Inventory Policy Decisions
Inventory Policy Decisions
Inventory Policy Decisions
Reduce Costs
Although holding inventories has a cost associated with it, it can indirectly reduce
operating costs in other activities and may more than offset the carrying cost.
First, holding inventories may encourage economies of production by allowing
larger, longer and more level production runs. Production output can be
decoupled from the variation in demand requirements when inventories exist to
act as buffers between the two.
The material in this handout is extracted from Chapter IV, Business Logistics Management by
Ronald H..Ballou, Prentice-Hall International Inc. Fourth Edition,
TYPES OF INVENTORIES
Inventories can be categorized in five distinct forms. First, inventories may be in
the pipeline. These are inventories that are in transit between stocking or
production points because movement is not instantaneous. Where movement is
slow and/or movement is over long distances and/or movement must take place
between many stages, the amount of inventory in the pipeline may well exceed
that held at the stocking points.
Second, some stocks may be held for speculation, but they are still part of the
total inventory base that must be managed. Raw materials such as copper, gold,
and silver are purchased as much for price speculation as they are to meet
operating requirements. Where price speculation takes place for time periods
beyond the foreseeable needs of operations, such resulting inventories are
probably more the concern of financial management than logistics management.
However, when inventories are built up in anticipation of seasonal selling or
occur due to forward buying activities, these inventories are likely to be the
responsibility of logistics.
Third, stocks may be regular or cyclical in nature. These are the inventories
necessary to meet the average demand during the time between successive
replenishments. The amount of cycle stock is highly dependent on production lot
rises and falls through their life cycles, many products have a selling life that is
sufficiently long to be considered infinite for purposes of planning. Even though
brands turn over at the rate of 20 percent per year, a life cycle of three to five
years can be long enough to justify treating them as having a perpetual demand
pattern.
On the other hand, some products are highly seasonal or have a one-time, or
spike, demand pattern. Inventories that are held to meet such a demand pattern
usually cannot be sold off without deep price discounting. A single order for
inventories must be placed with little or no opportunity to reorder or return
goods if demand has been inaccurately projected. Fashion clothing, Christmas
trees, and political campaign buttons, are examples of this type of demand
pattern. ,
Similarly, demand may display a lumpy, or erratic, pattern. The demand may be
perpetual, but there are periods of little or no demand followed by periods of
high demand. The timing of demand is not as predictable as for seasonal
demand, which usually occurs at the same time every year. Items in inventory
are typically a mixture of lumpy and perpetual demand items. A reasonable test
to separate these is to recognize that lumpy items have a high variance around
their mean demand level. If the standard deviation of the distribution of demand,
or the forecast error, is greater than the average demand, or forecast, the item
is probably lumpy. Inventory control of such items is best handled by intuitive
procedures or by mathematical procedures.
There are products whose demand terminates at some predictable time in the
future, which is usually longer than one year. Inventory planning here involves
maintaining inventories to just meet demand requirements, but some reordering
within the limited time horizon is allowed. Textbooks with planned revisions,
spare parts for military aircraft, and pharmaceuticals with a shelf life are
examples of products with a defined life. Because the distinction between these
products and those with a perpetual life is often blurred, they will not be treated
differently from perpetual-life products for the purposes of developing a
methodology to control them. Finally, the demand pattern for an item may be
derived from demand for some other item. The demand for packaging materials
is derived from the demand for the primary product. The inventory control of
such dependent demand items is best handled by materials requirements
planning (MRP) and distribution requirements planning (DRP) procedures.
Management Philosophy
There are two basic philosophies around which inventory management is
developed. First, there is the pull approach. This philosophy views each stocking
INVENTORY OBJECTIVES
Inventory
management
involves balancing product
availability,
or
customer
service, on the one hand with
the costs of providing a given
level of product availability on
the other. Since there may be
more than one way of meeting
the customer service target,
we
seek
to
minimize
inventory-related costs for
each level of customer service
See Figure. Let us begin the
development of the methodology to control inventories with a way to define
product availability and an identification of the costs relevant to managing
inventory levels.
Product Availability
A primary objective of inventory management is to assure that product is
available at the time and in the quantities desired. This is commonly judged on
the basis of the probability of being able to fill a request for a product from
current stock. This probability, or item fill rate, is referred to as the service level,
and, for a single item, can be defined as
Service level =
1-
time. Therefore, the probability of filling the customer order completely can be of
greater concern than single-item service levels. For example, suppose that five
items are requested on an order where each item has a service level of 0.95,
that is, only a 5 percent chance of not being in stock. Filling the entire order
without any item being out of stock would be 0.95 x 0.95 x 0.95 x 0.95 x 0.95 =
0.77. The probability of filling the order completely is somewhat less than the
individual item probabilities.
A number of orders from many customers will show that a mixture of items can
appear on anyone order. The service level is then more properly expressed as a
weighted average fill rate (WAFR). The WAFR is found by multiplying the
frequency with which each combination of items appears on the order by the
probability of filling the order completely, given the number of items on the
order. If a target W AFR is specified, then the service levels for each item must
be adjusted so as to achieve this desired WAFR.
A specialty chemical company receives orders for one of its paint products. The
paint product line contains three separate items that are ordered by customers in
various combinations. From a sampling of orders over a period of time, the items
appear on orders in seven different combinations with frequencies as noted in
the Table. Also from the company's historical records, the probability of having
each item in stock is SL1=0.95; SL2=0.90; and SL3=0.80. As the calculations in
the Table show, the WAFR is 0.801. There will be about one order in five where
the company cannot supply all items at the time of the customer request.
Table :Computation of the weighted Average Fill Rate
Item
(1) Frequenc
(2) Probability of
(3)=(1)
(2)
Combination on
y
filling order
Marginal
Order
of Order
complete
value
A
0.1
(0.95)=0.950
0.095
B
0.1
(0.90)=0.900
0.090
C
0.2
(0.80)=0.800
0.160
A,B
0.2
(0.95)(0.90)=0.855
0.171
A,C
0.1
((0.95)(0.80)=0.760
0.076
B,C
0.1
(0.90)(0.80)=0.720
0.072
A,B,C
0.2
(0.95)(0.90)(0.80)=0.
0.137
684
1.0
WAFR 0.801
Relevant Costs
Three general classes of costs are
important to determining inventory
policy: procurement costs, carrying
costs, and out-of -stock costs. These
costs are in conflict, or in trade-off, with
each other. For determining the order
Capital Costs
Capital costs refer to the cost of the money tied up in inventory. This cost may
represent more than 80 percent of total inventory cost, yet it is the most
intangible and subjective of all the carrying cost elements. There are two reasons
for this. First, inventory represents a mixture of short-term and long-term assets,
as some stocks may serve seasonal needs and others are held to meet longerterm demand patterns. Second, the cost of capital may vary from the prime rate
of interest to the opportunity cost of capital.
The exact cost of capital for inventory purposes has been debated for some time.
Many firms use their average cost of capital, whereas others use the average
rate of return required of company investments. The hurdle rate has been
suggested as most accurately reflecting the true capital cost. The hurdle rate is
the rate of return on the most lucrative investments forgone by the firm.
Inventory Service Costs
Insurance and taxes are also a part of inventory carrying costs because their
level roughly depends on the amount of inventory on hand. Insurance coverage
is carried as a protection against losses from fire, storm, or theft. Inventory taxes
are levied on the inventory levels found on the day of assessment. Although the
inventory at the point in time of the tax assessment only crudely reflects the
average inventory level experienced throughout the year, taxes typically
represent only a small portion of total carrying cost. Tax rates are readily
available from accounting or public records.
Inventory Risk Costs
Costs associated with deterioration, shrinkage (theft), damage, or obsolescence
make up the final category of carrying costs. In the course of maintaining
inventories, a certain portion of the stock will become contaminated, damaged,
spoiled, pilfered, or otherwise unfit or unavailable for sale. The costs associated
with such stock may be estimated as the direct loss of product value, as the cost
of re working the product, or as the cost of supplying it from a secondary
location.
Out of Stock Costs
Out-of-stock costs are incurred when an order is placed but cannot be filled from
the inventory to which the order is normally assigned. There are two kinds of
out-of- stock costs: lost sales costs and back order costs. Each presupposes
certain actions on the part of the customer, and, because of their intangible
nature, they are difficult to measure accurately. A lost sales cost occurs when the
customer, faced with an out-of-stock situation, chooses to withdraw his or her
request for the product. The cost is the profit that would have been made on this
particular sale and may also include an additional cost for the negative effect
that the stock out may have on future sales. Products for which the customer is
very willing to substitute competing brands, such as bread, gasoline, or soft
drinks, are those that are most likely to incur lost sales. A back order cost occurs
when a customer will wait for his or her order to be filled so that the sale is not
lost, only delayed. Back orders can create additional clerical and sales costs for
order processing, and additional transportation and handling costs when such
orders are not filled through the normal distribution channel. These costs are
fairly tangible, so measurement of them is not too difficult. There also may be
the in- tangible cost of lost future sales. This cost is very difficult to measure.
Products that can be differentiated in the mind of the consumer (automobiles
and major appliances) are more likely to be back ordered than substituted.
When the tuna boats are sent to the fishing grounds, a packer of tuna products
must process all the tuna caught since storage is limited, and, for competitive
reasons, the company does not want to sell the excess of this valued product to
other packers. Therefore, this packer processes all fish brought in by the fleet
and then allocates the production to its three field warehouses on a monthly
basis. There is only enough storage at the plant for one month's demand. The
current production run is 125,000 lb.
For the upcoming month, the needs of each warehouse were forecasted, the
current sock levels checked, and desired stock availability level noted for each
warehouse. The findings are tabulated in the following Table
Warehouse
Current
Level
1
2
3
5,0001b
15,000
30,000
Stock
Forecasted
Demand
10,0001b
50,000
70,000
130,000
Forecast
Errora
(std.
Dev.)
2,0001b
1,500
20,000
Stock
Levelb
Availability
90%
95%
90%
From the normal distribution curve z= 1.28. Hence, the total requirements for
each warehouse would be 12,560 = 10,000 + (1.28 2,000). Other warehouse
total requirements are computed similarly.
Net requirements are found as the difference between total requirements and
the quantity on hand in the warehouse. Summing the net requirements
(110,635) shows that 125,000 -110,635 =14,365 which is the excess production
that needs to e prorated to the warehouses.
Proration of excess 14,365 lb is made in proportion to the average demand rate
at each warehouse. Average demand for warehouse 1 is 10,000 lb. against a
total demand rate for all warehouses of 130,000 lb. The proportion of the excess
allocated to warehouse 1 should be (10,000+ 130,000) (14,365) =1,105.
Compute the proration for the remaining warehouses in a similar manner. The
total allocation to a warehouse is the sum of its net requirements plus its
proration of the excess. The results are tabulated in the following Table.
Table: Allocation of Tuna production to Three warehouses
Warehouse
(1)Total
Requirements
(2) On Hand
1
2
3
12,560 lb.
52,475
95,600
160,635
5,000
15,000
30,000
(3) =(1)-(2)
Net
Requirements
7,560 lb.
37,475
65,600
110,635
(4) Proration
of Excess
1,105 lb.
5,525
7,735
14,365
(5)
=(3)
+(4)
Allocation
8,665 lb
43,000
73,335
125,000
This says that we should continue to increase the order quantity until the
cumulative probability of selling additional units just equals the ratio of the Profit
/ (Profit + Loss).
A grocery store estimates that it will sell 100 pounds of its specially prepared
potato salad in the next week. The distribution of demand is normally distributed
with a standard deviation of 20 pounds. The supermarket can sell the salad for $
5.99 per pound for the ingredients. Because no preservatives are used, any
unsold salad is given to charity at no cost.
Finding the quantity to prepare that will maximize profit requires that we first
compute CPn. That is
CPn. = Profit/ Profit + Loss = (5.99-2.50) / (5.99-2.50) + 2.50 = 0.583
From the normal distribution curve, the optimum Q* is at the point of 58.3
percent of the area under the curve (see the figure). This is a point where z
=0.21 The preparation quantity should be
Q* =100 lb. +0.21 (20 lb.) =104.2 lb.
When Demand is discrete, the order quantity may be between whole values. In
such cases, we will round up Q to the next higher unit to assure at least CPn is
met.
Number of parts
0
1
2
3
4
5
Frequency of Need
0.10
0.15
0.20
0.30
0.20
0.05
1.00
Cumulative Frequency
0.10
0.25
0.45
0.75
Q*
0.95
1.00
Q* =
IC
Q *
D
D
Q *
Replacement parts are supplied from an inventory by a manufacturer of
industrial machine tools. For a particular part, the annual demand is expected to
be 750 units. Machine set up costs are $50, carrying costs are 25 percent per
year, and the part is valued in inventory at $35 each. The economic order
quantity placed on production is
Q* =
2DS
2(750)(50)
=
= 92.58 or 93units
IC
(0.25)(35)
We can now introduce the idea of a reorder point, which is the quantity to which
inventory is allowed to drop before a replacement order is placed. Because there
is generally a time lapse between when the order is placed and when the items
are available in inventory, the demand that occurs over this lead time must be
anticipated. The reorder point (ROP) is
ROP = d x LT
where
ROP = reorder point quantity, units
d = demand rate, in time units
LT = average lead time, in time units
The demand rate (d) and the average lead time (LT) must be expressed in the
same time dimension.
Continuing the previous example for the machine replacement part, suppose that
it takes 1.5 weeks to set up production and make the parts. The demand rate is
d=750 (units per year) /52 (weeks per year)=14.42 units per week. Therefore,
ROP=14.42 1.5 = 21.6 or 22 units. We can now state the inventory policy:
When the inventory level drops to 22 units, place a replenishment
order for 93 units.
EOQ formula without changing the remaining cost and/or demand factors since
they remain constant. Notice the stability in values for Q*. If the incorrect order
quantity were used in these two cases, total would have been in error by only
0.11 percent and 0.62 percent, respectively.
Noninstantaneous Resupply
A built-in assumption to Ford Harris's original EOQ formula was that resupply
would be made instantaneously in a single batch of size Q*. In some
manufacturing and resupply processes, output is continuous for a period of time,
and it may take place simultaneously with demand. The basic sawtooth pattern
of on-hand inventory is modified, as shown in the Figure.
The order quantity now becomes the production run, or production order,
quantity (POQ), and we will label it Q p*. To find this, the basic order quantity
formula is modified as follows:
Q p* =
2D S
IC
p - d
where p is the output rate. Computing Q p only makes sense when the output
rate p exceeds the demand rate d.
Again, for the previous parts replacement problem, suppose that the production
rate for these parts is 50 units per week. The production run quantity is
Qp * =
2(70)( 50)
(0.25)( 35)
50
50 14.42
=
92.58 1.185 = 109.74, or 110 units
The ROP quantity remains unchanged.
point in time is the quantity on hand plus the stock on order less any
commitments against the inventory, such as customer back orders or allocations
to production. The entire quantity Q* arrives at a point in time offset by the lead
time. Between the time that the replenishment order is placed at the reorder
point and when it arrives in stock, there is a risk that demand wil1 exceed the
remaining amount of inventory. We control the probability of this occurring by
raising or lowering the reorder point, and by adjusting Q*.
In the Figure, the operation of the reorder point system is illustrated for a single
item where the demand during the lead time is known only to the extent of a
normal probability distribution. This demand during lead time (DDLT) distribution
has a mean of X' and a standard deviation of s'd. The values for X' and s'd are
usually not known directly, but they can be easily estimated by summing a
single-period demand distribution over the length of the lead time. For example,
suppose weekly demand for an item is normally distributed with a mean d = 100
units and a standard deviation of sd = 10 units. Lead time is 3 weeks. We wish to
roll up the weekly demand distribution into one 3-week DDLT distribution of
demand (see Figure below).
The mean of the DDLT distribution is simply the demand rate d times the lead
time LT; or X' = d x LT = 100 x 3 = 300. The variance of DDLT distribution is
found by adding the variances of the weekly demand distributions (see Figure
10-11). That is, s'd2= LT(sd2). The standard deviation is the square root of s'd2
which is .
s'd = sd LT = 10x 3 = 17.3. .
IC
2 ( 1 11
, 0 7 ) (1 0 )
=
( 0 .2 0 / 1 2 ) ( 0 .1 1)
1 1 , 0 0 8 u n i t s
ROP =d LT +z (s d)
where s d = sd LT = 3,099 1.5 = 3,795 units. The value for z is 0.67 from
Normal Tables where the fraction of the area under the normal distribution curve
is 0.75. Thus,
ROP = 11,107
1.5 + 0.67 3,795 =19,203 units. So, when the effective
inventory level drops to 19,203 units, place a reorder for 11,008 units.
It is not unusual for the reorder point quantity to exceed the order quantity, as
was the case in the example shown above. This frequently happens when lead
times are long and/or demand rates are high. To make the reorder point control
system work properly, we simply must make sure that in deciding when to
trigger a replenishment order, we base the decision on the effective inventory
level. Recall that the effective inventory level requires that we add all stock on
order to the current quantity on hand when deciding whether the reorder point
has been penetrated. When ROP > Q*, the result of this procedure is that a
second order will be placed before the first arrives in stock.
Average Inventory Level
The average inventory level (AIL) for this item is the total of the regular stock
plus safety stock. That is,
Average inventory = Regular stock + Safety stock
AI L = (Q/2) +z(s'd )
For the previous Tie Bar problem, the average inventory would be
AIL = 11,008 / 2 + 0.67 3,795 = 8047 units
TC =
D
Q
S + IC
Q
2
+ I C z s' d +
D
S
k s' d E(z )
where k is the out-of-stock cost per unit. The stock out cost term requires some
explanation. First, the combined term of s'd E(z) represents the expected number
of units out of stock during an order cycle. E(z) is called the unit normal loss
integral whose values are tabled as a function of the normal deviate z. Second,
the term D/Q is the number of order cycles per period of time, usually a year.
Hence, the number of order cycles times the expected number of units out of
stock during each order cycle gives the total expected number of units out of
stock for the entire period. Then, multiplying by the out-of-stock cost yields the
total period cost.
Continuing the Tie Bar example, suppose the stockout cost is estimated at $0.01
per unit. The total annual cost for the item would be
1 1 , 1 0 7 ( 1 2 )( 1 0 )
TC =
1 1 , 0 0 8
1 1,0 0 8
+ 0 . 2 0 ( 0 .1 1) (
1 1,1 0 7 ( 2 )
) + ( 0. 2 0 ) ( 0 .1 1) (0 . 6 7 ) ( 3 7 9 5 ) +
1 1,0 0 8
( 0 .0 1) ( 3 7 9 5 ) (0 .1 5 0 )
(z)
=E
(0.67)
Service Level
The customer service level, or item fill rate, achieved by a particular inventory
policy was previously defined. Restating it in the symbols now being used, we
have
S L = 1-
(D / Q ) ( s ' d E (z ))
D
= 1-
s ' d E ( z ))
Q
= 0 .9 4 8
1 1,0 0 8
That is, the demand for Tie Bars can be met 94.8 percent of the time. Note that
this is somewhat higher than the probability of a stock out during the lead-time
of P = 0.75.
tray is brought from reserve storage and inserted into its position. This action is
the trigger to place a replenishment order on production. Little or no paper work
is needed to make a rather sophisticated inventory control system operate
effectively.
Q =
IC
2. Compute the probability of being in stock during the lead time from
P = 1-
QIC
D k
Find s'd. Find the z value that corresponds to P in the normal distribution table.
Find E(z) from the unit normal loss integral table.
3. Determine a revised Q from a modified EOQ formula, which is
2 D [ S + k s' d E ( z ) ]
Q =
IC
2D S
Q =
IC
2 (1 1,1 0 7 ) ( 1 0 )
=
( 0 . 2 0 / 1 2 ) ( 0 .1 1)
1 1 , 0 0 8 u n i t s
Estimate P
P = 1-
QIC
D k
= 1
11,008(0.2 0)(0.11)
0 . 8 2
1 1 , 1 0 7 ( 1 2 )( 0 . 0 1 )
From Normal Table [email protected] = 0.91 from Normal Loss Table E(0.91) =0.0968.
Revise Q
E( z ) ]
2 ( 1 1 , 1 0 7 ) ( 1 2 ) [ 1 0 + 0 . 0 1 ( 3 7 9 5 )( 0 . 0 9 6 8 ) ]
=
IC
= 1 2 , 8 7 2
0 . 2 0 ( 0 . 1 1 )
u n it s
Revise P
P = 1
12,872(0.2 0)(0.11)
0 . 7 9
1 1 , 1 0 7 ( 1 2 )( 0 . 0 1 )
Now, z
@ 0.79
Revise Q
2 ( 1 1 , 1 0 7 ) ( 1 2 ) [ 1 0 + 0 . 0 1 ( 3 7 9 5 )( 0 . 1 1 8 1 ) ]
Q =
= 1 3 , 2 4 6
0 . 2 0 ( 0 . 1 1 )
u n it s
We continue this revision process until the changes in P and Q are sufficiently
small that further calculation is not practical. The final results are P=0.78:
Q*=13,395 units; ROP =19,583 units with a total relevant cost of TC =$15,019
and an actual service level of SL =96%.
The Reorder Point Method with Demand and Lead Time Uncertainty
Accounting for uncertainty in the lead time can extend the realism of the reorder
porn' model. What we wish to do is find the standard deviation (sd) of the DDLT
distribution based on uncertainty in both demand and lead time. This is found by
adding the variance of demand to the variance of lead time, giving us a revised
formula for s'd of
2
s'
L T s + d
L T
s'
1. 5 ( 3 0 9 9 )
+ 1 1 1 0 7 ( 0 .5 )
= 6 ,7 2 7 u n i t s
Combining demand and lead-time variability in this way can greatly increase s'd
and the resulting safety stock. Brown warns that demand and lead-time
distributions may not be independent of each other. Rather, when a
replenishment order is placed, a fair idea is known as to the lead-time for that
order. Therefore, application of Equation (10-18) may lead to an overstatement
of s'd and the resulting amount of safety stock. If lead times do vary
Alternately and less precisely, the longest lead time may be used as the average
lead time with s LT set at zero (0). The standard deviation is then computed as s'd
= sd LT.
Suppose inventory is to be maintained on a distributor's shelf for an item whose
demand is forecasted to be d=100 units per day and s d =10 units per day. A
reorder point is the method of inventory control. There are multiple points
throughout the supply channel where time is incurred in the product flow
between source point and customer. The distributions of these times that form
the order replenishment lead time are shown in the following figure. No
significant amounts of inventory are maintained at the pool point or in the trucks.
Solution
The reorder point n
i ventory control method applies. However, determining the
statistics of the demand-during-lead-time distribution requires taking the leadtime for the entire channel into account.
Recall:
2
s'
L T s + d
L T
Where
2
= s
L T
+ s
p
+ s
i
LT
'
7 x1 0
+ 1 0 0
x 1. 3 5 =
1 4 , 2 0 0 = 1 1 9 .1 6 u n i t s
and
Finally, the average inventory level is
1 2 0 8
AIL =
+ 2 . 3 3 (1 1 9 .1 6 ) = 8 8 2 u n i t s
2.
3.
4.