineering, manufacturing or financial decisions are being made – through a new Engineering Technical University (ETU) course, a Pratt & Whitney wiki page and updated Engineering Standard Work.
Steve Morford, vice president of Core Systems Engineering, echoed the importance of the group's work. "Data-based decision making is fundamental to Core Systems Engineering and beyond," he said. "The framework that statistics provides for turning data into actionable knowledge is critical to our success as an organization and a company."
An example of where the Statistics group's efforts are being brought to bear is a substantial effort currently underway involving the new Next Generation Product Family (NGPF) business and the corresponding switch to fleet management programs (FMP).
"We're using advanced statistical analysis techniques to predict how often engines will need to come off-line and planning the number of shop visits necessary," McCullough said.
"By statistically analyzing engine time on-wing, we'll be able to quantify with some certainty our shop visit rate and how that will support our FMP business," Reinman added.
Dave Houston, manager of Core Structures in Engineering, has been one of the group's most vocal customer advocates, coining the term "Statistical Revolution" within his group as a call to arms for changing the culture to one where everyone is thinking in terms of uncertainty and variation instead of single-point, deterministic answers.
"When it comes to determining the structural capability of our parts, there is uncertainty in material properties, geometry, usage and more," Houston said. "Without capturing this variation, we could be overdesigning parts and losing out on potential weight and cost improvements."
Reinman, who's been a Pratt & Whitney fellow since the early 1990s, added, "We tend to consider variation only by looking at what we believe to be the worst case. The fact is actual outcomes vary from one engine to the next, sometimes in unexpected ways, especially in large production runs. The disciplined use of statistical data allows you to engineer and develop products after exploring the entire range of possible outcomes."
He noted that when an organization asks the question, "How do I allow for variability or uncertainty in my decision making?" it should involve the Statistics team. "That should be their number one question, regardless of the decision to be made," he said.