It is tempting to label everything involving data as big data these days as if the qualifier makes the topic inherently sexier. The Wall Street Journal is guilty of this in recent headline on “manufacturing execution systems” (How Many Turns in a Screw? Big Data Knows, May 15). While the headline may be hyperbole, the basic idea of these systems is pretty cool.
Raytheon is one of many manufacturers installing more sophisticated, automated systems to gather and analyze factory-floor data. The company uses software known as manufacturing execution systems, or MES, which has been around since the 1980s. Semiconductor and other high-tech companies were early adopters, but now “others are catching up,” says Tom Comstock, an executive vice president at Apriso Corp., one of the suppliers of this software. …
Manufacturers are looking harder at data partly because of increasing pressure from customers to eliminate defects and from shareholders to squeeze out more costs. Regulators are also demanding more data collection to trace safety problems. The cost of computers, scanners and other hardware has also come down, and technology for storing and moving data has improved.
At the same time, factory equipment has “got smarter,” says Mike Lackey, a vice president at SAP. The newest equipment comes with computerized controls that make it easier to collect data and share it with the rest of the company or suppliers.
It makes some sense that semiconductor manufacturers were out in front on these systems. Making chips requires lots of expensive, sophisticated equipment with relatively little direct labor (in comparison to, say, assembling cars). Hence, the machines truly control the process so one could think of doing sophisticated experiments without worrying about humans introducing unwarranted variability. Now it seems that similar approaches are being developed in other forms of manufacturing that don’t necessarily happen in a clean room and involve working at a micron level. But how can this kind of analysis be useful?
One obvious answer is tighter quality control.
At Harley-Davidson Inc.’s newly renovated motorcycle plant in York, Pa., software keeps a constant record of the tiniest details of production, such as the speed of fans in the painting booth. When the software detects that fan speed, temperature, humidity or some other variable is drifting away from the prescribed setting, it automatically adjusts the machinery.
“It allows us to be more consistent,” says John Dansby II, vice president for global manufacturing at the motorcycle maker. In the past, he says, operators had a bit of leeway on paint jobs and each could do the work in a slightly different way. Now it is supposed to be an exact science, not an art.
Note that this implies some real discipline in how the process is designed and executed. They are limiting discretion in how workers are doing their jobs in order to assure consistency. Of course, process standardization is often step one in many lean or quality programs. It may make the job more rote but it does increase productivity.
A second use of these system is to identify bottlenecks.
Harley has also used the software to find bottlenecks that could keep it from its goal of completing a motorcycle every 86 seconds. Recently, by studying the data, Harley managers determined that installation of the rear fender was taking too long. They changed a factory configuration so those fenders would flow directly to the assembly line rather than having to be put on carts and moved across an aisle.
This is an interesting twist. I suspect that front-line managers would have a pretty good idea of where issues in product flow arise. Automating the data collection, however, can put concrete numbers on what kind of changes are needed or what kind of payoff can be expected from saving a handful of seconds on a task.