What Does It Mean?

what does it mean?

It's not enough to gather information. You have to make sense of it, too.

By Jean Thilmany, Associate Editor

Like all large companies, General Electric Co. runs the risk of duplicating work again and again. In such a huge, disparate company, it's hard for the left hand to know what the right hand is doing.

GE, headquartered in Fairfield, Conn., makes everything from washing machines to aircraft engines to locomotives to medical imaging equipment to generators and gas turbines at many locations. Employees face the same kinds of problems regardless of what they're designing. Solving the same problem again and again, without realizing that someone at another location might already know the answer, wastes time. Business information that could be shared throughout the company is instead frittered away, according to the makers of a type of technology meant to store and reuse that knowledge.

Employees in each GE division continually brainstorm, design, manufacture, and market new products, generating hoards of information each step of the way that may simply sit in one thinker's individual computer.

Makers of what is called knowledge management software argue that companies should record everything that happens in each division—from the generation of ideas for a new product right through to tracking its sales—and have all the information available for other employees, whether they know exactly what they're looking for or not. These companies say their software helps engineers and others avoid reinventing a wheel and also helps them discover how someone in another division worked through the problem.

At plastic conveyor-belt maker Intralox, software records injection-molding machine information and uses it for a number of purposes.

Knowledge, the broad term used for information gathered in a software system, is only as powerful as the way it's used. A large book of the notes taken during a new product brainstorming session or a huge printout of plant floor machine operating parameters measured each day won't do anyone any good. What counts is how that information is retained, made user-friendly, and turned into easy-to-understand reports or reference materials. And it's those things that knowledge management software aims to do.

GE uses knowledge management software from Sopheon Corp. to track and retain all kinds of information and ideas that are generated by employees during product development.
According to Andy Michuda, chief executive officer of Sopheon in Minneapolis, the software, called Accolade, includes a feature that lets users get advice via phone or e-mail from one of 50 on-call experts in science and technology. For example, if an engineer encounters a particular problem when designing a product, he can call on an expert. The back-and-forth e-mails are then stored in the knowledge management software's database and can be accessed by anyone at GE who encounters the same problem.

In fact, a dialog box might pop up on the engineer's computer while he grapples with the problem. The box might say that a GE engineer has already solved this problem and may provide a link to the warehoused e-mail correspondence. That engineer could also phone the original engineer to ask more questions, if necessary.

"If I sit in a corporation of 40,000 people, I can blast an e-mail to everyone when I'm faced with a problem," Michuda said. "But that just increases the noise factor. I need to find the people who can help me."

Notes From a Brainstorm

In the first stage of product development, the brainstorming session, members of a new product development committee bring up their ideas. At this stage, they can answer questions included on an Accolade customized template, which gives them an idea of whether or not an idea is valid, Michuda said.

The template includes questions that the company's executives ask when they're pitched a new product: What technology would be needed to make it? Which products will compete with it? What companies compete?

The software comes loaded with answers to questions that may come up during a brainstorming session, Michuda said. For example, information on how similar products have been received might be included in the software. Committee members also can call the Sopheon consultants, who will answer questions or help at brainstorming sessions.

Executives refer to the templated information, while committee members pitch the new idea. If every question in the template is answered, a new product has a good chance of approval, Michuda said.

The questions also ensure that committee members agree on important facts before they pitch the product concept to higher-ups.

"If I've done my homework, I have a much higher probability of getting my product approved," Michuda said. "Today, in those meetings, people answer these questions and then they leave. They don't capture the answers."

Another Sopheon customer, the mobile telecommunications provider Vodafone Group Plc of Berkshire, England, uses the templates before each product meeting. Because many of the questions are automatically answered using information already in the system, filling out the template takes about 45 seconds.

"They're extracting the data from other sources and then there are gaps. They need to complete that part themselves," Michuda said. "Once the template is complete, it's stored in the system because someone will come along with a similar product and the information can be used to populate the next document, where relevant. Some companies have thousands of product teams going on at once, and they don't talk to each other."

When a proposed product is given the green light and passed off to the mechanical engineer, that person can refer to the templates to see what product managers originally envisioned when they came up with the idea. It stands in for picking a manager's brain about the idea, Michuda said.

Meaningful Measurements

Software that gathers and massages information so it's easily interpreted by users—who might need it for a number of reasons—can take other forms.

In 1998, Intralox, a maker of modular plastic conveyor belts, installed new product line monitoring software after executives grew dissatisfied with older software that monitored the production line. The reason is that it did only that, said Ned Dudoussat, plastics scheduler at Intralox. The company, in Harahan, La., just outside New Orleans, is a subsidiary of Laitram Corp., which was started by James Martial Lapeyre, inventor of the automated shrimp peeler.

Shrimp attracts bacteria, so plastic is the natural choice for a shrimp conveyor because it doesn't promote bacterial growth, Dudoussat said. Intralox was spun off from Laitram specifically to make the plastic conveyor belts used on the shrimp processing line.

Now, the conveyors are used predominantly in the food processing industry. Dudoussat said chances are good that if you've had a soda or a beer recently, it once rolled down an Intralox belt. The plastics injection-molding machines that make those belts must be continually monitored and adjusted.

"Roll back the clock to 1989, when we installed a production monitoring system for our injection-molding machines to monitor cycle times and the number of cavities a mold has, which shows the number of parts that will fall out of a machine," Dudoussat said. "This is fundamental process control and data."

Intralox uses the processing monitoring software Shotscope, from Moldflow of Wayland, Mass.

But a system that was cutting edge in the 1980s didn't fit the manufacturer's needs by the mid-1990s. For one thing, programs written in those days were original to themselves, as Dudoussat put it. The software was closed architecture, so data gathered couldn't be extracted and used in other software applications. For example, you might want to look at the history of a particular injection-molding machine's output. You couldn't write or buy software to allow that with the old monitoring system, Dudoussat said. Plus, the former system just didn't give back reliable information.

"We'd go into a meeting and say, 'Here are the numbers,' and people could shoot holes in it," said Trey Diaz, an operations analyst at Intralox.

The company uses 300 plastic molds, has 70 injection-molding machines on its line, and makes 2,000 plastic parts. In 1998, it installed a new process-monitoring system called Shotscope, from Moldflow of Wayland, Mass.

The software tracks process data and records exactly the state of the injection-molding machine as it makes each shot, Dudoussat said. All pressure, temperature, and machine settings are recorded and monitored for trends over time. It also tracks production data, which shows how many parts were made, and how quickly, and gives the rate of scrap.

No Margin for Error

Plastics suppliers traditionally live and die by scrap because they typically have no margin for error, Dudoussat said. That means plastic molds must be efficiently designed and injection-molding machines on the line must produce parts efficiently.

What's done with the information gathered is where knowledge management comes in. The software records every machine parameter for every shot made. Some machines record about 15 to 20 parameters, including temperature and scrap, per shot. The amount of information can be staggering and data is simply raw until it's assembled in a meaningful way. How the data is massaged to derive meaning changes according to who's looking at the numbers: a process engineer, a line maintenance person, or a production scheduler.

Shortly after the system was installed, Intralox analysts wrote a software application for an intranet paging device tied to the system. If a machine starts functioning out of its prescribed parameters, the software automatically sends a page to the beeper of a maintenance person, saying the machine is out of alignment and describing the problem.

That was the first homegrown application. Others followed, and so have uses for the original software.

"Once we got the system in and had some faith in our ability to keep track of rates, several projects grew out of that," Dudoussat said. "Now I'm interested in the scheduling of the plant. We're able to make some software tools that can access the data in the Shotscope database and determine where molds should run on a daily basis. You're looking at 70 machines. Say someone says you have to add 20 parts that day. When do you want to start them? Where do you want to start them? How do you want to run to meet your inventory needs?"

Information that is collected right on the plant floor can be compiled into meaningful Web reports.

Diaz helped write an application that extracts data from the monitoring system and places it on the company intranet so select employees have access to production floor data. An easy-to-understand Web-generated report calculates efficiencies for the plant, for example, which can be seen on the intranet and easily understood.

"Most monitoring software has a very out-of-the-can reporting method for determining your production efficiency," Diaz said. "But our analysts figure that method didn't tell you whether a machine was having a problem with the rate of production because of the cycle time you were running or because of the number of cavities available at a time, or because of scrap. The numbers didn't tell you what area you were having a problem with."

That homegrown Web-generated reporting application opened the company's eyes to other uses for the monitoring system, Dudoussat added. Now system information that previously would have been difficult to obtain or share is available to employees at all levels—from managers to those on the shop floor.

"Say a production supervisor wants to see the performance of an individual machine if he has a gut feeling its efficiency is suffering," Dudoussat said. "In the old days, he'd have to approach a system administrator and ask for a specific range of the machine's data for a specific part for a range of dates, and the administrator would have to manually extract the data and put it into Excel and calculate to figure efficiency, scrap rate, and production rate. Now the supervisor does that process on the floor. He puts in a part number and a date, and presto, it takes about two minutes."

The shop floor scheduling component has changed the way production scheduling is done at Intralox. Many of the company's molds can be run on one of several injection-molding machines. Deciding how to schedule production to derive the full use of those machines can be tricky, Dudoussat said. Schedulers feed the critical dimensions of a mold and a machine into the monitoring system's database, which returns a list of possible machines on which the mold can run.

"This plant runs 24 hours a day seven days a week and if a crisis happens in the middle of the night or on a weekend, the production supervisor can take his own initiative to move the mold to where it's never been before," Dudoussat said.
And it's on-call information taken from the process-monitoring system and compiled into a series of reports or notifications that keeps the factory running smoothly.