Geoffrion and Powers (1980) present the main
reasons for the need of optimization capability when
solving comprehensive distribution planning problems.
It allows the analyst to evaluate different
'what-if' questions that must be considered when
studying the flexibility of the model. Among these
what-if questions, we have changes in demands and
costs, closing of facilities and transportation channels,
plant capacity expansions, introduction of a
new product or elimination of one, inventory policy
comparisons, tradeoffs between customer service and
costs, and changes in echelon structures.
Lee and Billington (1992) present a qualitative
discussion concerning the pitfalls and opportunities
of managing supply chain inventory. The main pitfalls
that can influence model formulation for a
global supply chain are the inadequate performance
measures for the complete supply chain, the inadequate
definition of customer service, the exclusion of
uncertainties, poor coordination, the consideration of
incomplete methods of shipment, the incorrect treatment
of inventory costs, and the separation of supply
chain design from operational decisions.
Ballou (1994) states the importance of the approach
that analysts adopt when aggregating customers
in location studies. He presents two major
sources of error: the estimation of the total transportation
cost to the cluster rather than to each
customer individually, and the allocation of customers
to facilities based on clustered demand instead
of individual demands. The main goal of Ballou's
research is to determine the influence of the
number of clusters, their size, and the number of
source points, on the total error obtained in location models for the continental United States. He provides
a table to select the minimum number of
clusters that guarantees a maximum allowable transportation
cost error. The author also states that
grouping customers by proximity is a reasonable
way to form clusters, and controlling cluster size can
reduce the costing error. However, according to the
author, the results of this research are not easy to
generalize to other product types and other regions
of the world, because the costing error is very sensitive
to many variables. Ballou claims that the general
patterns found in the study, nevertheless, should be
valid in other scenarios.