CH APTE R 14 14 THE LOGISTICS PLANNING PROCESS 3
The MILES module within the LOGWARE software is used to solve this problem. It computes distance based on the great circle distance formula using longitude and latitude. (a) The estimated road distance is 1,380 miles. (b) The estimated road distance is 830 miles. (c) Since both latitudes are in the same hemisphere, no adjustments need to be made. The estimated distance is 244 miles, or 244×1.61 = 393 km. (d) In this case, one point is east and the other west of the Greenwich line. Therefore, we need to set a sign convention. Let's set west longitudes as + and east longitudes as −. Thus, 2.20o E longitude is entered into MILES as −2.20 o . The estimated distance is 250 miles, or 250×1.61 = 402.5 km. 4
Suppose that a certain linear grid coordinate system has been overlaid on a map of the United States. The grid numbers are calibrated in miles, miles, and there is a road road circuity factor of 1.21. Find the expected road distances between the following following pairs of points: Equation 14-1 in the text is used to approximate distances from linear coordinates. The K factor factor in the equation is set at 1.21. (a) Lansing, MI to Lubbock, TX L o c a t i on a. From To b. From To c. From To d. From To
D
= 1.21
X C oor di nat e Lansing, MI Lubbock, TX El Paso, TX Atlanta, GA Boston, MA Los Angeles, CA Seattle, WA Portland, OR
( 924.3 − 1, 488.6)2
Y Co o r din a t e 924.3 1488.6 1696.3 624.9 374.7 2365.4 2668.8 2674.2
+
(1, 675. 2 − 2, 579. 4) 2
+
( 2, 769.3 − 2, 318. 7)2
=
1 67 5 . 2 2 579.4 2 7 6 9. 3 2 318.7 1 3 2 6. 6 2 76 3 . 9 1 900.8 2 0 3 9. 7
1, 290 miles
(b) El Paso, TX to Atlanta, GA D = 1.21 (1, 696.3 − 624.9)2
(c) Boston, MA to Los Angeles, CA
202
=
1, 406 miles
763.9)2 D = 1.21 ( 374.7 − 2,365.4)2 (1, 326. 6 − 2, 76
=
2, 971 miles
(d) Seattle, WA to Portland, OR D = 1.21 ( 2,668.8 − 2, 674.2)2
+
(1, 900.8 − 2, 039. 7)2
=
168 miles
5
The plot of the truck class rates is shown in Figure Figure 14-1. The rates show a high degree of linearity. A linear regression was found with aid of the MULREG module in LOGWARE. The rate equation was determined to be: R = 5.1745 + 0.0041× D
The standard error of the estimate S E is 0.9766. The coefficient of determination r 2 is 0.928. The best single estimate of the rate at 500 miles is: R = 5.1745 + 0.0041×500 = $7.23/cwt.
Assuming the error around the regression line is normally distributed, a 95 percent confidence band would give a range for the actual rate. That is, Y = R ± 1.96×S E = 7.23 ± 1.914
where 1.96 is the normal deviate for the normal distribution representing 95 percent of the area in a two-tailed two-tailed distribution. The range of the estimate is: is: $5.32/cwt. ≤ Y ≤ $9.14/cwt. The r 2 value of 0.928 indicates that a linear rate equation explains about 93 percent of the variation in the data with distance. Such a simple relationship seems to represent the rates quite well.
203
FIGURE 14-1 Plot of Truck Class Rates 20 18 16 . 1 4 t w c 1 2 / $ , e 1 0 t a r 8 s s 6 a l C 4
Estimating line
2 0 0
500
1000
1500
2000
2500
3000
3500
Distance, miles 6
A plot of the average inventory level versus warehouse throughput is shown in Figure 142. The multiple regression software software in LOGWARE was used to test two equation forms. The first was of the form I
=
aTP b
and the other was of the form I
=
a + bTP
Both forms showed high r 2 values, with the exponential form being slightly higher at 0.9406. It was selected as the equation form to use. This equation was: I
=
0.83 0704 . × TP
where TP and and I are are both expressed in thousands of dollars. We can now estimate that for an annual warehouse throughput of $50,000,000, the average inventory would be: 704 × 50, 0000.83 I = 0.70 =
5,59393 593939 . 9, or $5,593,939 $5,593,939
This type of relationship is very useful in network planning, especially warehouse location, to estimate how inventory levels will change when sales are reallocated to a varying number of warehouses.
204
FIGURE 14-2 Plot of Inventory Levels and Warehouse Throughput for California Fruit Growers’ Association )
s 12 n oi ll i
M( $ e v el yr ot n e v ni e
10
l, 8
6 4
g 2 ar e v 0 A
Estimating line
0
20
40
60
Annual warehouse thruput, $(Millions)
205
80
100
USEMORE SOAP COMPANY Teaching Note
The purpose of this case study is to provide students with the opportunity to evaluate and design a large-scale production-distribution network using real data and cost relationships. To assist in the substantial amount of computational effort in this this problem, an interactive computer program (WARELOCA) is available in the LOGWARE collection of software modules. Major Issues The text of the case suggests a number of questions that are critical to productiondistribution network network design. These reduce to three major major issues, namely:
(1) Should plant capacity be added and, if so, when and where? (2) How many warehouses are optimal and where should they be located? (3) Should the current customer service level be retained? Although no change can be made in the network without potentially affecting other variables, the attempt here will be to treat these questions sequentially to converge on a good network design. Numerous computer runs were made to provide the basic information needed ne eded in the analysis. The more meaningful runs are summarized in Appendix A to this note. Tables 1 and 2 compare selected runs for both the current-year and the future-year time periods. This information is used throughout the analysis of the major issues. The Plant Expansion Issue An attempt to meet 5-year growth goals using current plant capacity will cause the system having a total capacity of 1,630,000 cwt. to be out of capacity in 1.7 years. That is,
5th-year demand Current demand Net increase
1,908,606 cwt. −1,477,026 431,580 cwt.
Therefore, the average annual growth rate rate is 431,580/5 = 86,316 cwt. So, in (1,630,000 − 1,477,026)/86,316 = 1.7 years all available capacity will be depleted. If no expansion of plant capacity occurs, then 1,908,606 − 1,630,000 = 278,606 cwt. will potentially be lost by the 5th year. Sales are $100 million on 1.477 million cwt. in volume for a product product value of $67.7/cwt. With a profit margin of 20 percent, the profit per cwt. would be 20%×$67.7/cwt, or $20/1.477, = $13. Thus, 278,606×13 = $3.16 million are lost in sales. The weighted profit loss over the five-year period would be: 2/5× (0) + ( 3/5)× ((0 + 3.6))/2) = $1.08m/yr.
206
TABLE 1 Current-Year Comparison of Network Alternatives ($000s)
C o s t t y pe
B e n c hm a rk
Production $30,762 Warehouse operations 1,578 Order processing 369 Inventory carrying 457 Transportation Inbound 2,050 Outbound 6,896 Total costs $42,112
I m p r o v ed b e n c hm a rk
O p t i m um n u m b er of whses
O p t i m um n u m b er of o f whses
R e l a x ed se s ervice (1)
R e l a x ed se s ervice (2)
Maxim um op p o r tu t unit y
$30,678 1,468 354 431
$3 $ 30,673 1,608 370 508
$3 $ 30,675 1,572 358 490
$3 $ 30,678 1,296 349 390
$30,673 1,420 354 445
$30 ,3 86 1 ,5 29 3 58 5 00
1,802 6,991 $41,725
1,976 6,310 $41,447
1,860 6,365 $41,321
1,249 7,238 $41,201
1,178 6,698 $41,043
1, 1 78 6, 4 58 $4 0, 4 09
Customer service: ≤
300 mi.
93 %
93 %
98%
92%
75%
88%
81 %
≤
600 mi.
98 %
98 %
100%
100%
98%
100%
9 4%
No. of stock stockin ing g poi n t s
22
21
31
30
19
26
40
N o . o f pl an t s
4
4
4
4
4
4
6
Savings Savings vs. benchmark
$0
$387
$665
$ 79 1
$9 1 1
$1,069
$1 ,7 0 3
$0
$278
$ 40 4
$5 2 4
$
$ 1, 31 6
Service to match benchma r k
600 mi constraint on current warehouses
600 mi constraint on opt no. of warehouses
Savings Savings vs. improve improved d benchmark $0 Comments:
682
U nli m it e d se rv i ce , wh ses , an d plant cap.
207
TABLE 2 Future-Year Comparison of Alternatives ($000s)
C o s t t y pe
Production Warehouse operations Order processing Inventory carrying Transportation Inbound Outbound Total costs
No p l a nt expa nsio n
A dd p l a nt @ M e m p h is
Add plant @ M e m p h is & C h i c a go
M e m p h is a n d o pt n o . of w h s e s
Mem p his a nd o p t n o . of w h s e s
$33,965 1,496 393 431
$39,517 1,842 462 505
$ 3 9,5 4 8 1, 84 7 45 4 49 7
$39,524 2,028 470 591
$ 39 , 52 2 1 , 97 6 4 60 57 3
1,647 7,230 $45,164
2,350 9,030 $53,705
2,0 0 0 9 , 03 6 $ 5 3, 38 2
2,614 8,117 $53,342
2 , 42 6 8 , 22 2 $ 53 , 17 9
Customer service: ≤
300 mi
98%
94 %
95%
98%
92%
≤
600 mi
99%
98 %
98%
100%
1 00 %
No. of stock stockin ing g poi n t s
20
21
20
31
30
N o . o f pl an t s
4
5
6
5
5
Comments:
Not all demand met
High service level
S er v ic e he l d a t be n ch ma r k
TABLE 2 Future-Year Comparison of Alternatives ($000s)
C o s t t y pe
Production Warehouse operations Order processing Inventory carrying Transportation Inbound Outbound Total costs
No p l a nt expa nsio n
A dd p l a nt @ M e m p h is
Add plant @ M e m p h is & C h i c a go
M e m p h is a n d o pt n o . of w h s e s
Mem p his a nd o p t n o . of w h s e s
$33,965 1,496 393 431
$39,517 1,842 462 505
$ 3 9,5 4 8 1, 84 7 45 4 49 7
$39,524 2,028 470 591
$ 39 , 52 2 1 , 97 6 4 60 57 3
1,647 7,230 $45,164
2,350 9,030 $53,705
2,0 0 0 9 , 03 6 $ 5 3, 38 2
2,614 8,117 $53,342
2 , 42 6 8 , 22 2 $ 53 , 17 9
Customer service: ≤
300 mi
98%
94 %
95%
98%
92%
≤
600 mi
99%
98 %
98%
100%
1 00 %
No. of stock stockin ing g poi n t s
20
21
20
31
30
N o . o f pl an t s
4
5
6
5
5
Comments:
Not all demand met
High service level
S er v ic e he l d a t be n ch ma r k
208
Based on a simple rate of return on investment, capturing this profit potential would yield 1.08/4 = 27 percent annually on a $4,000,000 investment investment for expansion. The return would increase to 90 percent per year with the full loss in the 5th year. year. The potential seems great enough to justify one unit of expansion (1,000,000 cwt.). Two units of expansion probably cannot be justified, since adequate capacity would be available from the first capacity unit to meet demand requirements. requirements. The only benefit would be from from the network design improvement. The savings would be about $323,000 per year in the fifth year (see Table 2) comparing one additional plant with two additional plants and keeping the current number of warehouses. The simple return on investment using fifth-year savings would only amount to about 8 percent (323,000×100/4,000,000 = 8.1%). The next question is: Where should the expansion take place at at an existing plant or at one of the two proposed locations? From a test of expanding any of the four existing plants or the two proposed plant locations (runs 10 through 16 in Appendix A of this note), it would appear that Memphis would be the lowest cost site in the 5th year with Chicago next at only an additional cost of $76,000 per year (compare runs 14 and 15 in Appendix A). Adding a plant at a new location rather than expanding an existing plant
Based on a simple rate of return on investment, capturing this profit potential would yield 1.08/4 = 27 percent annually on a $4,000,000 investment investment for expansion. The return would increase to 90 percent per year with the full loss in the 5th year. year. The potential seems great enough to justify one unit of expansion (1,000,000 cwt.). Two units of expansion probably cannot be justified, since adequate capacity would be available from the first capacity unit to meet demand requirements. requirements. The only benefit would be from from the network design improvement. The savings would be about $323,000 per year in the fifth year (see Table 2) comparing one additional plant with two additional plants and keeping the current number of warehouses. The simple return on investment using fifth-year savings would only amount to about 8 percent (323,000×100/4,000,000 = 8.1%). The next question is: Where should the expansion take place at at an existing plant or at one of the two proposed locations? From a test of expanding any of the four existing plants or the two proposed plant locations (runs 10 through 16 in Appendix A of this note), it would appear that Memphis would be the lowest cost site in the 5th year with Chicago next at only an additional cost of $76,000 per year (compare runs 14 and 15 in Appendix A). Adding a plant at a new location rather than expanding an existing plant site saves a minimum of $281,000 annually (compare runs 11 and 14 in Appendix A of this note), which results from placing plant capacity closer to warehouses. Selecting Warehouses A simple test on the number of warehouses in the network shows that transportation costs are dropping more rapidly than inventory related costs are increasing (see Figure 1). This means that 40 active warehouses will have the lowest total total cost. However, some of these warehouses will have low throughput. throughput. In order to maintain a minimum replenishment frequency and shipment size, size, a minimum throughput needs to be met. Approximately a truckload every two weeks, or 10,400 cwt. of throughput per year, is the minimum activity needed to open a warehouse. Therefore, any warehouse showing less than this throughput will be eliminated from consideration. Under various assumptions about plants and their capacities, demand growth, and service levels, 30 to 31 warehouses seem most economical with no deterioration on service over the benchmark network. The following table shows selected results.
209
T y p e of
P l a nt
Percent of d e m a n d
T o t al
N o. of
≤
co s t
w hs es
r un
Y e ar
c ap aci ti es
Ben chmar k Improved ben chmar k
Current
C urren t
93
$ 4 2, 11 2
22
Current
C urren t
93
4 1 , 72 5
21
5th yr.
C urren t + Memphis
94
53,70 5
31
92
41,32 1
3 30 0
5th year
C u rr e n t C u r re n t + Memphis
92
53,17 9
30
Improved ben chmar k Current yr. whses 5th yr. whses
3 00 m i .
Note that this conclusion about the number of warehouses depends on the previous conclusion that a Memphis plant should should be added by the fifth year. year. The number of warehouses should be increased from the present 22 in both the current year and the fifth year. 42
100 99
41.8
98 Service (right scale)
97
) s 0 0 0 , 41.6 0 0 0 ( $ , t s o c 41.4 l a t o T
96 95 94
Cost (left scale)
93
41.2
. i m 0 0 3 < d n a m e d f o %
92
Practical design
91 41
90 22
26
30
31
36
40
Number of warehouses
FIGURE 1 Cost and Customer Service Profiles for Alternative Network Designs
More detailed economic analysis shows that if the plants are held at current throughput levels, a savings realized from 30 warehouses would be $41,725,000 − 41,321,000 = $404,000 (see previous table). If current plant capacities are used and the Memphis plant is on-stream in year five, the savings of the added warehouse would be: $53,705,000 − 53,179,000 = $526,000
210
On the average, there can be savings of approximately ($404,000 + 526,000)/2 = $465,000 per year by increasing the number of warehouses to 30 from the current 22. Since these are public warehouses, little or no investment would be required to implement the change. Although the number of warehouses remains relatively unchanged from the current year to the 5th year, there is some shifting among the particular warehouses in the mix. The 30 warehouses in the current year should be numbers: 1,2,3,4,5,7,8,11,13,14,15,16,17,18,19,20,21,25,28,31,32,33,34,35,36,37,38,40,44,45 providing that the loading lo ading on the current plants is allowed up to the limits of their current capacity. When the Memphis plant is brought brought on-stream by the end of the second year, the warehouse mix should begin to evolve to numbers: 1,2,3,4,5,7,8,11,13,14,15,17,18,19,20,21,25,28,29,31,32,34,35,36,37,38,40,44,45,47 As the Memphis plant is bought on stream, the Memphis public warehouse is closed and the volume is is shifted to the Memphis plant as a warehouse. In addition, the Richmond, VA warehouse is closed and the Las Vegas, NV warehouse warehouse is opened. The number of warehouses remains at 30. Both in the base year and in the future year, the throughputs in the plants serving as warehouses are within acceptable limits as the following summary shows. P l a nt as a wa r e hou s e
Covin vington New York Arl ingto n Long Beach
T h r u p ut l imi t s
C u r r e n ty e ar s o l u t i on
Fu tu re y e ar so lut io n
450,000 000 cwt. wt. 380,000 140,000 180,000
254,47 ,471 cwt cwt. 302,043 66,592 95,943
306, 06,478 cwt. 38 0,523 6 6,161 11 7,288
Customer Service Currently, a high proportion of demand (93 percent) is located within 300 miles of a stocking point. Since the service distance may be up to 600 miles and still meet the company's service policy, should the service level be reduced somewhat to effect a cost saving? For example, using the improved benchmark as the the base case (run 2), 93 percent of the demand is within within 300 miles and 97.5 percent is within 600 miles. If a 600-mile 600-mile constraint is applied to the current network configuration (run 23), 75 percent of the demand is within 300 miles and 98 percent is still still within 600 miles. The total costs are reduced from $41,725,000 to $41,201,000, or a savings of $524,000 per year. In addition, if the number of warehouses in the network is optimized, the costs can be reduced by another $158,000 per year year (run 23 vs. run 22). However, $278,000 of the total $524,000 + 158,000 = $628,000 can be realized without a service change. This leaves approximately $404,000 that can be saved by a relaxed service restriction. The question now becomes one of whether the higher costs associated with the more restrictive service level level are justified. Since there is no sales-service sales-service relationship for this problem, we can only estimate the worth of the service. That is, can enough sales be
211
generated to cover the higher service level? If physical physical distribution costs for the com pany are 15 percent of sales, which is probably a conservative estimate, then 1/0.15 = $6.70 in sales must be generated for each dollar that is added to distribution costs. Therefore, to cover $404,000 in cost would require $404,000 × $6.70 , cwt. = 38124 $0.71 / lb l b.×100lb./cwt. increase in sales. In terms of overall demand, this would be 38,124×100/1,477,026 = 2.5 percent. But not all customers customers would experience a higher service level. Comparing the demand centers for 299,818 cwt. of demand shows a reduction in warehouse to customer miles. Thus, moving from a minimum cost network to one with with a high service level, where the percent of demand less than 300 miles increases from 75 percent to 93 percent, requires that the 38,124 cwt. increase in demand occur in the 299,818 cwt. of demand affected by the change. This would be a 13 percent increase. increase. The products are not highly differentiated from others in the marketplace so that service plays an important role role in selling selling these products. Whether a 93 − 75 = 18 percentage points increase in service can result in a 2.5 percent increase in overall sales s ales cannot be judged by the distribution department alone. The sales department must play an important part in indicating whether the additional sales are possible. possible. If they are not likely to be realized, there is no incentive for a network other than the minimum cost one. If this information is not available from sales, the conclusion is likely to be to maintain the status status quo as represented by the benchmark. That is, one-day one-day service is most likely to guide the design. Overall Analysis and Summary The recommended design would involve an immediate increase in the number of warehouses from 22 to 30. In addition, there should be an immediate reallocation reallocation of demand among the existing plants. No reduction in the customer service level seems justified at this time. Therefore, a total cost reduction of $42,112,000 − 41,321,000 = $791,000 per year seems immediately achievable (run (run 1 vs. run run 18). By the end of the 2nd year, the Memphis plant should be brought on stream and the network should begin to evolve from the current design design (run 24) to that for the fifth fifth year (run 25). The addition of a plant is justified from the high rate of return realized from the profit potential of being able to continue meeting the growth in demand. For the current year, a breakdown of the service and the cost changes show the following:
212
Cos t typ e Production Whse operations Order processing I n v e n t o r y c a r r y in g Transportation I nboun d Outbound Total costs ($000s)
Benchmark
Currentyear design
Change from be n ch m a r k
$3 0 ,7 6 2 1,57 8 36 9 45 7
$ 3 0 ,6 7 5 1,57 2 35 8 49 0
$ -87 - 6 -11 +33
- 0 . 3% -0 -0.4 - 3 .0 +7.2
2, 0 5 0 6,89 6 $ 4 2, 1 1 2
1,86 0 6 , 36 5 $ 4 1, 3 2 0
-190 -531 $-792
-9.3 -7.7 -1 . 9 %
By the fifth year, total distribution costs should be $53,179,000, or $53,179,000/1,908,606 = $27.86, compared with the current-year cost of 42,112,463/1,477,026 = $28.51 per cwt. If current year year costs are projected to the fifth year demand level, the 5th-year 5t h-year production/distribution costs might be 28.51×1,908,606 = $54,414,357, or a savings of $54,414,357 − 53,179,000 = $1,235,357 per year. Of course, these savings can only be realized through the addition of capacity at Memphis for $4,000,000. If this this capacity is useful for at least 15 years, the the amortization of $4,000,000/15 = $267,000 per year would yield a net savings of $532,000 per year. Overall, the design change appears app ears to be justified.
213
APPENDIX A Listing of Selected Computer Runs R u n . R un
No of
P l a nt
D e m a nd
S e r v i ce c o n-
No of
T o t al
P e r c e n t of d e m a nd w i t h i n
n o. d e s c r i p t i o n
plant s
c a p a c i ty
l e v el
s t r a i nt
w h s es
c osts
≤
1 Benchmark 2 Im I mproved benchmark 3 No No serv constraint 4 Ma M ax opportunity 5 Future yr-imp bmk 6 Test 27 whses 7 Test 32 whses 8 Test 37 whses 9 Test 42 whses 10 E Ex xp Covington 11 Ex Exp New York 12 E Ex xp Arlington 13 Exp Long Beach 14 Add Memphis 15 Add Chicago 16 Add Mem & Chi ea 17 No plant expansion served 18 Optimum whses capacity 19 Op Optimum whses 20 O Op ptimum whses 21 Test cust service 22 Test cust service 23 Te T est cust service 24 Optimum whses 25 Optimum whses
4 4 4 6 4 4 4 4 4 4 4 4 4 5 5 6
Cu rr en t C u rr en t Cu rre n t Cr nt +1 m C u rr en t C u rr en t C u rr en t C u rr en t C u rr en t C u rr en t C u rr en t C u rre n t Cu r r en t Cu r r en t C u rr en t
C u rr en t - mi Current 30 3 00 C ur re n t 9 00 0 Current 9000 5th yr. 300 Current 300 Current 300 Current 300 Current 300 5th yr. 3 30 00 5th yr. 30 300 5th yr. 3 30 00 5th yr. 30 3 00 5th yr. 300 5th yr. 300 5th yr. 300
22 21 18 40 21 26 31 36 40 21 21 21 21 21 20 20
$42,112 41,725 40,896 40,409 53,777 41,744 41,615 41,501 41,486 54,145 53,986 54,709 55,251 53,705 53,781 53,382
4
C u rr en t
5th yr.
30 0
20
4
C u rr en t
C u r ren t
300
4 5 4 4 4 4 5
S e e cm t C u rr en t C u rr en t C u rr en t C u rr en t C u rr en t S e e cm t
C ur re n t 5th yr. C ur re n t C ur re n t C ur re n t C u r ren t 5th yr.
30 300 3 30 00 60 0 60 0 60 6 00 375 375
300
≤
600 600
Comm Commen ents ts
93% 93 93 71 71 81 81 93 95 98 99 99 9 94 4 93 93 9 94 4 94 94 94 94 95
98% 98 89 94 98 1 00 1 00 1 00 1 00 98 98 98 98 98 98 98
Current network design No investment required
Covington cap + 1m cwt New York + 1m cwt Arlington cap + 1m cwt Long Beach cap + 1m cwt Add Memphis at 1m cwt Add Chicago at 1m cwt Add Chi & Mem at 1m cwt
45,164
98
100
Only 85.4% of demnd
31
41,447
98
100
Plants at current
30 31 31 26 19 30 30
41,563 53,342 40,996 41,043 41,201 41,321 53,179
97 97 9 98 8 80 88 75 75 92 92
100 100 100 100 98 100 100
Plants at current thruput Memphis at 1m cwt Whses at opt no = 31 Whses from opt no = 31 Whses from current 22 Service level at bmk Serv at bmk/Mem @ 1m cwt
Added plants at 1m cwt Plant cap + 1m cwt
214
ESSEN USA Teaching Note Strategy Essen USA is concerned with entire supply channel performance. The supply channel consists of four echelons ranging from factory to customers. The purpose of this case study is for the student to manipulate the supply channel variables using a channel simulator in in order to improve individual individual member and system-wide performance. The channel variables include forecasting methods, inventory policies, transportation services, production lot sizes, order processing costs, and stock availability levels. Students should seek to optimize channel performance, although it is not expected that the optimum actually can be found or verified. However, improving performance over existing levels is achievable. The SCSIM module of LOGWARE is used to simulate the demand and product flows throughout the multi-echelon multi-echelon supply chain. SCSIM is an ordinary Monte Carlo day-today type of simulator. simulator. Using a simulator for performance improvement requires thinking thinking of it in terms of as an experimental methodology. That is, a single run of the simulator is
ESSEN USA Teaching Note Strategy Essen USA is concerned with entire supply channel performance. The supply channel consists of four echelons ranging from factory to customers. The purpose of this case study is for the student to manipulate the supply channel variables using a channel simulator in in order to improve individual individual member and system-wide performance. The channel variables include forecasting methods, inventory policies, transportation services, production lot sizes, order processing costs, and stock availability levels. Students should seek to optimize channel performance, although it is not expected that the optimum actually can be found or verified. However, improving performance over existing levels is achievable. The SCSIM module of LOGWARE is used to simulate the demand and product flows throughout the multi-echelon multi-echelon supply chain. SCSIM is an ordinary Monte Carlo day-today type of simulator. simulator. Using a simulator for performance improvement requires thinking thinking of it in terms of as an experimental methodology. That is, a single run of the simulator is a particular event sequence generated from random numbers. Changing the seed number in the simulator causes a different set of random numbers to be generated and possibly another outcome from the same input data. A simulation run with a specified seed number should be viewed as a single statistical observation and multiple outcomes from various seed numbers should be treated as a statistical sample and analyzed accordingly, i.e., comparing means and standard deviations. Each simulation is run for a period of 11 years with results taken from years 2 through 11. The first year is not used since it can show unstable results due to to startup conditions. The results appear to reach steady steady state by the second year, and the results results for the 10 years thereafter are averaged to give a reasonable representation of channel performance for a given run. The database used to represent the current cu rrent performance of the channel, as derived from the case study, is summarized in the Appendix A of this note and a typical run report is shown in Appendix B. This case provides students with the opportunity to observe the operation of a multiechelon supply channel and to assess the impact of changing key operating variables on individual members as well as on channel-wide performance. The effect on cost and customer service as well as sales, inventory, and back order levels of demand patterns, demand forecasting methods, inventory control methods, transportation performance, production lot sizing, order processing procedures, procedu res, and item fill rates can be observed in both graphical and report forms. Most importantly, students can see the effects of supply chain decisions rather than project the results statistically. Questions
1. What can you say about the logistics performance performance throughout the supply channel for Essen and its customers?
215
General observations It is recognized that Essen must deal with demand that has significant seasonal peaks at gift-giving times of the the year as shown in in Figure 1. Compared with a smooth demand pattern, this can cause increasing demand variability upstream from the customers, as illustrated in Figure 2. This “bull whip” effect is partly partly a result of the demand for an upstream member being derived from the order size and pattern of its immediate downstream channel member. Forecast accuracy, lead-time uncertainty, and inventory control method also affect demand variability and the resulting cost of that variability.
Figure 1 Typical Demand Pattern for Essen Over the Period of One Year
Retailer Distri warehouse
Essen Warewarehouse house
Factory Factory
-butor
Retailer Retailer
Figure 2 Increasing Demand Variability of Upstream Channel Members for a Four-Year Period
216
Benchmark Running the simulator (SCSIM) with a seed number of 123456 and simulated period of 11 years with results taken from the last 10 years, the channel generates average annual sales of $109.5 million for a net average annual system profit contribution of $24.4 million, as shown in Table 1. The question arises as to whether channel performance can be improved and profits increased. At least two observations can be made that suggest there is room for improvement. improvement. First, the inventory levels for both the retailer’s warehouse and Essen’s warehouse are quite high compared with the Retailer (see Figure 3). It is possible that Retailer inventories are too low. low. However, the inventory turnover ratio is about seven for the Essen’s warehouse (see Table 1). This is not particularly high for a food product that might have a turnover at least in the range of 10 to 12. The turnover for the retailer’s warehouse appears more in line with industry norms of about 13 (see Table 1).
Retailer warehouse
Essen warehouse
Retailer
Figure 3 Inventory Levels for Four Years Using Benchmark Data
Second, the backorders at the Retailer level do not seem to recover well from the seasonal spike in demand. Correspondingly, the Retailer Retailer inventory turnover is 81 (see Table 1), which is quite high. The low percentage of demand filled filled on request (<50%) suggests that inadequate inventory is being maintained to meet reasonable fill rates. Third, customer service levels are also low for the retailer’s warehouse and Essen’s warehouse. Backorder occurrences are high for both channel members. Although inventory levels are adequate most of the time, seasonal demand rippling through the supply chain causes a significant number of back orders before inventory can be replenished. The observation is that there is an opportunity to improve channel performance, especially in terms of customer service. A major concern is how to mange the seasonal demand pattern that is causing the cyclical behavior throughout the echelons of the
217
channel. Current performance of the the channel members is summarized summarized in Table 1 for four simulation runs using different seed numbers. Table 1 Average Annual Performance of Channel Members and the System at Benchmark Channel member
Run 1
Run 2
Run 3
Run 4
Average
$73,105,904 37,918 $1,928
$72,967,088 37,846 $1,928
$72,616,192 37,664 $1,928
$72,477,376 37,592 $1,928
$72,791,640 37,755 $1,928
6.52 <50% $5,578,291 $147.02
6.59 <50% $5,549,202 $146.50
6.61 <50% $5,521,236 $146.32
6.58 <50% $5,540,447 $146.65
6.58 <50% $5,547,294 $146.62
12.95 <50% $3,873,236 $101.94
13.01 <50% $3,895,406 $102.15
12.96 <50% $3,853,890 $102.14
12.89 <50% $3,912,699 $103.07
12.95 <50% $3,883,808 $102.33
81.02 <50% 38,017 $2,884,527 $76.19
80.38 53.02% 37,983 $2,768,697 $72.89
80.45 54.09% 37,774 $2,939,464 $77.82
80.60 <50% 37,804 $2,981,607 $78.87
80.61 <50% 37,895 $2,893,574 $76.44
$24,426,593 22.23% 123456
$24,591,053 22.40% 444444
$24,235,498 22.20% 555555
$24,340,563 22.28% 666666
$24,398,426 22.28%
Essen’s factory Total cost
Units produced Cost per unit Essen’s warehouse TO ratio Fill rate Cost
Cost per unit Retailer’s warehouse TO ratio Fill rate Cost
Cost per unit Retailer TO ratio Fill rate Units sold Cost
Cost per unit System Profit Profit as % of sales Seed Number
2. What steps steps would you suggest taking to improve logistics performance throughout the channel? Do any of the changes involve Essen? If so, so, does the company directly realize any cost and/or operating performance improvements?
A number of actions can be taken to lower costs and improve customer service. Improving the forecast, shortening the lead times, changing the inventory control policy, and changing production lot sizes are all variables that can be altered for possible performance improvement. The interactions among these variables and the large number of variable variable combinations preclude finding the optimal set. However, they can be explored in a systematic way to find improvement. The primary focus of this analysis will be to increase the fill rates at the the risk of increasing increasing costs. Ultimately, revenues, through improved customer service, may be preserved or increased to more than compensate for reduced profits.
218
Retailer Level Start with the retailer because of the proximity to the customer. Fill rates need to be improved, probably in the 95-99% range as specified in the database. Inventory turns can be guided by the industry average of 12 turns per year. Where the two cannot be jointly met, service will prevail. Clearly, putting additional inventory at the retail point will improve customer service. Using the company’s current inventory policy of stocking to demand, the target level can be raised without changing the review time. Exploring different target levels shows 14 days to offer about 35 turns and a 99+% fill rate. Because of the high cost of a back order, total costs at the retail level drop significantly. Altering the forecasting method and the settings associated with the method yield little opportunity for improvement. Using an exponential smoothing smoothing model with a high smoothing constant to better follow the seasonal changes in demand results in increased costs. Lowering the the smoothing constant to 0.1 did not offer improvement either. Altering the number of periods in the moving average model did not improve costs and only degraded performance. Shortening the review time in the stock-to-demand reorder policy did have a positive effect on fill rate, but resulted in high costs and lower inventory turns. The tradeoff did not seem beneficial, given the fill rate and turnover targets. Retail Warehouse Level Determining an improved policy at this level is difficult because a 95% fill rate and 10 to 12 inventory turns is an illusive goal. Using service as the primary primary target, a stockto-demand control policy is used with a review period of 7 days and a target of 25 days of inventory. The forecasting forecasting method method is is moving average with a period period of 7 days. The performance achieved at this channel level is about nine inventory turns per year and a 97% fill rate. Essen’s Warehouse Level The performance at Essen’s warehouse level seems to mimic that at the retail warehouse level except that there there is more demand variability. Again, an inventory turnover ratio in the target range cannot be achieved while maintaining a high fill rate level. Trying to achieve high service levels with high levels of inventory is difficult, difficult, probably due to the extensive demand variability that filters back to this member of the channel. Multiple simulation simulation runs show that a high fill rate cannot consistently be achieved even when on the average inventory levels are high. However, average performance shows an 82% fill rate and 1.5 inventory turns per year based on a 7-day moving average forecasting model and a stock-to-demand inventory control policy with a review time of 7 days and an inventory target of 25 days. Essen’s Factory The concern with the factory level in the channel is whether product should be manufactured in a larger lot size, but with slightly higher production time variability. The reduced costs seem to out weigh the negative effects of increased variability. Producing in the larger lot size is favored.
219
Overall Using the objective of improving customer service, it is not surprising that supply channel costs increase as shown from the reduced profit in Table 2 compared with Table 2. The average fill rate has increased for all members of the channel, but the cost effects effects are spread disproportionately disproportionately among the members. Even with a higher fill rate, rate, the retailer benefits from a substantial reduction reduction in the cost per unit sold. On the other hand, the cost for handling a unit of the product at Essen’s warehouse is substantially increased. Essen should take advantage of the cost reduction from producing in the larger batch size, but this does not offset the higher cost at the company’s warehouse. As an upstream channel member, Essen undoubtedly suffers from variability in demand, which it cannot entirely control. The retailer benefits from the action to increase fill rates across the channel. However, Essen is put at a disadvantage and may take a counter action to improve its cost position. Essen may simply lower its inventory level by reducing the reorder target quantity from 25 to 10 days. This reduces Essen’s per-unit warehouse cost, but it also increases the costs for for the retailer. The reduced inventory level at the Essen Essen warehouse causes lower fill fill rates for for the downstream retailer. Unless the retailer can find an incentive to reward Essen for its good service, it will be difficult for Essen to provide the level of service that the retailer would like and that is economically beneficial to Essen. Table 2 Average Annual Performance of Channel Members and the System as Revised Channel member
Run 1
Run 2
Run 3
Run 4
Average
$68,638,738 35,950 $1,909
$74,761,258 39,286 $1,903
$78,418,188 41,268 $1,900
$75,400,666 39,622 $1,903
$74,304,713 39,032 $1,904
1.47 90.65% $12,060,444 $317.51
1.42 100% $12,449,170 $326.28
1.51 63.16% $11,895,743 $311.87
1.46 72.92% $12,178,581 $319.61
1.47 81.68% $12,145,984 $318.82
9.24 96.64% $4,018,143 $105.70
9.02 96.29% $4,080,160 $107.09
9.14 97.60% $4,043,973 $106.55
9.28 98.09% $3,992,791 $105.51
9.17 97.16% $4,033,767 $106.21
35.44 99.52% 38,017 $727,378 $19.13
35.36 99.53% 37,936 $717,363 $18.91
34.99 99.70% 37,979 $709,772 $18.69
35.14 99.40% 37,730 $748,198 $19.75
35.23 99.54% 37,916 $725,677 $19.12
$24,423,849 22.23% 123456
$17,627,666 16.08% 111111
$14,690,765 13.38% 222222
$17,184,464 15.69% 333333
$18,481,686 16.85%
Essen’s factory Total cost
Units produced Cost per unit Essen’s warehouse TO ratio Fill rate Cost
Cost per unit Retailer’s warehouse TO ratio Fill rate Cost
Cost per unit Retailer TO ratio Fill rate Units sold Cost
Cost per unit System Profit Profit as % of sales Seed Number
220
3. Would shipping by airfreight from Germany be a benefit to channel performance? To Essen?
No. Selling candies to end customers at $2,890 per thousand lb. using airfreight shipping results in obvious loss to Essen. The cost of shipping by air is is $1,833 per thousand lb., plus $1,000 material cost and $850 production cost results in a total cost much higher than selling price. There is no point to using airfreight. Running the simulation with the higher freight rate but lower variability confirms that channel profits would b e negative. 4. Is there a benefit to producing in the larger 20,000-pound batch size?
Yes. This was tested in question 2. From Tables 1 and 2, it can be seen that production costs drop from $1,928 per unit to $1,904 per unit. The overall channel cost reduction reduction overshadows the negative effects of greater length and variability in production time.
Appendix A Simulation Database for Essen USA Under Current Conditions Title: ESSEN USA Initialization 123456 11 2890
Seed value Length of simulation, years Annual price, $/unit
Customer demand pattern Generate daily demand 100 Average daily demand, units 15 Standard deviation of daily demand, units 1 Annual demand growth increment, % Monthly seasonal indices Month Index Month Index Month Index Month 1 0.25 4 0.75 7 0.75 10 2 1.25 5 0.75 8 0.75 11 3 1.25 6 0.75 9 0.75 12 Retailer/Level 1 Product item data 2220 1 35 25 1 0
Index 0.75 1.50 2.50
Item value in inventory, $/unit Customer order filling cost, $/unit Purchase order processing cost, $/order Inventory carrying cost, %/year Average customer order fill time, days Customer order fill time standard deviation, days
221
98 In-stock probability, % 670 Back order cost, $/unit Forecasting method Moving average 7 Number of periods Reorder policy Stock-to-demand control method 10 Target days of inventory 7 Review time in days Distributor/Level 2 Product item data 2220 Item value in inventory, $/unit 20 Retailer order filling cost, $/unit 75 Purchase order processing cost, $/order 25 Inventory carrying cost, %/year 2 Average retailer order fill time, days 0.2 Retailer order fill time standard deviation, days 95 In-stock probability, % 100 Back order cost, $/unit Forecasting method Moving average 30 Number of periods Reorder policy Stock-to-demand control method 45 Target days of inventory 30 Review time in days Warehouse/Level 3 Product item data 1710 Item value in inventory, $/unit 15 Distributor order filling cost, $/unit 75 Purchase order processing cost, $/order 20 Inventory carrying cost, %/year 3 Average distributor order filling time, days 0.3 Distributor order fill time, days 95 In-stock probability, % 25 Back order cost, $/unit Forecasting method Moving average 360 Number of periods Reorder policy Stock-to-demand control method 90 Target days of inventory 30 Review time in days
222
Factory/Source Produ Product ct item item data data
850 10 10 8 2 1000
Production cost, $/unit Minimum production lot size, units Warehouse order filling cost, $/unit Average production time, days Production time standard deviation, days Purchase cost, $/unit
Transportation Transpo Transport rt between between Distrib Distributor utor and Retailer Retailer
25 1 0
Transport cost, $/unit Average time in-transit, days Transit time standard deviation, days
Transpo Transport rt between between Warehou Warehouse se and Distrib Distributor utor
70 5 1
Transport cost, $/unit Average time in-transit, days Transit time standard deviation, days
Transpo Transport rt between between Factory Factory and Warehous Warehouse e
78 9 3
Transport cost, $/unit Average time in-transit, days Transit time standard deviation, days
Appendix B Benchmark Simulation Results with Seed Number 123456 and Simulation Length of 11 Years SUPPLY CHANNEL REPORT FOR SIMULATED YEARS 2 TO 11
Yearly average $109 $109,8 ,868 68,5 ,552 52 37,918,000 71,950,552
Sim ulat ed period $1,0 $1,098 98,6 ,685 85,5 ,520 20 379,180,000 719,505,520
32,230,300
322,303,000
949,025 2,656,010 2,957,604
9,490,250 26,560,100 29,576,040
38,017 759,890 569,145
380,168 7,598,900 5,691,450
1,638 683 675
16,380 6,825 6,750
FINANCIAL PERFORMANCE Reve Revenu nue e Cost of purchased goods Gross margin Production cost Transportation Transportation costs: Distributor to retailer Warehouse to distributor Factory to warehouse Sales orde Sales order r handl handlin ing g cost cost for: for: Customer orders Retailer orders Distributor orders Order Order pro process cessing ing cos cost t for: Orders to distributor Orders to warehouses Orders to factory Inventory Inventory costs
223
260,414 1,627,909 1,988,984
2,604,145 16,279,092 19,889,837
R e t a i l er D i s t r i b u t or Ware house
2,584,458 535,730 363,478
25,844,580 5,357,300 3,634,775
Back ord Back order er cost costs s Reta iler D i s t r i bu t o r W a r e h o us e
$24, 24,426, 26,593
$24 $244,265 265,928 928
Net prof rofit contrib ribution ion
Appendix C Simulation Database for Essen USA as Revised for Service Improvement Title: ESSEN USA Initialization 123456 11 2890
Seed value Length of simulation, years Annual price, $/unit
Customer demand pattern Generate daily demand 100 Average daily demand, units 15 Standard deviation of daily demand, units 1 Annual demand growth increment, % Monthly seasonal indices Month Index Month Index Month Index Month 1 0.25 4 0.75 7 0.75 10 2 1.25 5 0.75 8 0.75 11 3 1.25 6 0.75 9 0.75 12
Index 0.75 1.50 2.50
Retailer/Level 1 Product item data 2220 Item value in inventory, $/unit 1 Customer order filling cost, $/unit 35 Purchase order processing cost, $/order 25 Inventory carrying cost, %/year 1 Average customer order fill time, days 0 Customer order fill time standard deviation, days 98 In-stock probability, % 670 Back order cost, $/unit Forecasting method Moving average 7 Number of periods Reorder policy Stock-to-demand control method 14 Target days of inventory 7 Review time in days
224
Distributor/Level 2 Product item data 2220 Item value in inventory, $/unit 20 Retailer order filling cost, $/unit 75 Purchase order processing cost, $/order 25 Inventory carrying cost, %/year 2 Average retailer order fill time, days 0.2 Retailer order fill time standard deviation, days 95 In-stock probability, % 100 Back order cost, $/unit Forecasting method Moving average 7 Number of periods Reorder policy Stock-to-demand control method 35 Target days of inventory 7 Review time in days Warehouse/Level 3 Product item data 1710 Item value in inventory, $/unit 15 Distributor order filling cost, $/unit 75 Purchase order processing cost, $/order 20 Inventory carrying cost, %/year 3 Average distributor order filling time, days 0.3 Distributor order fill time standard deviation, days 95 In-stock probability, % 25 Back order cost, $/unit Forecasting method Moving average 7 Number of periods Reorder policy Stock-to-demand control method 25 Target days of inventory 7 Review time in days Factory/Source Produ Product ct item item data data
825 20 10 10 2.1 1000
Production cost, $/unit Minimum production lot size, units Warehouse order filling cost, $/unit Average production time, days Production time standard deviation, days Purchase cost, $/unit
225
Transportation Transpo Transport rt between between Distrib Distributor utor and Retailer Retailer
25 1 0
Transport cost, $/unit Average time in-transit, days Transit time standard deviation, days
Transpo Transport rt between between Warehou Warehouse se and Distrib Distributor utor
70 5 1
Transport cost, $/unit Average time in-transit, days Transit time standard deviation, days
Transpo Transport rt between between Factory Factory and Warehous Warehouse e
78 9 3
Transport cost, $/unit Average time in-transit, days Transit time standard deviation, days
226