Story Title: Nevada Industry Excellence Helps Starbucks Improve Process

Client Profile: Starbucks Coffee Company's Carson Valley Roasting Plant & Distribution Center roasts, packages, and distributes coffee and non-coffee items to retail locations. The company employs 140 people at its facility in Minden, Nevada.

Situation: Employees at Starbuck's Carson Valley Roasting Plant were working to minimize or eliminate their occasional packaging quality problem of ground coffee interfering with the packaging seal. "There were so many process variables that could impact a good seal," recalls Joe Berardino, World Class Performance Manager. Experimenting on a production and packing line posed a problem since traditional experimentation is done on only one factor at a time. The combinations for studying eight factors at three settings are 6,561, which would not only be expensive but would seriously hinder production. The company contacted the Nevada Industry Excellence, a NIST MEP network affiliate, for assistance.

Solution: Nevada Industry Excellence representative, Diane Updegrove, conducted Statistical Process Control and Process Capability training, which demonstrated to Starbucks personnel that Design of Experiments (DOE) was the answer to their complex system. DOE uses pre-set combinations and statistics to predict the results of those combinations and could provide significant results with minimal runs. Plus, unlike single factor experimenting, designs identify if there are interacting effects between variables providing even more meaningful results.

The cross-functional project team narrowed their suspect process variables to a list of eight. They identified a high, medium and low setting for each variable to run a Tagguchi L-27 Experimental Design. They also identified four quality measures that would respond to the effects of factor changes. Starbucks ran a total of 18 runs and gained some valuable information. This statistical experimental design quickly showed that one process variable contributed to the majority of the variation they were observing and five of the process variables had little to no effect on the responses, making procedural changes evident without much experimental costs.

Totally unexpected from the experiment was the finding that one portion of the process was so robust that a significant reduction in material use was observed. Additionally, because the process was proven in the control experiment, finished product sampling was significantly reduced saving time, product and money on the packaging line. An 88.3 percent reduction in affected packages per million has been observed since the process improvements were made, and costs associated with the experimentation were reduced by 99.7 percent by using the statistical experimental design.

Results:    
* Reduced problem with affected packages by 88.3 percent.
* Reduced costs by 99.7 percent.

Testimonial: "We have a long way to go; however; we are excited about the opportunities ahead of us to apply Statistical Process Control and Design of Experiments to other areas of our plant."
 - Joe Berardino, World Class Performance Facilitator

Activity Period: 2004-3