Page 161 - Six Sigma Demystified

142 Six SigMa DemystifieD

• Person entering order. Don, who is on the team, was chosen to represent
the more experienced order taker, whereas Karen is less experienced yet
comfortable with the current procedures.
• Number of line items. The minimum possible number of items per order
is one; the maximum usually experienced is four to five items, so five
was chosen.
• Time between call arrival. after some discussion, the team decided that this
single factor would adequately address the intended effects of Time of
Day/Call Volume and Number of Available Staff factors given the manner in
which the experiment would be conducted. after a brief review of phone
records, the maximum incoming call rate for a single staff member was set
to six calls per hour, with the minimum set to zero. During the experiment,
these values would be used to interrupt the order processing (to simulate
a new call) for seven minutes (an average sales inquiry call length) every
10 minutes (equivalent to six calls per hour).

a standard half- factorial (eight- run) experiment for the four factors was repli-
cated in random order (for a total of 16 experimental runs using 16 different
orders selected to match the design conditions). Using a 0.05 level of signifi-
cance for each parameter, only the Number of Line Items, the Time Between Call
Arrival, and the interaction of these two factors were found to be significant.
The regression accounted for nearly 80 percent of the total variation (using the
2
calculated R adjusted statistic). No anomalies were discovered in the residuals
analysis.
The results suggested two pieces of information that the team found inter-
esting:

1. The order processing time is influenced not only by the number of

line items but also by the number of interruptions. Furthermore, the
impact of interruptions is larger for more line items. Recall that the
actual wait time associated with each interruption was removed
from the response. These results suggest that the process proceeds
at a slower rate in the presence of interruptions than it does when
there are no interruptions.
2. The product family was not significant. This seemed to contradict the
baseline data, where the three products had very different cycle times.
in discussing these results, the team wondered if perhaps the differ-
ences in baseline data actually were a reflection of the number of line
items, which might be correlated with product family to some extent. a
review of the baseline data confirmed this result.