Taguchi is a statistician's statistician, known for extending the realm
of Statistical Quality Control, particularly in Design of Experiments,
with a branch now known as 'Taguchi methods'.
Taguchi characterized products and processes in terms of factors that
can be separated into two types:
- Controllable (or design) factors (CF) are those whose values
may be set or easily adjusted by the designer or process engineer.
These subdivide into:
- Target control factors (TCF) or signal factors are
those which the average levels of the response of interest.
- Variability control factors (VCF) affect the variability in
- Cost factors affect neither the mean response nor the
variability and can be adjusted to fit economic requirements.
- Uncontrollable (or noise) factors (NF), which are
sources of variation often associated with the production or
operational environment. Overall performance of the final product or
process should be designed to be insensitive to variation in noise
factors. These can be internal noise factors (from
deterioration and manufacturing imperfections) or external noise
factors (from operating conditions and environment).
He defined quality as:
"The loss imparted to society from the time the
product is shipped."
This includes soft factors such as customer dissatisfaction, reputation
His Quality Loss Function views 'in spec' as inadequate and uses
a (parabolic) calculation that gives an increasing value as it approaches
the central points between upper and lower specification limits.