A new "visual spreadsheet" helps users easily create and apply system models

Decision-making can be hard enough even when the choices seem simple: watch Leno or Letterman, cheer for the Yankees or the Mets, order Cherry Garcia or Chunky Monkey. But when the factors that affect the decision - and the impact of the decision itself - loom large, decision-making becomes a complex and critical task.

At that point, modeling the system in question can dramatically improve your ability to make effective decisions. If, for example, you're trying to eliminate bottlenecks in a manufacturing plant, creating a mathematical model of the production line helps pinpoint exactly where improvements would most help boost productivity and profits.

Unfortunately, for many reasons convincing decision-makers to take advantage of system models has proven challenging. First, no streamlined method exists for pooling the knowledge of decision-makers and appropriate experts to produce a comprehensive picture of the system. Programming the system model can be grueling and modifying it worse still. And when the user finally puts the program to work, the output may look as though it was spewed from a "black box" - giving no idea how the program performed the calculations or what the results really mean.

What's needed, then, is a rapid way to model systems, analyze changes, and present the results. At the same time, the modeling environment should be visual, intuitive, quick, easy, and customizable. That's a tall order, and no current off-the-shelf software meets these criteria. But GM researcher Jeff Alden decided to fill the void by developing an integrated systems modeling environment called InSight. Alden had one more requirement for his program as well - it had to be fun to use.

"In a lot of ways, InSight is like a visual spreadsheet," says Alden, Senior Staff Research Engineer in the Manufacturing Systems Research Laboratory. "Many people use spreadsheets like Excel to help them analyze problems. InSight goes further, by showing users how the various pieces of the puzzle fit together, what data they need to solve a problem, how the program will calculate the results, and how the results change by varying the inputs."

Click on the small images above for a closer look at how Excel and InSight differ in their handling of business case models.

With InSight, users begin developing a model by drawing an influence diagram, which shows how all the elements that comprise the system relate to each other. This calls for collaboration among hands-on personnel, subject-matter experts, and decision-makers to determine the model's inputs and outputs. The influence diagram for the Maintenance Operations Planning Tool, for instance, shows every element that could affect Cost and Revenue - and ultimately Profit - as a node or bubble in the model.

Click on the image above for a closer look at the influence diagram and equations entered for the Maintenance Operations Planning Model.

In the next step, experts enter all the math equations that correspond to those nodes. If the ultimate output of a model is profit, then the equation Profit = Revenue - Cost will appear as one of the calculations. Although that's certainly straightforward enough, nodes can use far more extensive computations, including optimization routines with thousands of lines of code.

During the third step, users choose the type of analysis they want. InSight contains a number of capabilities, including cases, cost drivers, optimization, simulation, and ranges.

Using InSight to solve problems

"The model helped determine how many people should be assigned to each level," says Alden. "We started with most people in reactive maintenance, then slowly moved them to predictive maintenance to see what happens. As we changed the number of personnel, the nodes that show the results began to grow or shrink. You can actually see the changes, as though you were running a movie."

And what did the results show? "At first, increasing the number of predictive personnel helped the plant meet demand, generate more products, and increase profit," Alden noted. "But eventually, the few reactive maintenance personnel left couldn't keep up with repairs. Then throughput dropped and profit took a big dive. But by looking at the influence diagram, you can see a sweet spot where the plant is using the optimal number of maintenance personnel in each category."

Not only is InSight effective for decision-making, but it's also a dynamic way to show how a system is evolving. Users have put InSight to work in a variety of different applications, from sourcing decisions to risk management to vehicle planning and evaluation. In recognition of the program's positive impact on GM's business and of the extraordinary effort Alden and his colleague Dan Reaume put into developing InSight, GM R&D awarded them the McCuen Award.

InSight obviously has application far beyond GM's business, and the rest of the world may soon get a chance to use it: GM is working with a software consulting firm to commercialize InSight, and Alden and Reaume are writing a book on decision analysis using influence diagrams.

"Within GM, InSight has proven to be a powerful tool that people can use to collaborate on building models, while having fun and making better decisions in the process," he says. "That's the reason I developed the program and where I still hope to see its major impact."

By Diane Kightlinger