way to do this is to find a hot project. The “Six Sigma Changeover Reduction”
case study below illustrates this. The hot project was intended to relieve a bot-
tleneck—injection molding—by eliminating changeover time. The project was a
success and saved almost $300,000 a year in labor cost for changeovers.
Unfortunately from a lean perspective, the result of this was larger batches and
a lot more inventory of molded parts and a higher total cost. And the elaborate
Six Sigma approach only reduced changeovers to 1.2 hours, which is very far
from world class.
This is not to say the hot project approach should be completely dismissed.
First, it’s a way to get some quick results and earn a license to do more thought-
ful, longer-term lean system building—it’s money in the bank. Second, it’s
something you should do anyway when you’re well along on your lean journey.
Once basic lean systems are in place and there is a basic level of stability, flow,
and leveling, and when people are in teams and have developed good problem-
solving skills, they will often be working on hot projects. These will be the
objective of kaizen. But it will not be the driver for lean transformation. It will
be part of a more natural process of kaizen.
Case Study: Six Sigma Changeover Reduction—
Reducing Changeover Time to Break the Bottleneck^2
In an auto parts plant that makes headlamps for vehicles, a young
engineer was working toward her Six Sigma Black Belt. She selected
as a project a major problem that the plant had had for years: an
inordinate amount of time and resources focused on changing over
plastic injection molding machines. This made injection molding the
bottleneck in the process.
Detailed data were collected. The model changeovers averaged 3.5
hours. There were three changeovers per week, times 34 machines. This
resulted in lost production of about 100 hours per week. The target for
improvement was set at 2.5 hours per changeover, with anything longer
defined as a defect. The project goal was to reduce 50 percent of the
changeovers to less than 2.5 hours, thus cutting defects in half. A
stretch goal was set at 90 percent.
A lot of data analysis was done to determine the probability distribution
of changeovers; whether there were statistically significant differences
across shifts, machines, and different molds. The system of measuring
length of changeovers and process stability were both statistically
verified, and a detailed process map for the changeovers was devel-
oped. Various statistical concepts were used like paired t-tests, Weibull404 THETOYOTAWAYFIELDBOOK
(^2) We would like to thank Lester Sutherland and Donald Lynch, who shared this case study with us.