Page 74 - Orlicky's Material Requirements Planning

CHAPTER 4 Inventory in a Manufacturing Environment 55

version stage. In the example, (purchased) steel is made into a forging blank, which, in
turn, is machined into a gear that then becomes one of a number of components used in
assembling the gear box, a major component of a transmission. The transmission will be
required for the building of some end-product vehicle, which is also an assembly.
The demand for the end product, it should be noted, may have to be forecast if insuf-
ficient backlog of orders is on hand for the end product. But none of the component items,
including raw materials, need to be forecast separately. When someone manufactures
wagons, for instance, he or she may have to forecast how many he or she will sell and
when. Having done this, however, the manufacturer need not forecast the wheels because
he or she knows that there are four wheels per wagon. This seems elementary, but the
point is that the wagon wheels can be forecast independently, and the most sophisticated
statistical techniques can be employed for this purpose. Some manufacturing companies
do, in effect, just this. The results, of course, are bound to prove disappointing.
Min/max stock replenishment systems or kanbans are oblivious to the relationship
between inventory items and to their dependence on one another. This order-point
approach looks at the demand behavior of every inventory item as though it had a life of
its own. This, however, is a totally false premise in a manufacturing environment.
Statistical forecasting, on which order-point systems depend, addresses only the prob-
lem of individual item demand magnitude, but for purposes of manufacturing, an added
requirement is that component inventory represent matched sets. When components are
forecast and ordered independent of each other, their inventories will tend not to match
assembly requirements, and the cumulative service level will be significantly lower than
the service levels of the parts taken individually. This is caused by the adding up of indi-
vidual forecast errors of a group of components needed at one time to make an assembly.
If the probability of having one item in stock at a time of need is 90 percent, two
related items needed simultaneously will have a combined probability of only 81 percent
(0.9 0.9 0.81). At 10 such items, the odds are against all of them being available (34.8
percent probability of all being available simultaneously). Even with the service level set
slightly higher at 95 percent, the probability of simultaneous availability for 10 different
component items is less than 60 percent, and at 14 items, it drops below 50 percent, as
shown in Table 4-2.
These combined probabilities make it evident that when an assembly is to be built
from 20 or 30 different components (a proposition not unrealistic—it would hold rough-
ly true for the gearbox in the preceding example) ordered by an order-point system; the
lack of a shortage (and the lack of a need to expedite) actually would be a fluke. Note that
these shortages are not caused by some unforeseen events but are generated by the
process itself.

“Lumpy” Demand
Another dimension of demand to be considered is its relative continuity and uniformity.
Order point, as already mentioned, assumes more or less uniform usage, in small incre-