Page 245 - Six Sigma Demystified

Part 3 s i x s i g m a to o l s 225

The factors that generally are not controlled in your operations are sometimes
called subsidiary factors. Taguchi referred to these as noise factors or the outer
array. Examples include ambient temperature, humidity, and vibration. As men-
tioned earlier, it is preferred to control these factors for the experiment.
Factors are selected for the designed experiment by brainstorming among
team members. This typically will result in a long list of potential factors. For
an effective yet small screening design, you’d like to limit the design to five or
seven key factors. Even though this may seem “impossible” for your process, it
is often the best practice. If cost and time are of little concern, then add more
factors as necessary. However, when cost or time is limited, the number of fac-
tors can be reduced using the nominal group technique or prioritization matrix.
Alternatively, you could decide to hold some factors constant, effectively
excluding them from the analysis. Factors that are neither held constant nor
included are potential lurking factors.
Lurking factors may be a source of bias or error in the analysis. Bias causes
you to confuse the effect of a factor with that of another factor, particularly if
the lurking factor happens to be coincident with another factor. In other cases,
it just limits the usefulness of the results, such as if you do not include the
effects of communications skills in a customer-support experiment.
You always should randomize the order of the trials to prevent any bias in
the estimates. In some cases, however, you cannot fully randomize the experi-
mental trials and, instead, run the experiments in blocks. Examples of blocking
factors include the day of the week, a batch of material, a run of the furnace, an
airline flight, and so on. In each of these cases, you may have multiple run con-
ditions that can be randomized within the block, but these blocking factors
cannot be randomized within the entire experiment.

Consider the baking of cookies. You might vary a number of parameters, such
as the ingredients (margarine versus butter, chocolate chips versus M&M’s), the
cooking temperature, and the cooking time. While the ingredients can be varied
within each batch, the time or temperature within a batch cannot be varied. In
this case, time and temperature are coincident with the batch and thus are
blocking factors.
In other cases, you cannot run all the sample combinations within a given
batch simply because the limited size of a batch (e.g., the size of the oven)
permits only so many factor combinations. In this case, the batch itself becomes
a blocking factor. The differences you see between the factor combinations in
batch 1 and batch 2 may be due to the factor effects or to the batch-to-batch
variation.