You are on page 1of 23

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No.

2, 2005

119

A COMPARISON OF THE ECONOMIC ORDER QUANTITY AND QUICK


RESPONSE INVENTORY REPLENISHMENT METHODS
by
Walter Zinn
The Ohio State University
and
John M. Charnes
The University of Kansas
Logistics managers have at their disposal at least two well-known competing methods of
determining order quantity: Economic Order Quantity (EOQ) and Quick Response (QR). Each
method takes a distinct approach and makes substantially different recommendations for order
quantity. Although both methods are discussed extensively in the literature and included in logistics and operations management textbooks, a comprehensive comparison of the two methods is
not currently available in the logistics literature. Accordingly, the goal of this paper is to compare
the EOQ and QR methods in order to propose a set of rules by which managers can select the most
appropriate method to use under various circumstances.
The EOQ method, introduced by Ford W. Harris (1915), balances the cost of ordering an
item with the inventory holding cost for that item. Given a known level of annual demand, a firm
must balance the cost of ordering smaller quantities more frequently to minimize holding cost,
against the cost of making a smaller number of larger-quantity purchases to minimize ordering
cost. The EOQ determines the optimal order quantity that minimizes the joint cost of ordering and
holding inventory.
In contrast, the QR method considers only the inventory holding cost and ignores the ordering cost. QR is a general term that describes the criteria used to determine the order quantity in
several well-known rapid-replenishment inventory management methods such as Just-In-Time
or Continuous Replenishment. The QR order quantity is just enough to support operations until the
next delivery. Therefore, the QR quantity is the product of the daily demand and the time between
deliveries, in days. Thus, each inventory replenishment method ignores factors that are included in
the other method. While the EOQ method ignores the time between deliveries in the determination
of order quantity, the QR method ignores three variables included in the EOQ: the cost of an order,
the products unit value, and the unit cost of holding inventory. Note that there are EOQ models

120

ZINN AND CHARNES

available that include safety stock in the computation of the EOQ. However, these models are outside the scope of this research in that we consider base stock only.
This comparison is timely, as the drive to reduce inventory investment is leading firms towards
use of QR and away from the EOQ. One reason for inventory reduction is that a growing number
of firms today are increasingly concerned with the risk associated with stocking the wrong inventory and not just the cost of holding it. This risk is currently not included explicitly in either the EOQ
or QR. As a result, some have questioned the value of the EOQ and argue that it has outlived its usefulness (Woolsey 1988). On the other hand, the EOQ remains a part of business school curricula and
continues to be used by some managers. Which method is better? Should firms always move
towards QR? Are there circumstances under which managers should favor the EOQ over QR? Our
goal is to answer these questions.
The equations used to compute the EOQ and QR quantities are below.

QEOQ =

720 Pd
HV

where:
QEOQ = Quantity delivered for the EOQ method
d
= Average daily demand in units
P
= Cost of an order
H
= Average annual cost of holding inventory
V
= Unit product value
QQR = td
where:
QQR = Quantity delivered for the QR method
t
= Time between deliveries, in days
We make the comparison between the EOQ and QR methods with the following assumptions:
1. Both methods are applied in a continuous review inventory system with deterministic
order quantity and deterministic time between deliveries where product is made to
stock.
2. The scope of the comparison is limited to base or cycle stock.
3. The cost of an order (P) is assumed to be the same under both inventory replenishment
methods.
The third assumption is needed because the cost of an order is measured differently when
each method is applied. The cost of an order is defined in the literature to be all out-of-pocket costs
incurred when an order is issued, starting with product search and negotiation. They also include order
transmittal, receiving product, placing it in storage, and processing the invoice for payment (Stock
and Lambert 2001). Freight charges and freight audit costs are also included (Lambert and Bennion
1986). In the EOQ method, product search and negotiation costs are higher because each order is

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

121

negotiated independently. In contrast, in the QR method product search and negotiation costs are
lower because individual orders are issued under a long-term supply contract. This reduces the
number of suppliers included in the search and limits the number of items to be negotiated with the
selected supplier. On the other hand, the QR method has a higher cost of receiving and handling product, and of freight and audit costs because the QR implies a larger number of smaller orders than the
EOQ method. Thus, the differences in the cost of an order between the two methods are assumed
to cancel out.
A final issue to note before comparing the costs of the EOQ and the QR methods is the relationship between QQR and QEOQ. The EOQ order quantity, QEOQ, is always greater than or equal to
the QR quantity, QQR, because in the QR method the firm receives only enough inventory to cover
the demand until the next delivery. If there is any less, the inventory will run out prior to the next
delivery. Recall that we assume a fixed quantity and fixed time between deliveries replenishment
system.
The next section presents a review of related work comparing the EOQ with Just-In-Time
(JIT) methods. This is followed by a section examining reasons for the shift by practitioners toward
QR and away from the EOQ. A quantitative comparison between the two inventory replenishment
methods follows. The final two sections contain a discussion of results and our conclusions and managerial implications.
LITERATURE REVIEW
Fazel (1997) computed the cost difference between the EOQ and Just-In-Time (JIT) methods
and estimated the demand and purchase price indifference points between the EOQ and JIT. He developed a model to estimate the demand above which EOQ is less costly than JIT and the maximum
unit purchase price below which JIT is preferable.
In a follow-up study, Fazel, Fischer, and Gilbert (1998) suggested that JIT should be used
at lower levels of demand whereas EOQ should be used at higher levels of demand. They also
identified factors that affect the cost difference between JIT and EOQ-based solutions. These two
studies considered the cost but not the risk of holding inventory.
Schneiderjans and Cao (2000 and 2001) expand on the Fazel, Fischer, and Gilbert (1998)
model by adding as a variable the reduction in space requirements caused by JIT. They conclude that
a JIT method is almost always preferable to an EOQ-based method, whether the demand is high or
low, because JIT always requires less warehouse space than the EOQ method.
Ramasesh (1990) developed a decision model to help companies decide whether to move
from an EOQ method to a JIT method. The model compares the fixed cost of initiating a JIT method
(such as the development of sources of supply, installation of quality assurance procedures, negotiating long-term contracts, etc.) with the savings obtained from operating a JIT method that has a
lower cost than the traditional EOQ method. Ramasesh concludes that the incentive to move away
from the EOQ is greater when the demand is large and the order cost is low.

122

ZINN AND CHARNES

Instead of choosing between the two methods, Johnson and Stice (1993) propose a hybrid
that they call Not Quite JIT. Their idea is to alleviate some of the costs and risks associated with
JIT by proposing a method that combines a lower level of inventory with JIT deliveries. The authors
list ordering and transaction costs as some that are higher in the JIT methods. They also list the potential for stockouts and the inflexibility associated with being locked into long-term contracts as risks
inherent to JIT.
FROM THE EOQ TO QR
In recent years, a growing number of firms adopted the QR method in an attempt to significantly
reduce or even eliminate inventory. As noted earlier, the QR order quantity is less than or equal to
the EOQ quantity. The goal of QR is to minimize only the cost of holding inventory, while the goal
of the EOQ is to minimize the joint cost of ordering and holding inventory. QR is therefore suboptimal because it disregards the cost of ordering. Thus, adopting QR apparently counters the notion
of integrated logistics in that it aims to minimize a single cost. Why, then, are firms increasingly adopting QR? We see two main reasons: the risk of holding inventory and the decline in the cost of an order.
Each is discussed.
The Risk Related to Holding Inventory
When holding inventory, firms incur the risk that inventory will lose value as a result of
changes in product technology or in consumer preferences. For example, computer retailers holding a large inventory incur the risk of having to sell product at a significant discount if the inventory is made obsolete by a new wave of technological innovation. Similarly, product proliferation
in industries such as automotive, packaged goods, or apparel offer consumers a wide choice of
products. This increases the likelihood that firms will hold the wrong inventory because the number of choices available complicates the task of forecasting consumer preference. Apparel merchandisers, for instance, often sell product at large discounts to reduce the inventory of low demand
items. Thus, by purchasing smaller quantities per order, retailers shift the risk of holding inventory
to suppliers.
Unless the risk of holding inventory is made explicit, the EOQ method will always appear to
be preferable to the QR method. As mentioned earlier, the reason is that the EOQ method optimizes the joint cost of ordering and holding inventory while the QR method minimizes only the holding cost. Thus, the need to manage the risk of holding inventory may substantially explain the shift
from the EOQ to the QR observed in many companies.
Our operational definition of risk (R) is the percent decline in the value of a unit of product resulting from changes in product technology or consumer preference. For example, R has a value of .3
if a computer part is sold at a 30 percent discount after it is made obsolete by a newly introduced
substitute part. While the determination of R for each product is beyond the scope of this research,
one could estimate it on the basis of historical discount values or with the help of subject matter experts.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

123

The Decline in the Cost of an Order


A second reason for the increased adoption of QR is the decline in the cost of an order (P).
Advances in computer technology triggered a decline in the cost of processing information, which
in turn has led to a significant decline in the cost of an order. As a result, the value of the EOQ declined
(Jones 1991). This helps to explain the decreased usage of the EOQ because it makes the cost of an
order considerably less important than the inventory holding cost, thereby reducing the importance
of considering the trade-off between the two costs.
THE TOTAL COST UNDER THE EOQ AND QR
As noted above, the purpose of this research is to compare the total cost of ordering and holding inventory under the alternatives of selecting either the EOQ or QR as a method to determine order
quantity. In the case of the QR, the total ordering and holding costs are:

CQR =

360dP QQR
+
VH
QQR
2

(1)

where:
CQR = Total cost if the firm buys the QR quantity
In the case of the EOQ, the total cost function is similar, except for a term added to capture the
monetary expression of the risk of holding inventory. We call this term the Expected Loss of Value
(ELV). The ELV is the product of the number of units of inventory at risk and the expected loss per
unit. The inventory at risk is the difference between the EOQ order quantity and the QR order
quantity (QQR-QEOQ). Recall that QQR is the minimum amount that a firm may order so that QEOQ
always equals or exceeds QQR. Therefore, the greater the quantity ordered beyond QQR, the higher
the inventory at risk. The potential loss per unit of inventory at risk is the product of the value of each
unit (V) and risk (R), the expected discount needed to sell the unit. The risk, R, varies from 0 to 1,
where 1 means 100 percent discount and 0 means no discount. Thus:
ELV = (QEOQQQR)VR

(2)

where:
ELV = Expected loss of value
R
= Risk (expected discount needed to sell the unit)
The total cost of ordering, holding and assuming risk in the EOQ method is:

C EOQ =

360 dP QEOQ
+
VH + (Q EOQ QQR )VR
Q EOQ
2

where:
CEOQ = Total cost if firm buys the EOQ quantity

(3)

124

ZINN AND CHARNES

A COST COMPARISON OF THE EOQ AND QR


In this section, we compare the total cost of using the EOQ vs. QR. Our goal is not just to compute the lowest cost, but principally to identify the impact of different variables on the decision. This
will ultimately enable managers to understand the differences between the two inventory replenishment
methods and to select one method or the other in a given situation.
The object of the comparison is to compute the difference between the total costs of adopting
the EOQ vs. QR. The difference is denoted as C = CQRCEOQ. When C > 0, firms should
adopt the EOQ method. In contrast, firms should adopt the QR method whenever C < 0. The cost
difference, C, can also be expressed as a function of the six variables defined previously in
Equations 1 and 3:

C =

360dP QQR
360dP QEOQ
VH
+

VH (QEOQ QQR )VR


QQR
QEOQ
2
2

(4)

The above equation has five terms, each corresponding to an identified cost. The first two
terms represent the cost of ordering (OCQR) and the holding cost (HCQR) in the QR method. The
next two terms represent the ordering (OCEOQ) and holding (HCEOQ) costs in the EOQ method. The
last term represents the expected loss of value (ELV).
We begin our cost comparison by selecting realistic ranges of values for each of the variables.
These values are based on a search of the literature for values of the same variables used in related
research projects. The input values are listed in Table 1 and the justification for their selection is given
in the next subsection of the paper. This is followed by the data analysis, which is presented in two
stages in the results section. In the first stage, we investigate the general relationship between each
of the 6 variables considered here and the cost differential C. The general relationship is categorized by the changes in the value of C with respect to each of the 6 variables. The general relationship
is also examined with respect to the five cost terms included in Equation 4, viz., OCQR, HCQR,
OCEOQ, HCEOQ, and ELV. At the end of this stage, we establish whether the general relationship
between each of the six variables and C is positive or negative.
However, because of interactions among the six variables, the general relationship may not
hold for the entire range of values considered. Therefore, in the second stage, we examine how
interactions affect the general relationship between C and the six variables. This is done by
taking the partial derivative of C with respect to each of the variables to see whether changes in
a second variable alter the sign of the derivative. For example, assume that the derivative of C with
respect to the order cost (P) is positive. The goal of the analysis in the second stage is to determine
whether the sign changes from positive to negative when the values for the other five input variables
are changed.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

125

Values for Input Variables


The literature review identified a reasonable medium value and/or range for each variable
included in the cost comparison. The low and high values for each variable were determined by systematically increasing and decreasing the medium value by 50%. The variables and the values
selected are given in Table 1.
TABLE 1
VALUES FOR INPUT VARIABLES
Variable

Low

Medium

High

.175

.35

.525

60

120

180

125

250

375

37.5

75

112.5

1.5

4.5

.125

.25

.375

The medium value of R corresponds to a 35% discount to sell a product that has lost value due
to changes in product technology or in consumer preference. A survey of websites selling discontinued computer parts reveals that the price differential with regularly available items is between 15
and 68%, which is a reasonable match with the selected range for R. Ranges for the remaining
variables are available in published research. The cost of an order (P) varies from 40 to 175 dollars
(Fazel 1997; Fazel, Fischer, and Gilbert 1998; Lambert and Bennion 1986); daily demand (d) from
50 to 500 units (Evers 1997, 1999; Schwarz and Weng 1999; Zinn, Marmorstein, and Charnes
1992); and product value (V) from 4 to 150 dollars (Johnson and Anderson 2000; Ramasesh 1990).
The time between deliveries (t) varies from 6 hours to 21 days, a range with a higher medium value
than the one adopted in this research (Closs et al. 1998; Evers 1997, 1999; Schwarz and Weng
1999; Tyagi and Das 1997; Zinn, Marmorstein and Charnes 1992). However, given that the nature
of this research is to compare the EOQ with QR (which assumes a relatively short time between deliveries), we opted for a shorter range for t. Finally, the inventory holding cost (H) averages about 25%
per year (Fazel 1997; Johnson and Anderson 2000; Ramasesh 1990). To be consistent with our
practice of determining the range of an input variable by adding plus or minus 50% to the medium
value, we selected the range displayed in Table 1.

126

ZINN AND CHARNES

RESULTS
Recall that the data analysis is done in two stages. In the first stage, we determine the general
relationship between each of the six variables considered here and the cost differential C. This is
done by plotting C versus each of the six variables to see whether the general relationship for each
of the six variables with respect to C is positive or negative. A second goal is to break down the
general relationship into the five cost terms defined above: OCQR, HCQR, OCEOQ, HCEOQ, and
ELV. This is followed by the second stage, where we examine how interactions affect the general
relationships between C and the six variables.
The General Relationships Among C, Input Variables and Cost Terms
The relationships among the cost differential (C), the five cost terms and each of the input variables are illustrated in Figures 1 through 6. Each figure has cost on the vertical axis and one of the
input variables on the horizontal axis. For consistency, the value specified for each of the other
five input variables is the lowest in the range shown in Table 1. For example, Figure 1 shows that
the general relationship between C and the variable risk (R) is negative. The values used to prepare
Figure 1 are the medium value for variable R and the low value for the other five input variables.
Note that OCEOQ and HCEOQ appear as the same line in all six figures because, by definition, ordering and holding costs are equal when the EOQ quantity is ordered.
Risk
The relationship between R and C is negative. The higher the risk, the less likely it is that a
firm will adopt the EOQ, because the level of inventory held under the QR method is lower than in
the EOQ method. Firms manage risk by keeping inventory as low as possible. In the equation for
C, R impacts only the Expected Loss of Value (ELV) term, which has a negative sign. Figure 1 illustrates that ELV is the only cost term that varies with risk.
Similarly, for low risk products managers are usually better off ordering EOQ quantities
because it minimizes the joint cost of ordering and holding inventory, while the QR minimizes the
holding cost only. At the extreme, when risk is zero, the EOQ is always the preferred method.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

127

FIGURE 1
RELATIONSHIPS BETWEEN RISK, COST DIFFERENTIAL, AND COST TERMS

20000
OCQR
HCQR

15000
10000

OCEOQ
HCEOQ

Cost 5000
0

ELV
C

-5000
-10000
0.175

0.350

0.525

Risk (R )

Cost of an Order
The effect of P on C is generally positive. There is an incentive to place fewer and larger orders
as the cost of an order increases. This favors the adoption of the EOQ because it is the only method
that considers the cost of an order as a variable in determining order size. In the equation for C,
P impacts most cost terms but, as Figure 2 shows, its greatest impact is on the ordering cost under
the QR method (OCQR).
The impact of OCQR on the choice between QR and the EOQ should not be underestimated.
There will be serious consequences if a firm switches from the EOQ to QR without first managing
to bring down the cost of an order (P). Adopting the QR when P is high will result in a very high
cost structure, because the QR requires more orders per year, which causes the annual cost of
ordering to grow accordingly.

128

ZINN AND CHARNES

FIGURE 2
RELATIONSHIPS BETWEEN THE COST OF AN ORDER,
COST DIFFERENTIAL, AND COST TERMS

50000
40000

OCQR
HCQR

30000
Cost

OCEOQ

20000

HCEOQ
ELV

10000

0
60.0

120.0

180.0

Cost of an Order (P )

Daily Demand
The general relationship between d and C is negative. As illustrated in Figure 3, when d is small,
C tends to be large, which favors the selection of the EOQ because the OCQR is relatively large
and also constant with respect to d. As d increases, C decreases and there is a greater tendency to
select the QR. This is because HCQR, OCEOQ, and HCEOQ grow as d grows. Therefore, QR is more
likely to be the best alternative for items with high daily demand, while the EO is likely to be the
best alternative for items with low daily demand.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

129

FIGURE 3
RELATIONSHIPS BETWEEN DAILY DEMAND, COST DIFFERENTIAL,
AND COST TERMS

16000
12000
Cost

OCQR
HCQR

8000

OCEOQ

4000

HCEOQ
ELV

0
-4000
125.0

250.0

375.0

Daily Demand (d )

Product Value
The general relationship between V and C is negative. As shown in Figure 4, OCQR remains
relatively large and is independent of V. Therefore, when V is small, the EOQ tends to be the preferred inventory replenishment method. As V increases in value, HCQR, OCEOQ, and HCEOQ increase
in value as well. The increase in OCEOQ and HCEOQ offsets OCQR and the growth in HCQR, resulting in a decline in C. Thus, QR becomes the lowest cost inventory replenishment method as V
increases.

130

ZINN AND CHARNES

FIGURE 4
RELATIONSHIPS BETWEEN PRODUCT VALUE,
COST DIFFERENTIAL AND COST TERMS

16000
OCQR
HCQR

12000
8000

OCEOQ

Cost

HCEOQ

4000

ELV

-4000
37.5

75.0

112.5

Product Value (V )

Time Between Deliveries


Within the range of values considered in this research, the general relationship between the time
between deliveries and C is negative and non-linear. The slope of C becomes less negative as t
increases. This implies that as t increases firms are more likely to adopt QR. This implication may
appear counterintuitive, since QR is usually associated with shorter time between deliveries. In
fact, results show that adopting QR with a short time between deliveries usually results in a very high
ordering cost because the short time between deliveries means more frequent deliveries. These
relationships are illustrated in Figure 5.
If we were to extend the range of t beyond the values considered in the research, the slope of
C would become positive. This occurs because OCQR is the dominant cost factor. The cost of
ordering in the QR method is high whenever t is short. As t increases, OCQR drives C down. At
the extreme, when t is sufficiently large, OCQR ceases to be the dominant cost factor and is mitigated
by a combination of a growing HCQR and decreasing ELV, which eventually drives C back up.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

131

FIGURE 5
RELATIONSHIPS BETWEEN TIME BETWEEN DELIVERIES,
COST DIFFERENTIAL AND COST TERMS

16000
OCQR
HCQR

12000
8000
Cost

OCEOQ
HCEOQ

4000

ELV
C

0
-4000
1.5

3.0
4.5
Time Between Deliveries (t )

Inventory Holding Cost


Figure 6 shows that the inventory holding cost (H) has a minimal impact on the cost differential (C). This is because there are 2 major effects of H on C and they cancel each other out. On
the one hand, as H increases, OCEOQ also increases because QEOQ is smaller. Recall that H is in the
denominator of the equation to compute the QEOQ quantity. In addition, HCEOQ increases because
the holding cost increases with the inventory holding cost (H). On the other hand, as H increases,
the ELV decreases because QEOQ carries a negative sign in the computation of the ELV. Thus, as shown
in Figure 6, these effects cancel out, rendering H as an unimportant factor to consider when deciding between QR and the EOQ. This is a somewhat counterintuitive conclusion because the inventory holding cost is often used as a justification to switch from the EOQ to QR.

132

ZINN AND CHARNES

FIGURE 6
RELATIONSHIPS BETWEEN INVENTORY HOLDING COST,
COST DIFFERENTIAL AND COST TERMS
16000
OCQR
HCQR

12000
Cost

OCEOQ
HCEOQ

8000

ELV

4000
0
0.125

0.250

0.375

Inventory Holding Cost (C )

The Effect of Interactions on the General Relationships between DC and Input Variables
In the second stage of the analysis we examine how interactions affect the general relationship
between C and each of the six input variables. This is done by taking the partial derivative of C
with respect to each of the variables, in turn. The algebraic signs of the partial derivatives correspond
to the general relationships described above for the ranges of values specified in Table 1. Note that
the partial derivative of C with respect to P is positive throughout its range, while the sign of the
partial derivative with respect to H varies from slightly positive to slightly negative. The partial derivatives with respect to each of the four remaining variables are negative. Recall also that in this second-stage analysis, the partial derivative of C is evaluated at the medium value of the variable with
respect to which the partial derivative is taken, and at the low values of the other five input variables.
We investigate interactions through a sensitivity analysis in which we increase the values of the variables held constant in the partial derivative, and then observe whether there are changes in the
signs of the partial. If a change in sign occurs, we conclude that there is an interaction between input
variables, and proceed to explain it.
Two of the input variables, H and R, are not included in the interaction analysis. The reason for
excluding H is that the general relationship is essentially flat and therefore there is no sign to
change. The reason for excluding R is that the derivative of C with respect to R is independent of
R. The partial derivative of C with respect to R is shown below, after Equation 4 is first reproduced
for convenience and then manipulated to include the expressions for QQR and QEOQ.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

C =

133

360 Pd QQR
360Pd QEOQ
+

VH (QEOQ QQR )VR


VH
2
QEOQ
2
QQR

360 P 1
720VPd
+ VHtd 720VHPd
R + VRtd
t
2
H
The partial derivative of C with respect to R is clearly independent of R:

(4)

C
720VPd
= V Q EOQ QQR = Vtd
H
R

(5)

Cost of an Order
The partial derivative of C with respect to P is presented in Equation 6 below. This is followed
by Table 2, which shows values of the derivative C with respect to P for various levels of the five
other input variables. For example, the number 182 at the top right corner of the table is the value
of the partial derivative of C with respect to P when the input value for P is high and the values for
the other input variables (t, R, d, V, and H) are low. Recall that input values are taken from Table 1.

C 360
180VHd
180Vd
=

R
P
t
P
PH

(6)

TABLE 2
VALUES OF THE DERIVATIVE OF C WITH RESPECT TO P FOR
VARIOUS LEVELS OF INPUT VARIABLES
P
low

med

high

short
med
long

139
19
21

169
49
9

182
62
22

low
med
high

139
81
22

169
127
86

182
148
114

low
med
high

139
98
66

169
139
117

182
158
139

low
med
high

139
98
66

169
139
117

182
158
139

low
med
high

139
139
133

169
169
165

182
182
179

134

ZINN AND CHARNES

As expected, the majority of the values in Table 2 carry a positive sign, since the general relationship between C and P is positive. The only exception is marked in bold italic font. The sign
of the relationship is negative when P is low and t is long. In this case, QR is the lower cost method
because OCQR is kept low by the combined high value of t and low value of P. Note, in Figure 2,
that OCQR is the dominant cost in the relationship between C and P.
Daily Demand
The partial derivative of C with respect to d is in Equation 7 below. Table 3 shows values of
the derivative C with respect to d for various levels of the five other input variables.
The values are interpreted in the same way as Table 2, with the difference that we expect most
values to carry a negative sign.
C VHt
180VHP
180 PV
=

R
+ VRt
(7)
d
2
d
Hd

TABLE 3
VALUES OF THE DERIVATIVE OF C WITH RESPECT TO D FOR
VARIOUS LEVELS OF INPUT VARIABLES
d
low

med

high

short
med
long

35
22
8

21
7
6

15
1
12

low
med
high

35
53
72

21
31
41

15
21
27

low
med
high

35
55
70

21
35
46

15
26
35

low
med
high

35
42
44

21
22
19

15
13
8

low
med
high

35
32
31

21
17
16

15
11
9

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

135

The only two positive values in Table 3 (marked bold italic) are for medium and high d when
t is long. They show an interaction between d and t whereby the firm is better off adopting the
EOQ as demand increases and the time between deliveries is long. This is explained mostly by
changes in the Expected Loss of Value (ELV) in Equation 4. When t is long and d increases, the ELV
will decrease because QEOQ increases slightly and QQR increases substantially. The reduction in ELV
makes the EOQ more attractive, changing the sign of the slope of C for medium and high values
of d.
Product Value
The derivative of C with respect to V is in Equation 8 below. Similar to the analysis above
for the other input variables, Table 4 shows values of the partial derivative C with respect to V for
various levels of the five other input variables. The values are interpreted in the same way as Table
3. As a recap, 48, the number on the top right corner of the table is the value of the partial derivative of C with respect to V when V has the high value indicated in Table 1 and the other input variables (P, t, R, d, and H) have the low values indicated in Table 1.

C Htd
180 HPd
180 Pd
=

R
+ Rtd
V
2
V
VH

(8)

Values in Table 4 are mostly negative, which is consistent with the results displayed in
Figure 4. The two exceptions correspond to interactions. The slope of C with respect to V turns positive for the cases when t is long and V is medium or high. This is explained mostly by changes in
the Expected Loss of Value (ELV) in Equation 4. When t is long and V increases, the ELV will
decrease because QEOQ also increases slightly. The reduction in ELV makes the EOQ more attractive, changing the slope of C.

136

ZINN AND CHARNES

TABLE 4
VALUES OF THE DERIVATIVE OF C WITH RESPECT TO V FOR
VARIOUS LEVELS OF INPUT VARIABLES
V
low

med

high

low
med
high

116
183
234

69
116
153

48
87
116

short
med
long

116
72
27

69
25
20

48
4
41

low
med
high

116
178
239

69
103
137

48
70
91

low
med
high

116
139
145

69
72
64

48
42
27

low
med
high

116
105
102

69
58
53

48
37
30

Time Between Deliveries


The partial derivative of C with respect to t is in Equation 9 below. Note, as illustrated in
Figure 5, that the input variable t has a non-linear effect on C that is different from the previous
input variables examined in this research. As explained earlier, the slope of C with respect to t
carries a negative sign when t is short and then bottoms out. For longer values of t that are beyond
the scope of this research, the slope of C turns positive. Table 5 below is similar to the equivalent
tables for other input variables included in the research. It shows values of the partial derivative of
C with respect to t for various levels of the five other input variables.
C
VHd 360 P
(9)
= VRd +
2
t
2
t
The time between deliveries (t) has interactions with every other input variable, always when
t is long or medium. Note, in Figure 5 that C attains a minimum at t = 4.5 days and that it crosses
zero (i.e., the point of indifference between the EOQ and QR) at t = 2.3 days. Therefore, the higher
the value of variables V, H, d, and R, the shorter the time between deliveries at which it is
advantageous to adopt the EOQ. Conversely, the larger the value of P, the longer the time between
deliveries must be before it becomes preferable to adopt the EOQ.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

137

TABLE 5
VALUES OF THE DERIVATIVE OF C WITH RESPECT TO T FOR
VARIOUS LEVELS OF INPUT VARIABLES
t
short

med

long

low
med
high

8487
18087
27687

1287
3687
6087

47
1020
2087

low
med
high

8487
7373
6260

1287
173
940

47
1160
2273

low
med
high

8487
7666
6846

1287
466
354

47
867
1687

low
med
high

8487
7373
6260

1287
173
940

47
1160
2273

low
med
high

8487
8194
7901

1287
994
701

47
340
633

CONCLUSIONS AND MANAGERIAL IMPLICATIONS


The goal of this paper is to compare the EOQ and the QR methods to replenish inventory, and
then propose a set of rules to help managers select the more appropriate method to use under various circumstances. Therefore, we began by writing cost equations for both the EOQ and QR and then
found the difference, denoted by C. The cost difference was analyzed both as a function of the six
individual input variables composing the cost equations and also as a function of five cost terms
obtained in the equation for C (Equation 4). The analysis focused on determining the general
relationship between C and each input variable. The general relationships were then explained with
respect to each of the five cost terms. This was followed by an analysis of the effect of interactions
in the relationship between C and each of the input variables.
The two main conclusions in this research are (1) the importance of reducing the cost of an order
(P) whenever managers are considering switching from the EOQ to QR, and (2) the fundamental
role of risk as justification for adopting the QR method. These two conclusions are supported by
observing in Figures 1 to 6 that OCQR and ELV are almost always the largest cost factors. Input
variable P is a key determinant of the cost term OCQR. Similarly the risk (R) is a key determinant
of the cost term ELV.

138

ZINN AND CHARNES

These two conclusions are further justified conceptually. The key difference between the EOQ
and the QR methods is that the former minimizes the joint cost of ordering and holding inventory
and the latter minimizes only the cost of holding inventory. Thus, the ordering cost in the QR
method is managed independently of the inventory holding cost, which may have important consequences because the ordering cost (OCQR) is almost always the highest cost in Figures 1 to 6. Similar reasoning applies to risk, which is a key factor in determining that firms hold the right inventory
as opposed to merely the right level of inventory, but has so far not been explicitly included by others in models to determine order quantity.
The EOQ method is clearly the lowest cost method if the time between deliveries is short and
the order cost is high. This result is counterintuitive because QR is usually associated with a short
time between deliveries. However, a short time between deliveries also increases order frequency
and therefore the cost of ordering. Thus, if QR is implemented in an environment where the cost of
an order P is high, the increase in ordering cost may negate any gains from the reduction in inventory. The cost of an order must be managed down before implementing the QR inventory replenishment method, or else using the QR method will be counterproductive.
The greater the risk related to holding inventory, the greater the advantage of adopting the
QR, mostly because the risk represents a significant addition to the cost of holding inventory in the
EOQ method. At the extreme, the EOQ method is always less costly when the risk is zero.
We reach the following additional conclusions:
1. The greater the cost of an order, the greater the advantage of adopting the EOQ. This
is because the ordering cost increases proportionately when QR is adopted, as QR
requires more orders per year.
2. The greater the daily demand for a product, the greater the advantage of adopting QR.
This is explained by the cost of ordering under the EOQ. An increase in demand
increases the ordering cost in EOQ, but is neutral with respect to the ordering cost
in QR.
3. The greater the unit value of a product, the greater the advantage of adopting QR. This
is partially explained by the impact of product value on the expected loss of value
(ELV). The cost of the EOQ method grows as the ELV grows.
4. The time between deliveries, especially when short, has a significant upward impact on
the ordering cost in QR. Therefore, the shorter the time between deliveries, the greater
the advantage of adopting the EOQ, although this result is also strongly affected by the
cost of an order, as explained earlier.
5. The cost of holding inventory has little relevance to the choice between the EOQ and
the QR. Its impact on the cost difference between QR and the EOQ is negligible within
the variable ranges considered in this research.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

139

6. The interactions between variables were investigated by showing how changes in the
value of one variable affect the partial derivative with respect to another variable.
Changes in the sign of partial derivatives were observed always in cases where the
time between deliveries is medium or long, especially the latter. Results are summarized
in Table 6 below.
TABLE 6
SUMMARY OF RESULTS
QR is Preferred when
Time between deliveries is moderate
or long and cost of an order is low
Cost of an order is low
Risk is medium or high
Product value is high

Daily demand is high

EOQ is Preferred when


Time between deliveries is short and the cost of an order
is high
Cost of an order is medium or high
Cost of an order is low and time between deliveries is long
Risk is low
Risk is high and time between deliveries is long
Product value is low
Product value is medium or high and time between
deliveries is long
Daily demand is low
Daily demand is high and time between deliveries is
medium or long

Table 6 shows that the EOQ should still be considered by managers, especially for the least important products, i.e., that have low risk, low value and low demand. It is crucial, however, that managers keep in mind that these conclusions are generalizations based on the values selected for this
research. Managers can determine the best approach to determine order quantity by substituting their
firms appropriate values into Equation 4. Finally, this research has shown that despite the shift
observed from the EOQ to QR inventory replenishment methods (explained by the simultaneous
decline in the cost of an order and the increase in the risk of holding inventory), use of the EOQ method
remains a viable option for managers.

140

ZINN AND CHARNES

NOTES
Closs, David J., Anthony S. Roath, Thomas J. Goldsby, James A. Eckert, and Steven M. Swartz (1998),
An Empirical Comparison of Anticipatory and Response-Based Supply Chain Strategies, International Journal of Logistics Management, Vol. 9, No. 2, pp. 21-34.
Evers, Philip T. (1999), Filling Customer Orders from Multiple Locations: A Comparison of Pooling Methods, Journal of Business Logistics, Vol. 20, No. 1, pp. 121-140.
Evers, Philip T. (1997), Hidden Benefits of Emergency Transshipments, Journal of Business
Logistics, Vol. 18, No. 2, pp. 55-76.
Fazel, Farzaneh (1997), A Comparative Analysis of Inventory Costs of JIT and EOQ Purchasing,
International Journal of Physical Distribution and Logistics Management, Vol. 27, No. 8,
pp. 496-504.
Fazel, Farzaneh, Klaus Fischer, and Erika Gilbert (1998), JIT Purchasing vs. EOQ with a Price
Discount: An Analytical Comparison of Inventory Costs, International Journal of Production
Economics, Vol. 54, No. 1, pp. 101-109.
Harris, Ford W. (1915), Operations and Cost, Chicago, IL: A.W. Shaw.
Johnson, Gene and James Stice (1993), Not Quite Just-In-Time Inventories, The National Public Accountant, Vol. 37, No. 3, pp. 26-29.
Johnson, M. Eric and Emily Anderson (2000), Postponement Strategies for Channel Derivatives,
International Journal of Logistics Management, Vol. 11, No. 1, pp. 19-36.
Jones, Daniel J. (1991), JIT and the EOQ Model: Odd Couple no More! Management Accounting, Vol. 72, No. 8, pp. 54-57.
Lambert, Douglas and Mark Bennion (1986), Establishing A Minimum Order Policy, Journal of
Business Logistics, Vol. 7, No. 2, pp. 91-109.
Ramasesh, Ranga (1990), Recasting the Traditional Inventory Model to Implement Just-In-Time
Purchasing, Production and Inventory Management Journal, Vol. 31, No. 1, pp. 71-75.
Schneiderjans, Marc and Qing Cao (2001), An Alternative Analysis of Inventory Costs of JIT and
EOQ Purchasing, International Journal of Physical Distribution and Logistics Management, Vol.
31, No. 2, pp. 109-123.

JOURNAL OF BUSINESS LOGISTICS, Vol. 26, No. 2, 2005

141

Schneiderjans, Marc and Qing Cao (2000), A Note on JIT Purchasing vs. EOQ with a Price Discount: An Expansion of Inventory Costs, International Journal of Production Economics, Vol. 65,
No. 3, pp. 289-294.
Schwarz, Leroy B. and Z. Kevin Weng (1999), The Design of a JIT Supply Chain: The Effect of
Leadtime Uncertainty on Safety Stock, Journal of Business Logistics, Vol. 20, No. 1, pp. 141-164.
Stock, James R. and Douglas M. Lambert (2001), Strategic Logistics Management, 4th Ed. New York:
McGraw-Hill Irwin.
Tyagi, Rajesh and Chandrasekhar Das (1997), A Methodology for Cost versus Service Trade-offs
in Wholesale Location-Distribution Using Mathematical Programming and Analytic Hierarchy
Process, Journal of Business Logistics, Vol. 18. No. 2, pp. 77-100.
Woolsey, Gene (1988), A Requiem for the EOQ: An Editorial, Production and Inventory Management Journal, Vol. 26, No. 3, pp. 68-72.
Zinn, Walter, Howard Marmorstein, and John Charnes (1992), The Effect of Autocorrelated
Demand on Customer Service, Journal of Business Logistics, Vol.13, No. 1, pp. 173-192.

ABOUT THE AUTHORS


Walter Zinn, Ph.D. Michigan State University, is Associate Professor of Logistics and Marketing at The Ohio State University. He is a former Systems Section Editor of the Journal of Business Logistics. Dr. Zinns research has been published in academic journals such as the Journal of
Business Logistics, European Journal of Operational Research, Journal of the Operational Research
Society, and The International Journal of Logistics Management. Professor Zinn is a member of the
Council of Supply Chain Management Professionals.
John M. Charnes, Ph.D. University of Minnesota, is Professor in the Finance, Economics, and
Decision Sciences Area of the School of Business at The University of Kansas, where he does
research and teaches courses in risk analysis, systems simulation, and applied statistics. His website is www.ku.edu/home/jcharnes.

You might also like