A recent online poll from Deloitte LLP lays out what it calls "the essential paradox." Today's vehicles are more reliable than ever, Deloitte says, and manufacturers have the capability to uncover safety problems much sooner than before. "Yet only a few are taking advantage of early-warning systems that can help detect trouble earlier." The result: "a manual, rearview-mirror approach to vehicle quality and safety."
Just a glance at some recent recalls makes clear what a bad idea that is. Problems range from minor glitches to life-threatening defects. Each month seems to bring a new announcement by some major automaker about everything from faulty airbags to windshield wiper motors. So how can industry do a better job of catching defects before they end up in cars that are sold to the public?
The solution to any problem begins with identifying it. Deloitte was motivated to undertake the poll by the record number of recalls that occurred in 2014 – twice that of any prior year, according to Derek Snaidauf, senior manager in the firm’s Analytic Insights Practice. In a larger sense, though, that period was no outlier. “We view it as an ongoing and persistent trend,” he says.
Industry knows it has a problem. More than 90 percent of respondents said recalls were affecting the working relationships between suppliers and original equipment manufacturers. (A state of affairs that takes a back seat, of course, to the many deaths and injuries that have resulted from this grievous lack of oversight.)
Nor are automakers completely blind to the possibility of future incidents. Just over 42 percent were expecting even more recalls in 2015 and 2016, Snaidauf notes. (What were the other 58 percent doing? Sitting back and assuming that the problem would go away?)
But even that limited level of awareness doesn’t mean that industry is taking adequate steps to head off defects during design and assembly. Only 8 percent of the Deloitte survey were using advanced predictive analytics to prevent or manage recalls. And 23 percent had no analytic capabilities related to anticipating product safety and recall issues.
Why advanced analytics? It’s not a question of companies lacking numbers. Many rely on business intelligence and reporting to smoke out quality problems, says Snaidauf. For example, a report on a particular vehicle model might show the number of repairs or warranty claims per unit.
What most automakers aren’t doing, he says, is tapping the potential of unstructured data – insights gleaned from the voices of customers, dealers and technicians.
In the past, it was nearly impossible to sift through millions of such narratives from the marketplace. But the combination of “big data,” advanced analytics and natural language processing allows manufacturers today to make sense of the “noise.”
Another effective use of modern analytics involves delving more deeply into warranty claims, affixing “a more interpretive, prescriptive lens” to the problem, says Snaidauf. The automaker might ask itself: Is this particular number of claims statistically improbable? Is it something I should be worried about? The scrutinizing of historical patterns over time can expose the true anomalies in fault and repair records, he says.
Automakers could also do a much better job of partnering with parts suppliers for early detection of defects. Snaidauf says the advanced analytics approach is “very exciting to them as well.” Suppliers, too, would like to have predictive capabilities, but they often don’t have access to critical data from OEMs. In the Deloitte survey, nearly 22 percent of respondents named OEM-supplier collaboration, hampered by ineffective communications, as one of their biggest challenges.
So why has industry been lagging in its reliance on analytics to prevent recalls? In part, says Snaidauf, because software tools are only now catching up with the huge volumes of data that are available to manufacturers. That said, the larger issue is a lack of talent. Many companies don’t have enough of the data scientists and analytics professionals say are necessary to building out that capability.
In general, Snaidauf says, the most effective organizational models are those that carve out formal responsibility for analytics on a company-wide basis, ideally reporting to a chief financial officer, chief information officer or even chief executive officer. A typical company might possess some analytics capability, but it will often exist in pockets of the organization and not be managed in a coherent fashion. One proven approach is to house analytics experts within a center of excellence, then loan them out various departments to perform specific tasks.
Snaidauf does see progress among automakers toward the acceptance of predictive analytics. Companies are beginning to employ such methods as correlating internal and external data sources, and relying on interactive visualizations.
Two or three years from now, the number of automotive companies adopting safety and recall analytics programs will likely be higher, Snaidauf says. The benefits are just too compelling to ignore. Such initiatives can help to reduce the total cost of quality by 10 to 15 percent, he said. More importantly, they can save lives.