Mapping Social Networks to Improve Product Quality
Professor finds lack of coordination causes defects, but network methods can help predict and prevent them.
ANN ARBOR, Mich. — When the phrase "social network" comes to mind, people generally think about Facebook or Twitter. Volumes of academic studies have been written on this relatively new phenomenon.
But engineers who design such complex products as automobiles and airplanes have been operating within their own social networks — or specific patterns of communication — for a long time. As with most patterns of communication, gaps are bound to occur. A recent study undertaken by operations expert Wallace Hopp and two colleagues describes such gaps in social networks as "coordination deficits." And such deficits can be costly if not corrected.
"Based on what we hear from managers in the industry, about 60 percent of their quality problems are manufacturing-based and 40 percent are design-based," says Hopp, the Herrick Professor of Manufacturing at Ross. "We found that roughly 20 percent of those design problems can be traced to inadequate communication. This means that coordination problems are responsible for as many errors as individual mistakes by engineers. That's very powerful information."
Hopp recently co-authored "The Impact of Misalignment of Organization Structure and Product Architecture on Quality in Complex Product Development" with Bilal Gokpinar of University College, London, and Seyed Iravani of Northwestern University.
The paper reveals results of the researchers' efforts to identify, measure, and quantify communication breakdowns inside a specific design network at a major automaker. Then it details the innovative way in which they constructed a social network using data culled from reams of engineering change orders. The approach produced a statistically sound model that can help managers better coordinate projects across global engineering centers. A wide variety of firms can apply the networking model since the goal is to predict and avoid quality problems in complex products.
The practical implications are clear. In 2006, a survey by Bloomberg BusinessWeek and the Boston Consulting Group showed that senior managers cite a lack of coordination as the second-biggest barrier to innovation.
A Blessing in Disguise
Hopp and his co-authors aren't the first researchers to use social networks to map a new product development organization. But virtually all previous studies have made use of surveys to collect data with which to build a network. Indeed, Hopp initially thought his team also would rely on a survey of engineers about who they talked to on a regular basis during the design process. But that approach turned out to be problematic. Hopp needed to study vehicles that had been on the market for at least a year in order measure the available warranty claims. This meant the design process had started as long as five years ago. Since vehicle design programs involve thousands of engineers, this also meant that many of the individuals had shifted jobs or left the company. Even if the right engineers could be surveyed, five years is a long time for someone to remember who they talked to about what.
Hopp suggested studying email and phone records to map communication, but the company nixed that idea as too intrusive. But Hopp and his co-researchers did have the engineering change orders – records that document virtually all of the steps involved in designing the parts that make up the vehicle. Eventually they were able to extract information from these records that provided a map of all the systems in the vehicle, the engineers assigned to them, and who was talking to whom and how often.
Hopp and his co-authors used the data to rank the vehicle systems from simple to complex and to measure how much attention each system and subsystem received, given the communication between the engineers working on them.
"This is the first time that I know of that anyone has used an archival engineering database to construct a social network," Hopp says.
From that, they were able to see which systems suffered from a lack of coordination. Any gap between the complexity of a system and the communication surrounding it was a "coordination deficit." That deficit had a positive correlation with later defects found in the warranty data.
Hopp found that systems of medium complexity experienced the highest coordination deficit and the highest likelihood of defects. That's because complex systems receive plenty of attention because everybody knows they're going to be difficult. Simple systems don't require as much coordination. The intermediate ones tend to be overlooked.
"If you are really concerned about seeing the mismatches between your organization and your product, what you need to do is quantify the mismatch between those two networks," Hopp says. "We came up with a mathematical way of doing it which should be easy for companies to do."
The company Hopp worked with is now using an online organization chart to help improve communication and coordination among engineers. The idea is to give all the engineers a big-picture view of the entire project and make it easier to find the right people to talk to during the design process. Hopp hopes to do follow-up research to evaluate the impact of this new tool on the alignment of the organization with the product.
Hopp and his team also have extended their research to help guide managers as companies move toward designing products in several global engineering centers at once. They found that when a complex subsystem is developed in several locations, the number of locations predicts the likelihood of delays in design tasks. This impact is particularly pronounced in complex subsystems (i.e., ones that are highly connected in the product architecture network).
Still, it's a tough balancing act because companies want to take advantage of global expertise. The best team for electrical systems might be in Europe, even if most of the design work is in North America.
"What our research suggests is that you don't have to have everything in one place, but if you can reduce the number of design locations on highly central systems, you'll do better," he says. "So if you can move a few key people and get your locations down to three if you're using four, or down to two if you're using three, you can cut down on delays. I think this type of analysis will become increasingly important as more and more firms globalize their design organizations."
For Hopp, the ability to use engineering change orders for network analysis is good news, because the field of network study is growing.
"I think the network perspective is here to stay and it integrates very nicely into the engineering change-order system," he says. "For complex products like vehicles and airplanes, it turns out we've been collecting data all along that's relevant to this network analysis. Nobody realized it. We didn't realize it either when we started this research, but it's right there and it's a gold mine."
Figure 1: Superposition of product and organization networks. When connectivity between subsystems in the product architecture network exceeds connectivity in the organizational coordination network, the resulting "coordination deficit" can lead to design flaws that cause quality problems and, hence, warranty claims.
For more information, contact:
Bernie DeGroat, (734) 936-1015 or 647-1847, firstname.lastname@example.org