BA/E 296-1Global Supply Chain ManagementInstructor: Jayashankar Swaminathan
HW 3 – Atlanta Transportation System
Group Member: Xiaoqing Ge, Sally Huang, Qian Wang Page 1 of 5
______
BA/E 296Global Supply Chain Management
Case 4 Report
Global Plant Location - Applichem
(Harvard Case)
Qian Wang
Tsai-Lan Huang
Xiaoqing Ge
March 15, 2000
- How would you measure the productivity at a plant and perform a fair comparison of the performance of Applichem's six plants?
We assume that the 'Number of People at Each Operation at Each Plant' means 'number of people for each product type at each plant'. Therefore the product variety difference among different plants should already be taken into consideration in Exhibit 3.
But the package variety is also an important factor for performing a fair comparison of the performance, since the package variety difference is very big among all the 6 plants. For example, Gary is carrying 80 package sizes while most of the others carrying only 1. Therefore we divide the package labor (in Exhibit 3) by the number of package sizes the plant runs, and calculate the subtotal of direct labor based on it. By this modified calculation, the plant productivity is as shown in table 1.2.
Table 1.1Direct labor
Mexico / Canada / Venezuela / Frankfort / Gary / SunchemOriginal package labor in Exhibit 3 / 10.4 / 5 / 6.2 / 14.6 / 11.3 / 2.7
Modified package labor / 10.4 / 5 / 6.2 / 14.6 / 0.14125 / 1.35
Table 1.2Productivity
Mexico / Canda / Venezuela / Frankfort / Gary / Sunchem / averageby direct labor / 0.882 / 0.215 / 0.318 / 0.828 / 0.591 / 0.278 / 0.518523
by direct labor-modified / 0.882 / 0.215 / 0.315 / 0.828 / 1.116 / 0.304 / 0.610116
by indirect labor / 0.691 / 0.167 / 0.373 / 0.945 / 0.405 / 0.241 / 0.47017
by total / 0.387 / 0.094 / 0.172 / 0.441 / 0.240 / 0.129 / 0.243886
by total-modified / 0.387 / 0.094 / 0.171 / 0.441 / 0.297 / 0.134 / 0.254144
Similarly, for the manufacturing cost, it is unfair to measure using total manufacturing cost which includes the packaging cost. The manufacturing cost before packaging is a more realistic measurement.
- Why do you think the plant have different productivity?
The productivity depends on the product make-up, equipment efficiency, employee loyalty, improvements, operator technology, national labor rules, product quality etc. The following table summarizes the differences in all aspects among the six plants:
Product Make-uprelease-ease / other / packages / Equipment Efficiency / Employee Loyalty / Improve-ment / Operator Technology / National Labor Rules / Product Quality
Mexico / 1-6-1 / 68, dry-78 / - / - / ok / - / -
Canda / 1-4-1 / 55 / non-union / - / - / - / high
Venezuela / 1-1-1 / 64 / - / Bad / low / - / -
Frankfort / 2-12-1 / 71-74, 1961 / - / Good / - / - / -
Gary / 8-19-80 / 59-64 / excellent / - / - / - / -
Sunchem / 1-1-2 / 57, redesign-69 / non- union / best / excellent / severe / high
- A new cost measurement. Set up a linear program using the demand data, transportation and manufacturing costs and capacities to find the optimal production for Applichem's global network for 1982.
Because there is no technology break-through from 1977-1982, we assume that the manufacturing cost is affected by exchange rate and inflation rate only. The exchange rates and inflation rates vary greatly by country. For example, Mexico has the most dramatic change between 1982 and 1981, while Japan has the lowest overall inflation rate change, and Venezuela has no exchange rate change at all. Since all the plants' equipment setup date and manufacturing starting date are before 1975, the plant's performance (manufacturing cost) should be based on the modified cost eliminating the effect of outside influence out of control of plants (exchange rate and inflation rate). For the 1982 manufacturing, if it was performed in 1977-81, the corresponding cost should be:
year / Mexico / Canda / Venezuela / Frankfort / Gary / Sunchem1982 (check) / 92.63 / 93.25 / 112.31 / 73.34 / 89.15 / 149.24
1981 / 216.7922 / 91.70873 / 103.9096 / 73.29401 / 86.71935 / 156.7062
1980 / 198.4 / 82.52094 / 91.30894 / 78.05058 / 78.40809 / 167.4084
1979 / 162.1102 / 73.9437 / 76.06035 / 82.23734 / 67.50937 / 120.3686
1978 / 137.6809 / 63.54112 / 69.66872 / 74.14891 / 59.82537 / 138.1361
1977 / 118.8233 / 63.51462 / 64.82935 / 63.887 / 55.66974 / 114.9789
And the ratio with average cost (among 6 plants) is:
Mexico / Canda / Venezuela / Frankfort / Gary / Sunchem1982 / 0.911234 / 0.917333 / 1.104833 / 0.721472 / 0.877 / 1.468127
1981 / 1.78398 / 0.75467 / 0.85507 / 0.603135 / 0.713612 / 1.289533
1980 / 1.710107 / 0.711288 / 0.787036 / 0.672756 / 0.675838 / 1.442975
1979 / 1.670581 / 0.762006 / 0.783818 / 0.847473 / 0.695698 / 1.240424
1978 / 1.521333 / 0.70211 / 0.769819 / 0.819323 / 0.661053 / 1.526362
1977 / 1.48004 / 0.791126 / 0.807502 / 0.795764 / 0.693412 / 1.432156
From the ratio comparison, we can see Mexico plant's manufacturing cost is reduced in 1982 dramatically mostly because of the exchange rate and inflation rate change. Overall, Gary did the best (around 70% of the average, depends on which year we are looking at), while Mexico manufacturing cost is between 78%-48% higher than average, depends on the year.
The linear programming is formulated as:
Minimize:
Subject to:
where, is the manufacturing cost vector, is the transportation cost from site i to site j, is the demand at site i, is the capacity, is the production, is the quantity of product transported from site i to site j.
We solved the LP using AMPL, and the code and solution is as attachment A. The final optimal cost is $68,300,900. From the optimal solution, we observed that we will allocate the manufacturing to the plant with lowest cost first (Frankfurt and Gary). For those with higher manufacturing cost, then we need to consider the tradeoff between transportation cost and manufacturing cost. It’s cheaper to produce at Canada and Mexico than transfer form other places, while it’s cheaper to transfer from other places than produce at Sunchem and Venezuela.
- Find the optimal production for Applichem's global network of sites for 1982 taking into account custom duties. What changes do you observe? How different is it from the current production?
This problem is different from problem 3 only in the transportation cost. The linear programming is as attachment B. The optimal production at each site and the optimal quantity transported between each two sites turn out to be exactly the same as we obtained for Question 3. But, of course, the total cost is increased to $69,693,700 due to the added duties.
- Find the optimal production for Applichem's global network of sites for 1977 taking into account custom duties assuming that the demand remains the same as in 1982. What do you observe?
The only difference between this and the model in Q4 is the manufacturing cost, where the 1977 manufacturing cost should be what we obtained in Q3. The linear programming is as attachment C. The calculation results shows that the optimal production and import amounts are quite different from the 1982 operations. And the total cost is the lowest if we produce at all plants. Therefore we can conclude that international exchange rates and price volatility (inflation rate) have tremendous influence on manufacturing planning.
- Suppose it was known that to operate each site it costs $1 billion to set up and $100 for 1000 pounds of capacity. How will you formulate an integer program to set up a new production network that determines which locations to setup and how much capacity to have and how much to produce at each of these plants for 1982 data?
According to the new situation, the integral programming model could be set up as follows:
Minimize:
Subject to:
is the setup cost =1,000,000,000, and is the binary variable to determine whether to produce or not. The linear programming is as attachment D.
Consider to the setup cost and scale operation cost, the optimal solution suggests us to produce only at Frankfurt, Gary and Mexico. We find these plants are again with the lowest manufacturing cost. But if we remove the capacity constraint of each plant, we will end up with solution that produce everything in Frankfurt and improve the objective value about 60%.
- How useful is a model such as the one developed in (5) under (a) uncertainty in demand; (b) uncertainty in currency fluctuations?
We build the model from Question 3~5 based on assumption of known manufacturing cost, transportation cost and demand required at different regions. When each of the assumption is in uncertainty, i.e (a) demand; (b) currency fluctuation, we can do the sensitivity analysis to find out the point when solutions change and be prepared for it. For example, we may use this model to analyze that when the demand at Venezuela is up to a certain level then we need to setup a plant there instead of transport from other plants. We can also use this linear programming to simulate different scenarios and have a whole view of possible results.
- How could you advice Joe Spadaro to configure his worldwide manufacturing systems?
Due to the different cultural, geographical, political and demographical environment that the subsidiaries of an international firm operate in, it is very important to set up a rational performance measurement system taking the above factors into consideration for the management to have a clear picture of the current operations.
Joe Spadaro should first understand that Applichem’s six plants have different product focus and their operation (such as packaging) specifications are different. The market situation and technology advancement level are also different. The foreign exchange rates as well as the tariff policies also exert influence on the performance of a local factory. It is therefore not advisable to use a rigid performance measurements to evaluate the performance of the six plants. Besides, instead of vertical comparison, we suggest Joe to compare the plants performance with similar local factories which might be more realistic benchmarks.
Apart from qualitative analyses, some quantitative tools can also be used in analyzing complex problems that involve multiple variables. For example, computer simulations could be used to analyze the behavior of an organization under the influence of uncertainties and provide supporting information for shielding risks. An enterprise-wide productivity study involving all the six plants has just been completed in Applichem, which has collected valuable data for quantitative analyses. An enterprise-wide resource leveling and allocation model could be set up based on the above information. As indicated above, Linear Programming is especially applicable to Applichem’s global network planning, which could solve some of the tactic problems that Joe is facing.