280 ◾ Appendix 11: Gage Repeatability and Reproducibility Measurement
Let’s analyze the measurement system this time in its ability of assessing
if the force meets the requirements of the customer. Later, we will assess the
measurement system’s capability in process improvement.
We now know the quality of the measurement system in the report in
Appendix 11: Gage Repeatability and Reproducibility Measurement ◾ 281
The measurement system is not capable of determining good parts from
bad. See the vertical marks in the scale named, “Can you sort good parts
from bad?”—42.5% of the variation is from the error in the measurement
system. Thus, it is difﬁcult to tell what the force actually was. The chart at
the bottom left suggests we should improve the Reproducibility of the mea-
surement system for the biggest improvement. The three operators are not
in agreement on the forces. Look for a bias between operators, such as one
operator consistently measuring a lower force at disconnect than the other
two. Then, use your DMAIC (demand, measure, analyze, improve, control)
training to determine the root cause of the variation, improve reproducibil-
ity, and perform a Measurement System Analysis (MSA) again to see if the
system is now capable.
Next, we will use the Gage R&R to assess the measurement system’s
ability to measure process improvement regardless whether the parts meet
the customer’s requirement. Highly reliable organizations meet the customer
requirements very easily and consistently and want to continuously improve
to become even more reliable. They need measurement systems that can tell
them if they are really improving.
Read the scale in the top left named, “Can you adequately assess pro-
cess performance?” Again, the answer is “no” because 41.3% of the variation
measured is not from the parts. It is from the measurement system itself.
Reproducibility is the major error. We would like the percent to be 10 or
less, especially if personal injury may occur if the force is not what we think
To demonstrate improving a measurement system, let’s take the sugges-
tion to look at each of the three operators and see if there is one operator
consistently recording a lower or higher value than the others. This is not to
suggest the one operator is wrong. Perhaps the other two have more error
and are not as well trained as the one operator. Figure A11.12 is a simple
statistical analysis showing if there is a statistically signiﬁcant difference
(bias) between operators.
282 ◾ Appendix 11: Gage Repeatability and Reproducibility Measurement
Boxplot of Measurements
Jan Adam Janelle
This is a boxplot chart. A boxplot chart shows the central tendency
(median and mode) and the variation. The “box” represents the middle two
quartiles of the data. The “whiskers” represent the remaining two quartiles
of the data. The horizontal line in the box is the median and the circle
within the box is the mean. Both the median and mean for Janelle is below
the median and mean for Adam and Jan. This suggests a bias. Again, maybe
Janelle is more correct than Adam and Jan. We need to apply our DMAIC
and ﬁnd out what the root causes are and improve.
After we think that we improved the system, we will run another MSA to
see if we can get the percent of Process and percent of Tolerance closer to