Solutions to Case Problems Manual to Accompany An Introduction To Management Science Quantitative Approaches To Decision Making Twelfth Edition David R. Anderson University of Cincinnati Dennis J. Sweeney University of Cincinnati Thomas A. Williams Rochester Institute of Technology R. Kipp Martin University of Chicago South-Western Cincinnati, Ohio Contents Preface Chapter 1: Introduction ♦ Scheduling a Golf League Chapter 9: Project Scheduling: PERT/CPM ♦ R.C. Coleman Chapter 2: An Introduction to Linear Programming ♦ Workload Balancing ♦ Production Strategy ♦ Hart Venture Capital Chapter 10: Inventory Models ♦ Wagner Fabricating Company ♦ River City Fire Department Chapter 3: Linear Programming: Sensitivity Analysis and Interpretation of Solution ♦ Product Mix ♦ Investment Strategy ♦ Truck Leasing Strategy Chapter 4: Linear Programming Applications in Marketing, Finance and Operations Management ♦ Planning an Advertising Campaign ♦ Phoenix Computer ♦ Textile Mill Scheduling ♦ Workforce Scheduling ♦ Duke Energy Coal Allocation Chapter 6: Distribution and Network Models ♦ Solution Plus ♦ Distribution Systems Design Chapter 11: Waiting Line Models ♦ Regional Airlines ♦ Office Equipment, Inc. Chapter 12: Simulation ♦ Tri-State Corporation ♦ Harbor Dunes Golf Course ♦ County Beverage Drive-Thru Chapter 13: Decision Analysis ♦ Property Purchase Strategy ♦ Lawsuit Defense Strategy Chapter 14: Multicriteria Decision Problems ♦ EZ Trailers, Inc. Chapter 15: Forecasting ♦ Forecasting Sales ♦ Forecasting Lost Sales Chapter 7: Integer Linear Programming ♦ Textbook Publishing ♦ Yeager National Bank ♦ Production Scheduling with Changeover Costs Chapter 16: Markov Processes ♦ Dealer’s Absorbing State Probabilities in Black Jack Chapter 8: Nonlinear Optimization Models ♦ Portfolio Optimization with Transaction Costs Chapter 21: Dynamic Programming ♦ Process Design Preface The purpose of An Introduction to Management Science is to provide students with a sound conceptual understanding of the role management science pays in the decision-making process. The text emphasizes the application of management science by using problem situations to introduce each of the management science concepts and techniques. The book has been specifically designed to meet the needs of nonmathematicians who are studying business and economics. The Solutions to Case Problems Manual contains solutions to the case problems. Note: The solutions to the end-of-chapter problems and learning objectives for each chapter are included in the Solutions Manual. Acknowledgements We would like to provide a special acknowledgement to Catherine J. Williams for her efforts in preparing the Instructor's Manual. We are also indebted to our acquisitions editor Charles E. McCormick, Jr. and our developmental editor Alice C. Denny for their support during the preparation of this manual. David R. Anderson Dennis J. Sweeney Thomas A. Williams R. Kipp Martin Chapter 1 Introduction Case Problem: Scheduling a Golf League Note to Instructor: This case problem illustrates the value of the rational management science approach. The problem is easy to understand and, at first glance, appears simple. But, most students will have trouble finding a solution. The solution procedure suggested involves decomposing a larger problem into a series of smaller problems that are easier to solve. The case provides students with a good first look at the kinds of problems where management science is applied in practice. The problem is a real one that one of the authors was asked by the Head Professional at Royal Oak Country Club for help with. Solution: Scheduling problems such as this occur frequently, and are often difficult to solve. The typical approach is to use trial and error. An alternative approach involves breaking the larger problem into a series of smaller problems. We show how this can be done here using what we call the Red, White, and Blue algorithm. Suppose we break the 18 couples up into 3 divisions, referred to as the Red, White, and Blue divisions. The six couples in the Red division can then be identified as R1, R2, R3, R4, R5, R6; the six couples in the White division can be identified as W1, W2,…, W6; and the six couples in the Blue division can be identified as B1, B2,…, B6. We begin by developing a schedule for the first 5 weeks of the season so that each couple plays every other couple in its own division. This can be done fairly easily by trial and error. Shown below is the first 5-week schedule for the Red division. Week 1 R1 vs. R2 R3 vs. R4 R5 vs. R6 Week 2 R1 vs. R3 R2 vs. R5 R4 vs. R6 Week 3 R1 vs. R4 R2 vs. R6 R3 vs. R5 Week 4 R1 vs. R5 R2 vs. R4 R3 vs. R6 Week 5 R1 vs. R6 R2 vs. R3 R4 vs. R5 Similar 5-week schedules can be developed for the White and Blue divisions by replacing the R in the above table with a W or a B. To develop the schedule for the next 3 weeks, we create 3 new six-couple divisions by pairing 3 of the teams in each division with 3 of the teams in another division; for example, (R1, R2, R3, W1, W2, W3), (B1, B2, B3, R4, R5, R6), and (W4, W5, W6, B4, B5, B6). Within each of these new divisions, matches can be scheduled for 3 weeks without any couples playing a couple they have played before. For instance, a 3-week schedule for the first of these divisions is shown below: Week 6 R1 vs. W1 R2 vs. W2 R3 vs. W3 Week 7 R1 vs. W2 R2 vs. W3 R3 vs. W1 Week 8 R1 vs. W3 R2 vs. W1 R3 vs. W2 A similar 3-week schedule can be easily set up for the other two new divisions. This will provide us with a schedule for the first 8 weeks of the season. For the final 9 weeks, we continue to create new divisions by pairing 3 teams from the original Red, White and Blue divisions with 3 teams from the other divisions that they have not yet been paired with. Then a 3week schedule is developed as above. Shown below is a set of divisions for the next 9 weeks. CP - 1 Chapter 1 Weeks 9-11 (R1, R2, R3, W4, W5, W6) (W1, W2, W3, B1, B2, B3) (R4, R5, R6, B4, B5, B6) (W1, W2, W3, B4, B5, B6) (W4, W5, W6, R4, R5, R6) (W1, W2, W3, R4, R5, R6) (W4, W5, W6, B1, B2, B3) Weeks 12-14 (R1, R2, R3, B1, B2, B3) Weeks 15-17 (R1, R2, R3, B4, B5, B6) This Red, White and Blue scheduling procedure provides a schedule with every couple playing every other couple over the 17-week season. If one of the couples should cancel, the schedule can be modified easily. Designate the couple that cancels, say R4, as the Bye couple. Then whichever couple is scheduled to play couple R4 will receive a Bye in that week. With only 17 couples a Bye must be scheduled for one team each week. This same scheduling procedure can obviously be used for scheduling sports teams and or any other kinds of matches involving 17 or 18 teams. Modifications of the Red, White and Blue algorithm can be employed for 15 or 16 team leagues and other numbers of teams. CP - 2 Chapter 2 An Introduction to Linear Programming Case Problem 1: Workload Balancing 1. Model DI-910 DI-950 Production Rate (minutes per printer) Line 1 Line 2 3 4 6 2 Profit Contribution ($) 42 87 Capacity: 8 hours × 60 minutes/hour = 480 minutes per day D1 = number of units of the DI-910 produced D2 = number of units of the DI-950 produced Let Max s.t. 42D1 + 87D2 3D1 + 6D2 4D1 + 2D2 D 1, D 2 ≥ 0 ≤ 480 ≤ 480 Line 1 Capacity Line 2 Capacity Using The Management Scientist, the optimal solution is D1 = 0, D2 = 80. The value of the optimal solution is $6960. Management would not implement this solution because no units of the DI-910 would be produced. 2. Adding the constraint D1 ≥ D2 and resolving the linear program results in the optimal solution D1 = 53.333, D2 = 53.333. The value of the optimal solution is $6880. 3. Time spent on Line 1: 3(53.333) + 6(53.333) = 480 minutes Time spent on Line 2: 4(53.333) + 2(53.333) = 320 minutes Thus, the solution does not balance the total time spent on Line 1 and the total time spent on Line 2. This might be a concern to management if no other work assignments were available for the employees on Line 2. 4. Let T1 = total time spent on Line 1 T2 = total time spent on Line 2 Whatever the value of T2 is, T1 ≤ T2 + 30 T1 ≥ T2 - 30 Thus, with T1 = 3D1 + 6D2 and T2 = 4D1 + 2D2 3D1 + 6D2 ≤ 4D1 + 2D2 + 30 3D1 + 6D2 ≥ 4D1 + 2D2 − 30 CP - 3 Chapter 2 Hence, −1D1 + 4D2 ≤ 30 −1D1 + 4D2 ≥ −30 Rewriting the second constraint by multiplying both sides by -1, we obtain −1D1 + 4D2 ≤ 30 1D1 − 4D2 ≤ 30 Adding these two constraints to the linear program formulated in part (2) and resolving using The Management Scientist, we obtain the optimal solution D1 = 96.667, D2 = 31.667. The value of the optimal solution is $6815. Line 1 is scheduled for 480 minutes and Line 2 for 450 minutes. The effect of workload balancing is to reduce the total contribution to profit by $6880 - $6815 = $65 per shift. 5. The optimal solution is D1 = 106.667, D2 = 26.667. The total profit contribution is 42(106.667) + 87(26.667) = $6800 Comparing the solutions to part (4) and part (5), maximizing the number of printers produced (106.667 + 26.667 = 133.33) has increased the production by 133.33 - (96.667 + 31.667) = 5 printers but has reduced profit contribution by $6815 - $6800 = $15. But, this solution results in perfect workload balancing because the total time spent on each line is 480 minutes. Case Problem 2: Production Strategy 1. Let Max s.t. BP100 = the number of BodyPlus 100 machines produced BP200 = the number of BodyPlus 200 machines produced 371BP100 + 8BP100 5BP100 2BP100 -0.25BP100 + + + + 461BP200 12BP200 10BP200 2BP200 0.75BP200 ≤ ≤ ≤ ≥ 600 450 140 0 BP100, BP200 ≥ 0 CP - 4 Machining and Welding Painting and Finishing Assembly, Test, and Packaging BodyPlus 200 Requirement Solutions to Case Problems BP200 80 Number of BodyPlus 200 . 70 Assembly, Test, and Packaging 60 50 Machining and Welding 40 BodyPlus 200 Requirement 30 20 10 Painting and Finishing BP100 30 40 50 60 70 80 90 100 Number of BodyPlus 100 Optimal Solution 0 10 20 Optimal solution: BP100 = 50, BP200 = 50/3, profit = $26,233.33. Note: If the optimal solution is rounded to BP100 = 50, BP200 = 16.67, the value of the optimal solution will differ from the value shown. The value we show for the optimal solution is the same as the value that will be obtained if the problem is solved using a linear programming software package such as The Management Scientist. 2. In the short run the requirement reduces profits. For instance, if the requirement were reduced to at least 24% of total production, the new optimal solution is BP100 = 1425/28, BP200 = 225/14, with a total profit of $26,290.18; thus, total profits would increase by $56.85. Note: If the optimal solution is rounded to BP100 = 50.89, BP200 = 16.07, the value of the optimal solution will differ from the value shown. The value we show for the optimal solution is the same as the value that will be obtained if the problem is solved using a linear programming software package such as The Management Scientist. 3. If management really believes that the BodyPlus 200 can help position BFI as one of the leader's in high-end exercise equipment, the constraint requiring that the number of units of the BodyPlus 200 produced be at least 25% of total production should not be changed. Since the optimal solution uses all of the available machining and welding time, management should try to obtain additional hours of this resource. CP - 5 Chapter 2 Case Problem 3: Hart Venture Capital 1. Let S = fraction of the Security Systems project funded by HVC M = fraction of the Market Analysis project funded by HVC Max s.t. 1,800,000S + 1,600,000M 600,000S 600,000S 250,000S S + + + 500,000M 350,000M 400,000M S,M ≥ M 0 ≤ ≤ ≤ ≤ ≤ 800,000 700,000 500,000 1 1 Year 1 Year 2 Year 3 Maximum for S Maximum for M The solution obtained using The Management Scientist software package is shown below: OPTIMAL SOLUTION Objective Function Value = 2486956.522 Variable -------------S M Value --------------0.609 0.870 Reduced Costs -----------------0.000 0.000 Constraint -------------1 2 3 4 5 Slack/Surplus --------------0.000 30434.783 0.000 0.391 0.130 Dual Prices -----------------2.783 0.000 0.522 0.000 0.000 OBJECTIVE COEFFICIENT RANGES Variable -----------S M Lower Limit --------------No Lower Limit No Lower Limit Current Value --------------1800000.000 1600000.000 Upper Limit --------------No Upper Limit No Upper Limit Current Value --------------800000.000 700000.000 500000.000 1.000 1.000 Upper Limit --------------822950.820 No Upper Limit No Upper Limit No Upper Limit No Upper Limit RIGHT HAND SIDE RANGES Constraint -----------1 2 3 4 5 Lower Limit --------------No Lower Limit 669565.217 461111.111 0.609 0.870 CP - 6 Solutions to Case Problems Thus, the optimal solution is S = 0.609 and M = 0.870. In other words, approximately 61% of the Security Systems project should be funded by HVC and 87% of the Market Analysis project should be funded by HVC. The net present value of the investment is approximately $2,486,957. 2. Security Systems Market Analysis Total Year 1 $365,400 $435,000 $800,400 Year 2 $365,400 $304,500 $669,900 Year 3 $152,250 $348,000 $500,250 Note: The totals for Year 1 and Year 3 are greater than the amounts available. The reason for this is that rounded values for the decision variables were used to compute the amount required in each year. To see why this situation occurs here, first note that each of the problem coefficients is an integer value. Thus, by default, when The Management Scientist prints the optimal solution, values of the decision variables are rounded and printed with three decimal places. To increase the number of decimal places shown in the output, one or more of the problem coefficients can be entered with additional digits to the right of the decimal point. For instance, if we enter the coefficient of 1 for S in constraint 4 as 1.000000 and resolve the problem, the new optimal values for S and D will be rounded and printed with six decimal places. If we use the new values in the computation of the amount required in each year, the differences observed for year 1 and year 3 will be much smaller than we obtained using the values of S = 0.609 and M = 0.870. 3. If up to $900,000 is available in year 1 we obtain a new optimal solution with S = 0.689 and M = 0.820. In other words, approximately 69% of the Security Systems project should be funded by HVC and 82% of the Market Analysis project should be funded by HVC. The net present value of the investment is approximately $2,550,820. The solution obtained using The Management Scientist software package follows: OPTIMAL SOLUTION Objective Function Value = Variable -------------S M Constraint -------------1 2 3 4 5 2550819.672 Value --------------0.689 0.820 Slack/Surplus --------------77049.180 0.000 0.000 0.311 0.180 Reduced Costs -----------------0.000 0.000 Dual Prices -----------------0.000 2.098 2.164 0.000 0.000 OBJECTIVE COEFFICIENT RANGES Variable -----------S M Lower Limit --------------No Lower Limit No Lower Limit Current Value --------------1800000.000 1600000.000 CP - 7 Upper Limit --------------No Upper Limit No Upper Limit Chapter 2 RIGHT HAND SIDE RANGES Constraint -----------1 2 3 4 5 4. Lower Limit --------------822950.820 No Lower Limit No Lower Limit 0.689 0.820 Upper Limit --------------No Upper Limit 802173.913 630555.556 No Upper Limit No Upper Limit If an additional $100,000 is made available, the allocation plan would change as follows: Security Systems Market Analysis Total 5. Current Value --------------900000.000 700000.000 500000.000 1.000 1.000 Year 1 $413,400 $410,000 $823,400 Year 2 $413,400 $287,000 $700,400 Year 3 $172,250 $328,000 $500,250 Having additional funds available in year 1 will increase the total net present value. The value of the objective function increases from $2,486,957 to $2,550,820, a difference of $63,863. But, since the allocation plan shows that $823,400 is required in year 1, only $23,400 of the additional $100,00 is required. We can also determine this by looking at the slack variable for constraint 1 in the new solution. This value, 77049.180, shows that at the optimal solution approximately $77,049 of the $900,000 available is not used. Thus, the amount of funds required in year 1 is $900,000 - $77,049 = $822,951. In other words, only $22,951 of the additional $100,000 is required. The differences between the two values, $23,400 and $22,951, is simply due to the fact that the value of $23,400 was computed using rounded values for the decision variables. The value of $22,951 is computed internally in The Management Scientist output and is not subject to this rounding. Thus, the most accurate value is $22,951. CP - 8 Chapter 3 Linear Programming: Sensitivity Analysis and Interpretation of Solution Case Problem 1: Product Mix Note to Instructor: The difference between relevant and sunk costs is critical. The cost of the shipment of nuts is a sunk cost. Practice in applying sensitivity analysis to a business decision is obtained. You may want to suggest that sensitivity analyses other than the ones we have suggested be undertaken. 1. Cost per pound of ingredients Almonds $7500/6000 = $1.25 Brazil $7125/7500 = $.95 Filberts $6750/7500 = $.90 Pecans $7200/6000 = $1.20 Walnuts $7875/7500 = $1.05 Cost of nuts in three mixes: Regular mix: .15($1.25) + .25($.95) + .25($90) + .10($1.20) + .25($1.05) = $1.0325 Deluxe mix .20($1.25) + .20($.95) + .20($.90) + .20($1.20) + .20($1.05) = $1.07 Holiday mix: .25($1.25) + .15($.95) + .15($.90) + .25($1.20) + .20($1.05) = $1.10 2. Let R = pounds of Regular Mix produced D = pounds of Deluxe Mix produced H = pounds of Holiday Mix produced Note that the cost of the five shipments of nuts is a sunk (not a relevant) cost and should not affect the decision. However, this information may be useful to management in future pricing and purchasing decisions. A linear programming model for the optimal product mix is given. The following linear programming model can be solved to maximize profit contribution for the nuts already purchased. Max s.t. 1.65R + 2.00D + 2.25H 0.15R 0.25R 0.25R 0.10R 0.25R R + + + + + 0.20D 0.20D 0.20D 0.20D 0.20D + + + + + 0.25H 0.15H 0.15H 0.25H 0.20H D H R, D, H ≥ 0 CP - 9 ≤ ≤ ≤ ≤ ≤ ≥ ≥ ≥ 6000 7500 7500 6000 7500 10000 3000 5000 Almonds Brazil Filberts Pecans Walnuts Regular Deluxe Holiday Chapter 3 The solution found using The Management Scientist is shown below. Objective Function Value = 61375.000 Variable -------------R D H Value --------------17500.000 10624.999 5000.000 Reduced Costs -----------------0.000 0.000 0.000 Constraint -------------1 2 3 4 5 6 7 8 Slack/Surplus --------------0.000 250.000 250.000 875.000 0.000 7500.000 7624.999 0.000 Dual Prices -----------------8.500 0.000 0.000 0.000 1.500 0.000 0.000 -0.175 OBJECTIVE COEFFICIENT RANGES Variable -----------R D H Lower Limit --------------1.500 1.892 No Lower Limit Current Value --------------1.650 2.000 2.250 Upper Limit --------------2.000 2.200 2.425 Current Value --------------6000.000 7500.000 7500.000 6000.000 7500.000 10000.000 3000.000 5000.000 Upper Limit --------------6583.333 No Upper Limit No Upper Limit No Upper Limit 7750.000 17500.000 10624.999 9692.307 RIGHT HAND SIDE RANGES Constraint -----------1 2 3 4 5 6 7 8 Lower Limit --------------5390.000 7250.000 7250.000 5125.000 6750.000 No Lower Limit No Lower Limit -0.000 3. From the dual prices it can be seen that additional almonds are worth $8.50 per pound to TJ. Additional walnuts are worth $1.50 per pound. From the slack variables, we see that additional Brazil nut, Filberts, and Pecans are of no value since they are already in excess supply. 4. Yes, purchase the almonds. The dual price shows that each pound is worth $8.50; the dual price is applicable for increases up to 583.33 pounds. CP - 10 Solutions to Case Problems Resolving the problem by changing the right-hand side of constraint 1 from 6000 to 7000 yields the following optimal solution. The optimal solution has increased in value by $4958.34. Note that only 583.33 pounds of the additional almonds were used, but that the increase in profit contribution more than justifies the $1000 cost of the shipment. Objective Function Value = 66333.336 Variable -------------R D H Value --------------11666.667 17916.668 5000.000 Reduced Costs -----------------0.000 0.000 0.000 Constraint -------------1 2 3 4 5 6 7 8 Slack/Surplus --------------416.667 250.000 250.000 0.000 0.000 1666.667 14916.667 0.000 Dual Prices -----------------0.000 0.000 0.000 5.667 4.333 0.000 0.000 -0.033 OBJECTIVE COEFFICIENT RANGES Variable -----------R D H Lower Limit --------------1.000 1.976 No Lower Limit Current Value --------------1.650 2.000 2.250 Upper Limit --------------1.750 3.300 2.283 Current Value --------------7000.000 7500.000 7500.000 6000.000 7500.000 10000.000 3000.000 5000.000 Upper Limit --------------No Upper Limit No Upper Limit No Upper Limit 6250.000 7750.000 11666.667 17916.668 15529.412 RIGHT HAND SIDE RANGES Constraint -----------1 2 3 4 5 6 7 8 5. Lower Limit --------------6583.333 7250.000 7250.000 4210.000 7250.000 No Lower Limit No Lower Limit 0.002 From the dual prices it is clear that there is no advantage to not satisfying the orders for the Regular and Deluxe mixes. However, it would be advantageous to negotiate a decrease in the Holiday mix requirement. CP - 11 Chapter 3 Case Problem 2: Investment Strategy 1. The first step is to develop a linear programming model for maximizing return subject to constraints for funds available, diversity, and risk tolerance. Let G = Amount invested in growth fund I = Amount invested in income fund M = Amount invested in money market fund The LP formulation and optimal solution found using The Management Scientist are shown. MAX .18G +.125I +.075M S.T. 1) 2) 3) 4) 5) 6) 7) G + I + M < 800000 .8G -.2I -.2M > 0 .6G -.4I -.4M < 0 -.2G +.8I -.2M > 0 -.5G +.5I -.5M < 0 -.3G -.3I +.7M > 0 .05G + .02I -.04M < 0 Funds Available Min growth fund Max growth fund Min income fund Max income fund Min money market fund Max risk OPTIMAL SOLUTION Objective Function Value = 94133.336 Variable -------------G I M Value --------------248888.906 160000.000 391111.094 Reduced Costs -----------------0.000 0.000 0.000 Constraint -------------1 2 3 4 5 6 7 Slack/Surplus --------------0.000 88888.898 71111.109 0.000 240000.000 151111.109 0.000 Dual Prices -----------------0.118 0.000 0.000 -0.020 0.000 0.000 1.167 OBJECTIVE COEFFICIENT RANGES Variable -----------G I M Lower Limit --------------0.150 -0.463 0.015 Current Value --------------0.180 0.125 0.075 CP - 12 Upper Limit --------------No Upper Limit 0.145 0.180 Solutions to Case Problems RIGHT HAND SIDE RANGES Constraint -----------1 2 3 4 5 6 7 Lower Limit --------------0.188 No Lower Limit -71111.109 -106666.641 -240000.000 No Lower Limit -8000.000 Current Value --------------800000.000 0.000 0.000 0.000 0.000 0.000 0.000 Upper Limit --------------No Upper Limit 88888.898 No Upper Limit 133333.313 No Upper Limit 151111.109 6399.999 Rounding to the nearest dollar, the portfolio recommendation for Langford is as follows. Fund: Growth Income Money Market Total Amount Invested $248,889 160,000 391,111 $800,000 Yield = 94,133 / 800,000 = .118 The portfolio yield is .118 or 11.8%. Note that the portfolio yield equals the dual price for the funds available constraint. 2. If Langford’s risk index is increased by .005 that is the same as increasing the right-hand side of constraint 7 by .005 (800,000) = 4000. Since this amount of increase is within the right-hand-side range, we would expect an increase in return of 1.167 (4000) = 4668. The revised formulation and new optimal solution are shown below. Except for rounding, the value has increased as predicted; the new optimal allocation is Fund: Growth Income Money Market Total Amount Invested $293,333 160,000 346,667 $800,000 The portfolio yield becomes 98,800/800,000 = .124 or 12.4% MAX .18G +.125I +.075M S.T. 1) 2) 3) 4) 5) 6) 7) G + I + M < 800000 .8G -.2I -.2M > 0 .6G -.4I -.4M < 0 -.2G +.8I -.2M > 0 -.5G +.5I -.5M < 0 -.3G -.3I +.7M > 0 .045G + .015I-.045M < 0 CP - 13 Chapter 3 OPTIMAL SOLUTION Objective Function Value = 3. 98800.000 Variable -------------G I M Value --------------293333.313 160000.000 346666.656 Reduced Costs -----------------0.000 0.000 0.000 Constraint -------------1 2 3 4 5 6 7 Slack/Surplus --------------0.000 133333.328 26666.666 0.000 240000.000 106666.664 0.000 Dual Prices -----------------0.124 0.000 0.000 -0.020 0.000 0.000 1.167 Since .16 is in the objective coefficient range for the growth fund return, there would be no change in allocation. However, the return would decrease by (.02) ($248,889) = $4978. A decrease to .14 is outside the objective function coefficient range forcing us to resolve the problem. The new formulation and optimal solution is as follows. MAX .14G +.125I +.075M S.T. 1) 2) 3) 4) 5) 6) 7) G + I + M < 800000 .8G -.2I -.2M > 0 .6G -.4I -.4M < 0 -.2G +.8I -.2M > 0 -.5G +.5I -.5M < 0 -.3G -.3I +.7M > 0 .05G + .02I-.04M < 0 OPTIMAL SOLUTION Objective Function Value = Variable -------------G I M 85066.664 Value --------------160000.016 293333.313 346666.688 CP - 14 Reduced Costs -----------------0.000 0.000 0.000 Solutions to Case Problems Constraint -------------1 2 3 4 5 6 7 4. Slack/Surplus --------------0.000 0.000 160000.000 133333.313 106666.688 106666.688 0.000 Dual Prices -----------------0.106 -0.010 0.000 0.000 0.000 0.000 0.833 Since the current optimal solution has more invested in the growth fund than the income fund, adding this requirement will force us to resolve the problem with a new constraint. We should expect a decrease in return as is shown in the following optimal solution. MAX .18G +.125I +.075M S.T. 1) 2) 3) 4) 5) 6) 7) 8) G + I + M < 800000 .8G -.2I -.2M > 0 .6G -.4I -.4M < 0 -.2G +.8I -.2M > 0 -.5G +.5I -.5M < 0 -.3G -.3I +.7M > 0 .05G + .02I-.04M < 0 G - I < 0 OPTIMAL SOLUTION Objective Function Value = 93066.656 Variable -------------G I M Value --------------213333.313 213333.313 373333.313 Reduced Costs -----------------0.000 0.000 0.000 Constraint -------------1 2 3 4 5 6 7 8 Slack/Surplus --------------0.000 53333.324 106666.664 53333.324 186666.656 133333.328 0.000 0.000 Dual Prices -----------------0.116 0.000 0.000 0.000 0.000 0.000 1.033 0.012 Note that the value of the solution has decreased from $94,133 to $93,067. This is only a decrease of 0.2% inyield. Since the yield decrease is so small, Williams may prefer this portfolio for Langford. 5. It is possible a model such as this could be developed for each client. The changed yield estimates would require a change in the objective function coefficients and resolving the problem if the change was outside the objective coefficient range. CP - 15 Chapter 3 Case Problem 3: Truck Leasing Strategy 1. xij = number of trucks obtained from a short term lease signed in month i for a period of j Let months yi = number of trucks obtained from the long-term lease that are used in month i Monthly fuel costs are 20 ($100) = $2000. Monthly Costs for Short-Term Leased Trucks Note: the costs shown here include monthly fuel costs of $2000. Cost $4000 + $2000 = $6000 2 ($3700) + $2000 = $9400 3 ($3225) + $2000 = $11,675 4 ($3040) + $2000 = $14,160 Decision Variables x11, x21, x31, x41 x12, x22, x32 x13, x23 x14 Monthly Costs for Long-Term Leased Trucks Since Reep Construction is committed to the long-term lease and since employees cannot be laid off, the only relevant cost for the long-term leased trucks is the monthly fuel cost of $2000. MIN 6000X11 + 9400X12 + 11675X13 + 14160X14 + 6000X21 + 9400X22 + 11675X23 + 6000X31 + 9400X32 + 6000X41 + 2000Y1 + 2000Y2 + 2000Y3 + 2000Y4 S.T. 1) 2) 3) 4) 5) 6) 7) 8) X11 + X12 X21 + X22 X31 + X32 X41 + X32 Y1 < 1 Y2 < 2 Y3 < 3 Y4 < 1 + + + + X13 X23 X23 X23 + + + + X14 X14 X22 X14 + + + + Y1 X13 X14 Y4 CP - 16 = + + = 10 X12 + Y2 = 12 X13 + Y3 = 14 8 Solutions to Case Problems Objective Function Value = 151660.000 Variable -------------X11 X12 X13 X14 X21 X22 X23 X31 X32 X41 Y1 Y2 Y3 Y4 Value --------------0.000 0.000 3.000 6.000 0.000 0.000 1.000 1.000 0.000 0.000 1.000 2.000 3.000 1.000 Reduced Costs -----------------3515.000 3725.000 0.000 0.000 2810.000 210.000 0.000 0.000 915.000 3515.000 0.000 0.000 0.000 0.000 Constraint -------------1 2 3 4 5 6 7 8 Slack/Surplus --------------0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Dual Prices ------------------2485.000 -3190.000 -6000.000 -2485.000 485.000 1190.000 4000.000 485.000 OBJECTIVE COEFFICIENT RANGES Variable -----------X11 X12 X13 X14 X21 X22 X23 X31 X32 X41 Y1 Y2 Y3 Y4 Lower Limit --------------2485.000 5675.000 10760.000 13950.000 3190.000 9190.000 10485.000 3190.000 8485.000 2485.000 No Lower Limit No Lower Limit No Lower Limit No Lower Limit Current Value --------------6000.000 9400.000 11675.000 14160.000 6000.000 9400.000 11675.000 6000.000 9400.000 6000.000 2000.000 2000.000 2000.000 2000.000 CP - 17 Upper Limit --------------No Upper Limit No Upper Limit 11885.000 15075.000 No Upper Limit No Upper Limit 11885.000 6915.000 No Upper Limit No Upper Limit 2485.000 3190.000 6000.000 2485.000 Chapter 3 RIGHT HAND SIDE RANGES Constraint Lower Limit -------------------------1 4.000 2 11.000 3 13.000 4 2.000 5 0.000 6 1.000 7 0.000 8 0.000 Current Value --------------10.000 12.000 14.000 8.000 1.000 2.000 3.000 1.000 Upper Limit --------------11.000 13.000 No Upper Limit 11.000 7.000 3.000 4.000 7.000 2. The total cost associated with the leasing plan is $151,660. 3. If Reep Construction is willing to consider the possibility of layoffs, we need to include driver costs of $3200 per month. Replacing the coefficients for y1, y2, y3, and y4 in our previous linear program with $5200 and resolving resulted in the following leasing plan: Month Length of Lease (Months) Leased 1 2 3 4 1 0 0 4 6 2 0 0 2 _ 3 0 0 _ _ 4 0 _ _ _ In addition, in month 3, two of the trucks from the long-term leases were used. The total cost of this leasing plan is $165,410. To see what effect a no layoff policy has, we can set y1 = 1, y2 = 2, y3 = 3, y4 = 1 and resolve the linear program using objective coefficients of $5200 for y1, y2, y3, and y4. The new optimal solution forces us to use all the available trucks from the long-term lease; the optimal leasing plan is shown below. Month Length of Lease (Months) Leased 1 2 3 4 1 0 0 3 6 2 0 0 1 _ 3 1 0 _ _ 4 0 _ _ _ The total cost associated with this solution is $174,060. Thus, if Reep maintains their current policy of no layoffs they will incur an additional cost of $174,060 - $165,410 = $8,650. CP - 18 Chapter 4 Linear Programming Applications in Marketing, Finance and Operations Management Case Problem 1: Planning an Advertising Campaign The decision variables are as follows: T1 = number of television advertisements with rating of 90 and 4000 new customers T2 = number of television advertisements with rating of 40 and 1500 new customers R1 = number of radio advertisements with rating of 25 and 2000 new customers R2 = number of radio advertisements with rating of 15 and 1200 new customers N1 = number of newspaper advertisements with rating of 10 and 1000 new customers N2 = number of newspaper advertisements with rating of 5 and 800 new customers The Linear Programming Model and solution using The Management Scientist follow: MAX 90T1+55T2+25R1+20R2+10N1+5N2 S.T. 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 1T1<10 1R1<15 1N1<20 10000T1+10000T2+3000R1+3000R2+1000N1+1000N2<279000 4000T1+1500T2+2000R1+1200R2+1000N1+800N2>100000 -2T1-2T2+1R1+1R2>0 1T1+1T2<20 10000T1+10000T2>140000 3000R1+3000R2<99000 1000N1+1000N2>30000 OPTIMAL SOLUTION Objective Function Value = Variable -------------T1 T2 R1 R2 N1 N2 2160.000 Value --------------10.000 5.000 15.000 18.000 20.000 10.000 CP - 19 Reduced Costs -----------------0.000 0.000 0.000 0.000 0.000 0.000 Chapter 4 Constraint -------------1 2 3 4 5 6 7 8 9 10 Slack/Surplus --------------0.000 0.000 0.000 0.000 27100.000 3.000 5.000 10000.000 0.000 0.000 Dual Prices -----------------35.000 5.000 5.000 0.006 0.000 0.000 0.000 0.000 0.001 0.000 OBJECTIVE COEFFICIENT RANGES Variable -----------T1 T2 R1 R2 N1 N2 Lower Limit --------------55.000 No Lower Limit 20.000 16.500 5.000 No Lower Limit Current Value --------------90.000 55.000 25.000 20.000 10.000 5.000 Upper Limit --------------No Upper Limit 66.667 No Upper Limit 25.000 No Upper Limit 5.500 Current Value --------------10.000 15.000 20.000 279000.000 100000.000 0.000 20.000 140000.000 99000.000 30000.000 Upper Limit --------------15.000 33.000 30.000 294000.000 127100.000 3.000 No Upper Limit 150000.000 109000.000 40000.000 RIGHT HAND SIDE RANGES Constraint -----------1 2 3 4 5 6 7 8 9 10 1. Lower Limit --------------0.000 0.000 0.000 269000.000 No Lower Limit No Lower Limit 15.000 No Lower Limit 93375.000 20000.000 Summary of the Optimal Solution T1 + T2 = 10 + 5 = 15 Television advertisements R1 + R2 = 15 + 18 = 33 Radio advertisements N1 + N2 = 20 + 10 = 30 Newspaper advertisements Advertising Schedule: Media Television Radio Newspaper Totals Number of Ads 15 33 30 78 Total Exposure Rating: Total New Customers Reached: Budget $150,000 99,000 30,000 $279,000 2,160 127,100 (Surplus constraint 5) CP - 20 Solutions to Case Problems 2. The dual price shows that total exposure increases 0.006 points for each one dollar increase in the advertising budget. Right Hand Side Ranges show this dual price applies for a budget increase of up to $294,000 - $279,000 = $15,000. Thus the dual price applies for the $10,000 increase. Total Exposure Rating would increase by 10,000(0.006) = 60 points A $10,000 increase in the advertising budget is a 3.6% increase. But, it only provides a 2.8% increase in total exposure. Management may decide that the additional exposure is not worth the cost. This is a discussion point. 3. The ranges for the exposure rating of 90 for the first 10 television ads show that the solution remains optimal as long as the exposure rating is 55 or higher. This indicates that the solution is not very sensitive to the exposure rating HJ has provided. Indeed, we would draw the same conclusion after reviewing the next four ranges. We could conclude that Flamingo does not have to be concerned about the exact exposure rating. The only concern might be the newspaper exposure rating of 5. A rating of 5.5 or better can be expected to alter the current optimal solution. 4. Remove constraint #5 for the linear programming model and use it to develop the objective function: MAX 4000T1+1500T2+2000R1+1200R2+1000N1+800N2 Solving provides the following Optimal Solution T1 + T2 = 10 + 4 = 14 Television advertisements R1 + R2 = 15 + 13 = 28 Radio advertisements N1 + N2 = 20 + 35 = 55 Newspaper advertisements Advertising Schedule: Media Television Radio Newspaper Totals Number of Ads 14 28 55 97 Total New Customers Reached: Budget $140,000 83,000 55,000 $279,000 139,600 Total Exposure Rating 90(10) + 55(4) + 25(15) + 20(13) + 10(20) + 5(35) = 2130 5. The solution with the objective to maximize the number of potential new customers reached looks attractive. The total number of ads is increased from 78 to 97 (24%) and the number of potential new customers reached is increased by 139,600 – 127,100 = 12,500 (9.8%). Maximizing total exposure may seem to be the preferred objective because it is a more general measure of advertising effectiveness. Exposure includes issues of image, message recall and appeal to repeat customers. However, in this case, many more potential new customers will be reached with the objective of maximizing reach, and the total exposure is only reduced by 2160 – 2130 = 30 points (1.4%). At this point, we would expect some discussion concerning which solution is preferred: the one obtained by maximizing total exposure or the one obtained by maximizing potential new customers reached. Expect students to have differing opinions on the final recommendation. Basically, there are two good media allocation solutions for this problem. CP - 21 Chapter 4 Case Problem 2: Phoenix Computer 1. The monthly cost of a new employee is $2,250 = $27,000/12. The training program is 3 months for a new employee and costs $1,500. So the total cost of hiring a new employee and training her/him to be a laptop specialist is $8,250 = 3($2,250) + $1,500. 2. The training program is only 2 months for current employees and costs $1,000. So a replacement employee will only need to be hired 2 months before the laptop specialist is needed in this case. So the incremental cost of putting a current employee through the training program is $5,500 = 2($2,250) + $1,000. 3. It is clear that 100 new employees will need to be hired either to become laptop specialists themselves or to replace current employees who will become laptop specialists. So the company's monthly payroll cost will increase by $225,000 = 100($2,250) in September over January. A linear program can be formulated and solved to minimize the cost of hiring and training over the period from February to August. The following variable definitions are used: Feb 3 • • • Jun 3 Mar 2 • • • Jul 2 = No. of new employees entering the 3-month program in February = No. of new employees entering the 3-month program in June = No. of current employees entering the 2-month program in March = No. of current employees entering the 2-month program in July IdleMay = No. of trained laptop specialists in excess of those needed in May • • • IdleSep = No. of trained laptop specialists in excess of those needed in September There are 16 constrains needed. The first 5 deal with meeting the needs for laptop specialists. Constraints (6) through (10) restrict the number of current employees that may enter the program while Constraints (11) through (16) deal with the capacity of the training center. The following output from The Management Scientist shows a model and solution that can be used to answer the questions in part 3. LINEAR PROGRAMMING PROBLEM MIN 8250Feb3+5500Mar2+2250IdleMay+8250Mar3+5500Apr2+2250IdleJun+8250Apr3+55 00May 2+2250IdleJul+8250May3+5500Jun2+2250IdleAug+8250Jun3+5500Jul2+2250IdleS ep CP - 22 Solutions to Case Problems S.T. 1) 2) 3) 4) 5) 1Feb3+1Mar2-1IdleMay=20 1Feb3+1Mar2+1Mar3+1Apr2-1IdleJun=30 1Feb3+1Mar2+1Mar3+1Apr2+1Apr3+1May2-1IdleJul=85 1Feb3+1Mar2+1Mar3+1Apr2+1Apr3+1May2+1May3+1Jun2-1IdleAug=85 1Feb3+1Mar2+1Mar3+1Apr2+1Apr3+1May2+1May3+1Jun2+1Jun3+1Jul2-IdleSep =100 6) 1Mar2<15 7) 1Mar2+1Apr2<35 8) 1Mar2+1Apr2+1May2<35 9) 1Mar2+1Apr2+1May2+1Jun2<40 10) 1Mar2+1Apr2+1May2+1Jun2+1Jul2<50 11) 1Feb3<25 12) 1Mar2+1Mar3<25 13) 1Apr2+1Apr3<25 14) 1May2+1May3<25 15) 1Jun2+1Jun3<25 16) 1Jul2<25 OPTIMAL SOLUTION Objective Function Value = Variable -------------Feb3 Mar2 IdleMay Mar3 Apr2 IdleJun Apr3 May2 IdleJul May3 Jun2 IdleAug Jun3 Jul2 IdleSep 698750.000 Value --------------10.000 10.000 0.000 15.000 0.000 5.000 25.000 25.000 0.000 0.000 0.000 0.000 0.000 15.000 0.000 CP - 23 Reduced Costs -----------------0.000 0.000 2250.000 0.000 2250.000 0.000 0.000 0.000 4500.000 2250.000 0.000 2250.000 0.000 0.000 10500.000 Chapter 4 Constraint -------------1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Slack/Surplus --------------0.000 0.000 0.000 0.000 0.000 5.000 25.000 0.000 5.000 0.000 15.000 0.000 0.000 0.000 25.000 10.000 Dual Prices -----------------0.000 2250.000 -2250.000 0.000 -8250.000 0.000 0.000 0.000 0.000 2750.000 0.000 0.000 2250.000 2250.000 0.000 0.000 Case Problem 3: Textile Mill Scheduling Let X3R = Yards of fabric 3 on regular looms X4R = Yards of fabric 4 on regular looms X5R = Yards of fabric 5 on regular looms X1D = Yards of fabric 1 on dobbie looms X2D = Yards of fabric 2 on dobbie looms X3D = Yards of fabric 3 on dobbie looms X4D = Yards of fabric 4 on dobbie looms X5D = Yards of fabric 5 on dobbie looms Y1 = Yards of fabric 1 purchased Y2 = Yards of fabric 2 purchased Y3 = Yards of fabric 3 purchased Y4 = Yards of fabric 4 purchased Y5 = Yards of fabric 5 purchased Profit Contribution per Yard Fabric 1 2 3 4 5 Manufactured 0.33 0.31 0.61 0.73 0.20 Purchased 0.19 0.16 0.50 0.54 0.00 1 2 3 4 5 Regular — — 0.1912 0.1912 0.2398 Dobbie 0.21598 0.21598 0.1912 0.1912 0.2398 Production Times in Hours per Yard Fabric Model may use a Max Profit or Min Cost objective function. CP - 24 Solutions to Case Problems Max 0.61X3R + 0.73X4R + 0.20X5R + 0.33X1D + 0.31X2D + 0.61X3D + 0.73X4D + 0.20X5D + 0.19Y1 + 0.16Y2 + 0.50Y3 + 0.54Y4 or Min 0.49X3R + 0.51X4R + 0.50X5R + 0.66X1D + 0.55X2D + 0.49X3D + 0.51X4D + 0.50X5D + 0.80Y1 + 0.70Y2 + 0.60Y3 + 0.70Y4 + 0.70Y5 Regular Hours Available 30 Looms x 30 days x 24 hours/day = 21600 Dobbie Hours Available 8 Looms x 30 days x 24 hours/day = 5760 Constraints: Regular Looms: 0.192X3R + 0.1912X4R + 0.2398X5R ≤ 21600 Dobbie Looms: 0.21598X1D + 0.21598X2D + 0.1912X3D + 0.1912X4D + 0.2398X5D ≤ 5760 Demand Constraints X1D + Y1 X2D + Y2 X3R + X3D + Y3 X4R + X4D + Y4 X5R + X5D + Y5 = 16500 = 22000 = 62000 = 7500 = 62000 OPTIMAL SOLUTION Objective Function Value = Variable -------------X3R X4R X5R X1D X2D X3D X4D X5D Y1 Y2 Y3 Y4 Y5 62531.91797 Value --------------27711.29297 7500.00000 62000.00000 4669.13672 22000.00000 0.00000 0.00000 0.00000 11830.86328 0.00000 34288.70703 0.00000 0.00000 CP - 25 Reduced Costs -----------------0.00000 0.00000 0.00000 0.00000 0.00000 0.01394 0.01394 0.01748 0.00000 0.01000 0.00000 0.08000 0.06204 Chapter 4 Constraint -------------1 2 3 4 5 6 7 Slack/Surplus --------------0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Dual Prices -----------------0.57531 0.64821 0.19000 0.17000 0.50000 0.62000 0.06204 OBJECTIVE COEFFICIENT RANGES Variable -----------X3R X4R X5R X1D X2D X3D X4D X5D Y1 Y2 Y3 Y4 Y5 Lower Limit --------------0.50000 0.71606 0.18252 0.31426 0.30000 No Lower Limit No Lower Limit No Lower Limit 0.18000 No Lower Limit 0.48606 No Lower Limit No Lower Limit Current Value --------------0.61000 0.73000 0.20000 0.33000 0.31000 0.61000 0.73000 0.20000 0.19000 0.16000 0.50000 0.54000 0.00000 Upper Limit --------------0.62394 No Upper Limit No Upper Limit 0.34000 No Upper Limit 0.62394 0.74394 0.21748 0.20574 0.17000 0.61000 0.62000 0.06204 Current Value --------------21600.00000 5760.00000 16500.00000 22000.00000 62000.00000 7500.00000 62000.00000 Upper Limit --------------28156.00000 8315.23047 No Upper Limit 26669.13672 No Upper Limit 35211.29297 84095.07813 RIGHT HAND SIDE RANGES Constraint -----------1 2 3 4 5 6 7 Lower Limit --------------16301.60059 4751.55957 4669.13672 10169.13672 27711.29297 0.00000 34660.54688 Production/Purchase Schedule (Yards) Regular Looms Fabric 1 2 3 4 5 27711 7500 62000 Projected Profit: $62,531.92 CP - 26 Dobbie Looms 4669 22000 Purchased 11831 34289 Solutions to Case Problems Value of 9th Dobbie Loom Dual Price (Constraint 2) = 0.64821 per hour dobbie Monthly Value of 1 Dobbie Loom (30 days)(24 hours/day)($0.64821) = $466.71 Note: This change is within the Right-Hand Side Ranges for Constraint 2. Discussion of Objective Coefficient Ranges For example, fabric one on the dobbie loom shares ranges of 0.31426 to 0.34 for the profit maximization model or 0.64426 to 0.67 for the cost minimization model. Note here that since demand for the fabrics is fixed, both the profit maximization and cost minimization models will provide the same optimal solution. However, the interpretation of the ranges for the objective function coefficients differ for the two models. In the profit maximization case, the coefficients are profit contributions. Thus, the range information indicates how price per unit and cost per unit may vary simultaneously. That is, as long as the net changes in price per unit and cost per unit keep the profit contributions within the ranges, the solution will remain optimal. In the cost minimization model, the coefficients are costs per unit. Thus, the range information indicates that assuming price per unit remains fixed how much the cost per unit may vary and still maintain the same optimal solution. Case Problem 4: Workforce Scheduling 1. Let tij = number of temporary employees hired under option i (i = 1, 2, 3) in month j (j = 1 for January, j = 2 for February and so on) The following table depicts the decision variables used in this case problem. Option 1 Option 2 Option 3 Jan. t11 t21 t31 Feb. t12 t22 t32 Mar. t13 t23 t33 Apr. t14 t24 t34 May t15 t25 June t16 Costs: Contract cost plus training cost Option 1 2 3 Contract Cost $2000 $4800 $7500 Training Cost $875 $875 $875 Min. 2875(t11 + t12 + t13 + t14 + t15 + t16) + 5675(t21 + t22 + t23 + t24 + t25) + 8375(t31 + t32 + t33 + t34) One constraint is required for each of the six months. CP - 27 Total Cost $2875 $5675 $8375 Chapter 4 Constraint 1: Need 10 additional employees in January t11 = number of temporary employees hired under Option 1 (one-month contract) in January t21 = number of temporary employees hired under Option 2 (two-month contract) in January t31 = number of temporary employees hired under Option 3 (three-month contract) in January t11 + t21 + t31 = 10 Constraint 2: Need 23 additional employees in February t12 , t22 and t32 are the number of temporary employees hired under Options 1, 2 and 3 in February. But, temporary employees hired under Option 2 or Option 3 in January will also be available to satisfy February needs. t21 + t31 + t12 + t22 + t32 = 23 Note: The following table shows the decision variables used in this constraint Jan. Option 1 Option 2 Option 3 t21 t31 Feb. t12 t22 t32 Mar. Apr. May June Constraint 3: Need 19 additional employees in March Option 1 Option 2 Option 3 Jan. Feb. t31 t22 t32 Mar. t13 t23 t33 Apr. May June t31 + t22 + t32 + t13 + t23 + t33 = 19 Constraint 4: Need 26 additional employees in May Jan. Option 1 Option 2 Option 3 Feb. Mar. t32 t23 t33 Apr. t14 t24 t34 May June t32 + t23 + t33 + t14 + t24 + t34 = 26 Constraint 5: Need 20 additional employees in May Jan. Feb. Option 1 Option 2 Option 3 Mar. Apr. t33 t24 t34 t33 + t24 + t34 + t15 + t25 = 20 CP - 28 May t15 t25 June Solutions to Case Problems Constraint 6: Need 14 additional employees in June Jan. Feb. Mar. Option 1 Option 2 Option 3 Apr. May June t16 t25 t34 t34 + t25 + t16 = 14 Optimal Solution: Total Cost = $313,525 Option 1 Option 2 Option 3 Jan. 0 3 7 Feb. 1 0 12 Mar. 0 0 0 Apr. 0 0 14 May 6 0 June 0 2. Option 1 2 3 3. Number Hired 7 3 33 Total: Contract Cost $14,000 $14,400 $247,500 $275,900 Training Cost $6,125 $2,625 $28,875 $37,625 Total Cost $20,125 $17,025 $276,375 $313,525 Hiring 10 full-time employees at the beginning of January will reduce the number of temporary employees needed each month by 10. Using the same linear programming model with the righthand sides of 0, 13, 9, 16, 10 and 4, provides the following schedule for temporary employees: Option 1 Option 2 Option 3 Option 1 2 3 Total: Jan. 0 0 0 Feb. 4 0 9 Number Hired 7 3 13 23 Mar. 0 0 0 Apr. 0 3 4 May 3 0 Contract Cost $14,000 $14,400 $97,500 June 0 Training Cost $6,125 $2,625 $11,375 Total Cost $20,125 $17,025 $108,875 $146,025 Full-time employees cost: Training cost: 10($875) = $8,750 Salary: 10(6)(168)($16.50) = $166,320 Total Cost = $146,025 + $8750 + $166,320 = $321,095 Hiring 10 full-time employees is $321,095 - $313,525 = $7,570 more expensive than using temporary employees. Do not hire the 10 full-time employees. Davis should continue to contract with WorkForce to obtain temporary employees. CP - 29 Chapter 4 4. With the lower training costs, the costs per employee for each option are as follows: Option 1 2 3 Cost $2000 $4800 $7500 Training Cost $700 $700 $700 Total Cost $2700 $5500 $8200 Resolving the original linear programming model with the above costs indicates that Davis should hire all temporary employees on a one-month contract specifically to meet each month's employee needs. Thus, the monthly temporary hire schedule would be as follows: January - 10; February - 23; March - 19; April - 26; May - 20; and June - 14. The total cost of this strategy is $302,400. Note that if training costs were any lower, this would still be the optimal hiring strategy for Davis. Case Problem 5: Duke Energy Coal Allocation A linear programming model can be used to determine how much coal to buy from each of the mining companies and where to ship it. Let xij = tons of coal purchased from supplier i and used by generating unit j The objective function minimizes the total cost to buy and burn coal. The objective function coefficients, cij , are the cost to buy coal at mine i, ship it to generating unit j, and burn it at generating unit j. Thus, the objective function is ∑ ∑ cij xij . In computing the objective function coefficients three inputs must be added: the cost of the coal, the transportation cost to the generating unit, and the cost of processing the coal at the generating unit. There are two types of constraints: supply constraints and demand constraints. The supply constraints limit the amount of coal that can be bought under the various contracts. For the fixedtonnage contracts, the constraints are equalities. For the variable-tonnage contracts, any amount of coal up to a specified maximum may be purchased. Let Li represent the amount that must be purchased under fixed-tonnage contract i and Si represent the maximum amount that can be purchased under variable-tonnage contract i. Then the supply constraints can be written as follows: ∑x ij = Li for all fixed-tonnage contracts ≤ Si for all variable-tonnage contracts j ∑x ij j The demand constraints specify the number of mWh of electricity that must be generated by each generating unit. Let aij = mWh hours of electricity generated by a ton of coal purchased from supplier i and used by generating unit j, and Dj = mWh of electricity demand at generating unit j. The demand constraints can then be written as follows: ∑a x ij ij = Dj for all generating units i Note: Because of the large number of calculations that must be made to compute the objective function and constraint coefficients, we developed an Excel spreadsheet model for this problem. Copies of the data and model worksheets are included after the discussion of the solution to parts (a) through (f). CP - 30 Solutions to Case Problems 1. The number of tons of coal that should be purchased from each of the mining companies and where it should be shipped is shown below: Miami Fort # 5 Miami Fort # 7 Beckjord East Bend Zimmer 0 0 61,538 288,462 0 Peabody 217,105 11,278 71,617 0 0 American 0 0 0 0 275,000 Consol 0 0 33,878 0 166,122 Cyprus 0 0 0 0 0 Addington 0 200,000 0 0 0 Waterloo 0 0 98,673 0 0 RAG The total cost to purchase, deliver, and process the coal is $53,407,243. 2. The cost of the coal in cents per million BTUs for each generating unit is as follows: Miami Fort #5 111.84 3. Miami Fort #7 136.97 Beckjord 127.24 East Bend 103.85 Zimmer 114.51 The average number of BTUs per pound of coal received at each generating unit is shown below: Miami Fort #5 13,300 Miami Fort #7 12,069 Beckjord 12,354 East Bend 13,000 Zimmer 12,468 4. The sensitivity report shows that the shadow price per ton of coal purchased from American Coal Sales is -$13 per ton and the allowable increase is 88,492 tons. This means that every additional ton of coal that Duke Energy can purchase at the current price of $22 per ton will decrease cost by $13. So even paying $30 per ton, Duke Energy will decrease cost by $5 per ton. Thus, they should buy the additional 80,000 tons; doing so will save them $5(80,000) = $400,000. 5. If the energy content of the Cyprus coal turns out to be 13,000 BTUs per ton the procurement plan changes as shown below: Miami Fort # 5 Miami Fort # 7 Beckjord Zimmer 0 0 61,538 288,462 0 Peabody 36,654 191,729 71,617 0 0 American 0 0 0 0 275,000 Consol 0 0 33,878 0 166,122 Cyprus 0 0 85,769 0 0 Addington 200,000 0 0 0 0 Waterloo 0 0 0 0 0 RAG 6. East Bend The shadow prices for the demand constraints are as follows: Miami Fort #5 21 Miami Fort #7 20 Beckjord 20 CP - 31 East Bend 18 Zimmer 19 Chapter 4 The East Bend unit is the least cost producer at the margin ($18 per mWh), and the allowable increase is 160,000 mWh. Thus, Duke Energy should sell the 50,000 mWh over the grid. The additional electricity should be produced at the East Bend generating unit. Duke Energy’s profit will be $12 per mWh. The Excel data and model worksheets used to solve the Duke Energy coal allocation problem are as follows: Duke Energy Coal Allocation Model (Data) Duke Energy Coal Allocation Model (Solution) CP - 32 Chapter 6 Distribution and Network Models Case Problem 1: Solution Plus 1. This case can be formulated as a transportation problem with the Cincinnati and Oakland production facilities as the origins. The locations of the railway stations are the destinations. Each objective function coefficient is the sum of the production cost at an origin and the freight cost to ship from the origin to a destination. The minimum cost solution has a value of $1,318,985 and 773,522 gallons of cleaning fluid are produced and shipped. So, the average cost per gallon of producing the cleaning fluid and shipping it to the railway stations is $1.705168. Shown below is the optimal shipping plan. Santa Ana El Paso Pendleton Houston Kansas City Los Angeles Glendale Jacksonville Little Rock Bridgeport Sacramento Cincinnati 0 6800 39636 100447 24570 0 0 68486 148586 111475 0 500000 Total Oakland 22418 0 40654 0 0 64761 33689 0 0 0 112000 273522 From the shipping plan it can be seen that the Cincinnati plant is at capacity. The dual price for the Cincinnati capacity constraint indicates that if more capacity can be made available the total cost can be decreased by $.11 for every additional gallon up to 40,654. 2. Since it costs Solutions Plus $1.705168 per gallon to produce and deliver the cleaning fluid to the railroad, this is the breakeven point. If Mr. Miller bids any less, Solutions Plus will lose money on the contract. 3. A 15% profit margin corresponds to a price of 1.15($1.705168) = $1.9609432 per gallon. Rounding up, we would recommend the president bid $1.97 per gallon if a 15% profit margin is desired. 4. If management of Solutions Plus expects oil prices to go up, they can also expect freight rates to go up. The breakeven point found by solving the minimum cost transportation problem is using freight rates for the first year. So, management should perhaps bid a bit higher to allow for an increase in freight rates in the second year of the contract. You might offer to resolve the problem assuming, say, a 10% increase in freight rates to see what the breakeven point would be in this situation. On the other hand, if freight rates are expected to go down in the second year, management might want to bid a bit lower to increase the chances of winning the contract. CP - 33 Chapter 6 Solutions Plus Production Cost Santa Ana Origin Cincinnati 10 Oakland 0.22 22418 Demand CP - 34 Santa Ana Origin Cincinnati 11.2 Oakland 1.87 22418 Demand Cincinnati Oakland 1.2 1.65 El Paso Pendleton 0.84 0.83 0.74 0.49 6800 80290 FREIGHT COST Destination Houston Kansas City Los Angeles Glendale Jacksonville Little Rock Bridgeport Sacramento Supply 0.45 0.36 10 10 0.34 0.34 0.34 10 500000 10 10 0.22 0.22 10 10 10 0.15 500000 100447 24570 64761 33689 68486 148586 111475 112000 El Paso Pendleton 2.04 2.03 2.39 2.14 6800 80290 FREIGHT + PRODUCTION COST Destination Houston Kansas City Los Angeles Glendale Jacksonville Little Rock Bridgeport Sacramento Supply 1.65 1.56 11.2 11.2 1.54 1.54 1.54 11.2 500000 11.65 11.65 1.87 1.87 11.65 11.65 11.65 1.8 500000 100447 24570 64761 33689 68486 148586 111475 112000 Model Total Per Gallon Min Cost 1318984.93 1.705167959 Santa Ana Origin 0 Cincinnati Oakland 22418 22418 Total = 22418 El Paso Pendleton 6800 39636 0 40654 6800 80290 = = 6800 80290 Destination Houston Kansas City Los Angeles Glendale Jacksonville Little Rock Bridgeport Sacramento 100447 24570 0 0 68486 148586 111475 0 0 0 64761 33689 0 0 0 112000 100447 24570 64761 33689 68486 148586 111475 112000 = = = = = = = = 100447 24570 64761 33689 68486 148586 111475 112000 Total 500000 <= 500000 273522 <= 500000 773522 Solutions to Case Problems Case Problem 2: Distribution Systems Design Three related linear programming models were developed and solved to answer the questions in this case. First, we developed a linear programming formulation of the network model shown on the following page; we refer to this network model and the corresponding linear program as the Part 2 Network Model and the Part 2 Linear Program, respectively. Variations of the Part 2 Linear Program were then used to answer the questions in parts 1 and 3. For instance, to answer the question concerning the existing system in part 1 we added constraints to the Part 2 Linear Program which set the flow equal to 0 over the distribution centercustomer arcs that Darby does not currently use. For Part 3, we added three plant to customer arcs to the Part 2 Network Model and corresponding linear program: El Paso - San Antonio, San Bernardino - Los Angeles, and San Bernardino - San Diego. The decision variables used in each of the linear programs use 4 letters to describe the amount of flow over each arc. The first two letters in each variable name identify the “from” node and the second two letters identify the “to” nodes. For instance, EPFW represents the amount shipped from El Paso to Ft. Worth and LVDE represents the amount shipped from Las Vegas to Denver. A description of the LP models that provide the basis for answering the questions in the managerial report follows the network model for the questions in part 2. CP - 35 Chapter 6 Part 2 Network Model: Dallas 6300 San Antonio 4880 Wichita 2130 Kansas City 1210 Denver 6120 Salt Lake City 4830 Phoenix 2750 0.3 2.1 Ft. Worth 3.1 4.4 6.0 3.20 30,000 5.4 El Paso 2.20 4.5 5.2 6.0 4.20 2.7 Santa Fe 4.7 3.90 20,000 3.4 3.3 2.7 San Bernardino 1.20 5.4 3.3 Las Vegas 2.4 2.1 2.5 Los Angeles San Diego CP - 36 8580 4460 Solutions to Case Problems LP Model and Solution for Part 2 MIN 13.7EPFW + 12.7EPSF + 14.7EPLV + 13.9SBSF + 11.2SBLV + .3FWDA + 2.1FWSA + 3.1FWWI + 4.4FWKC + 6.0FWDE + 5.2SFDA + 5.4SFSA + 4.5SFWI + 6.0SFKC + 2.7SFDE + 4.7SFSL + 3.4SFPH + 3.3SFLA + 2.7SFSD + 5.4LVDE + 3.3LVSL + 2.4LVPH + 2.1LVLA + 2.5LVSD S.T. 1) 2) 3) 4) EPFW SBSF FWDA SFDA + + + + 5) 6) 7) 8) 9) 10) 11) 12) 13) 14) LVDE FWDA FWSA FWWI FWKC FWDE SFSL SFPH SFLA SFSD + + + + + + + + + + SFSD EPSF SBLV FWSA SFSA LVSL SFDA SFSA SFWI SFKC SFDE LVSL LVPH LVLA LVSD + EPLV < 30000 < 20000 + FWWI + FWKC + FWDE - EPFW = 0 + SFWI + SFKC + SFDE + SFSL + SFPH + SFLA + EPSF - SBSF = 0 + LVPH + LVLA + LVSD - EPLV - SBLV = 0 = 6300 = 4880 = 2130 = 1210 + LVDE = 6120 = 4830 = 2750 = 8580 = 4460 Objective Function Value = Variable -------------EPFW EPSF EPLV SBSF SBLV FWDA FWSA FWWI FWKC FWDE SFDA SFSA SFWI SFKC SFDE SFSL SFPH SFLA SFSD LVDE LVSL 600942.000 Value --------------14520.000 6740.000 0.000 0.000 20000.000 6300.000 4880.000 2130.000 1210.000 0.000 0.000 0.000 0.000 0.000 6120.000 0.000 0.000 0.000 620.000 0.000 4830.000 CP - 37 Reduced Costs -----------------0.000 0.000 1.800 2.900 0.000 0.000 0.000 0.000 0.000 4.300 3.900 2.300 0.400 0.600 0.000 1.200 0.800 1.000 0.000 2.900 0.000 Chapter 6 Variable -------------LVPH LVLA LVSD Value --------------2750.000 8580.000 3840.000 Reduced Costs -----------------0.000 0.000 0.000 LP Model and Solution for Part 1 Questions (Existing System) The following constraints were added to the above LP Model so that each customer can only be served by the distribution center it is currently served by. 15) 16) 17) 18) 19) 20) 21) 22) 23) 24) SFDA SFSA SFWI SFKC SFLA SFSD LVDE FWDE LVSL LVPH = = = = = = = = = = 0 0 0 0 0 0 0 0 0 0 The optimal solution shows that the total cost (manufacturing plus distribution) of the existing system is $620,770. Details of this solution and the modifications suggested in questions 2 and 3 are contained in the “Summary of Optimal Solutions.” LP Model and Solution for Part 3 (Plant-Customer Shipments) The network model for the part 2 questions must be modified by adding 3 arcs: San Bernardino - Los Angeles, San Bernardino - San Diego, and El Paso - San Antonio. The corresponding linear program must also be modified to incorporate the new variables. CP - 38 Solutions to Case Problems A summary of the optimal solutions suggested by questions (1), (2), and (3) follows. Summary of Optimal Solutions Part 1 Part 2 Costs Existing System Any Distribution Center Part 3 Plant Customer Shipments Manufacturing Cost $426,710 $423,230 $423,230 194,060 177,712 130,304 $620,770 $600,942 $553,534 EPFW 14520 14520 9640 EPSF 13700 6740 6740 EPLV 0 0 0 SBSF 0 0 0 SBLV 13040 20000 6960 FWDA 6300 6300 6300 FWSA 4880 4880 0 FWWI 2130 2130 2130 FWKC 1210 1210 1210 FWDE 0 0 0 SFDA 0 0 0 SFSA 0 0 0 SFWI 0 0 0 SFKC 0 0 0 SFDE 6120 6120 6120 SFSL 4830 0 0 SFPH 2750 0 620 SFLA 0 0 0 SFSD 0 620 0 LVDE 0 0 0 LVSL 0 4830 4830 LVPH 0 2750 2130 Shipping Cost Total Cost Decision Variables CP - 39 Chapter 6 Part 1 Part 2 Existing System Any Distribution Center Part 3 Plant Customer Shipments LVLA 8580 8580 0 LVSD 4460 3840 0 EPSA 4880 SBLA 8580 SBSD 4460 Discussion 1. With the current system of assigning customers to distribution centers, the minimum total cost is $620,770; the total cost is the sum of $426,710 for manufacturing and $194,060 for shipping. One might suspect that restricting customers to a single distribution center is not optimal. 2. Allowing customers to be serviced by any distribution center reduces total cost to $600,942; the total cost consists of $423,230 of manufacturing cost and $177,712 of shipping cost. This is a 3.19% decrease in total cost and an 8.42% decrease in shipping cost. Only one customer, San Diego, is served by more than one distribution center. 3. Allowing some direct shipments from plants to customers causes a substantial reduction in total cost. The total cost is $553,534, a 10.83% reduction over the current plan. Note, however, there is a 32.85% reduction in shipping cost. Again, one customer, Phoenix, receives shipments from 2 distribution centers. 4. Total capacity at the two plants (50,000 units) is adequate to handle the growth in demand. However, the company might want to consider expanding the more efficient San Bernardino plant. The dual price from Part 3 output (not shown here) indicates that each unit of added capacity will reduce distribution costs by $2.50. However, the range of feasibility shows this dual price is only applicable for the net 620 units of capacity. Additional computer runs with higher capacity levels for San Bernardino are recommended. CP - 40 Chapter 7 Integer Linear Programming Case Problem 1: Textbook Publishing An integer programming model can be used advantageously to assist in developing recommendations. Let x i = book i is scheduled for publication { 10 ifotherwise The subscripts correspond to the books as follows: Book Business Calculus Finite Math General Statistics Mathematical Statistics Business Statistics Finance Financial Accounting Managerial Accounting English Literature German i 1 2 3 4 5 6 7 8 9 10 An integer programming model for maximizing projected sales (thousands of units) subject to the restrictions mentioned is given. Max 20x1 + 30x2 + 15x3 s.t. 30x1 + 16x2 + 24x3 40x1 + 24x2 30x3 x3 x1 + x2 + 10x4 + 25x5 + + 20x4 + 10x5 + 24x4 + x4 + + 18x6 + 25x7 + 50x8 + 20x9 + 30x10 + 40x9 ≤ 60 John + 24x7 + 28x8 + 34x9 + 50x10 ≤ 40 Susan 16x5 + 14x6 + 26x7 + 30x8 + 30x9 + 36x10 ≤ 40 Monica x5 ≤ 2 No. of Stat Books x7 + x8 ≤ 1 Account Book = 1 Math Book xi = 0, 1 for all i The optimal solution (x2 = x5 = x6 = 1) calls for publishing the finite math, the business statistics and the finance books. Projected sales are 73,000 copies. (1) If Susan can be made available for another 12 days, the optimal solution is x2 = x8 = 1. This calls for publishing the finite math and managerial accounting texts for projected sales of 80,000 copies. (2) If Monica is also available for 10 more days, a big improvement can be made. The new optimal solution calls for producing the finite math book, the business statistics book, and the managerial accounting book. Projected sales are 105,000 copies. CP - 41 Chapter 7 (3) The solution in (2) above does not include any new books. In the long run this would appear to be a bad strategy for the company. A variety of modifications can be made to the model to examine the short run impact of postponing a revision. For instance, a constraint could be added to require publication of at least one new book. Case Problem 2: Yeager National Bank A mixed integer linear programming (MILP) model can be used advantageously to assist in preparing a report for Mr. Wolff. Since the annual fixed costs of operating the lockboxes are not known exactly we formulate a model that can be used without that information. Then determine if information more precise than the $20,000 - $30,000 estimate is needed to make a final decision. We formulate a model that can be solved to find the best set of locations with a given number of lockboxes. It is then solved 4 times to find the best set of locations with 1 lockbox, up to 2 lockboxes, up to 3 lockboxes, and up to 4 lockboxes. We then can address the question of whether annual operating cost information is needed from some or all of the potential lockbox locations. We use the following variable definitions: P = 1 if a lockbox is in Phoenix; 0 otherwise S = 1 if a lockbox is in Salt Lake City; 0 otherwise A = 1 if a lockbox is in Atlanta; 0 otherwise B = 1 if a lockbox is in Boston; 0 otherwise PNW = 1 if the Northwest region is assigned to Phoenix; 0 otherwise PSW = 1 if the Southwest region is assigned to Phoenix; 0 otherwise . . . BSE = 1 if the Southeast region is assigned to Boston; 0 otherwise The MILP model has 24 variables and 26 constraints. The objective function calls for minimizing lost interest income. To see how the objective function coefficients are computed, consider assigning the Northwest region to Phoenix. Daily collections are $80,000 and it takes 4 days to receive and process payments. At a 15% rate Yeager could save $48,000 = (.15)(4)($80,000) annually over this assignment if the Northwest collections could be credited to their account instantaneously. A similar calculation is made for all the other potential assignments. The objective function coefficients for P, S, A, and B are zero since we are not including the cost of operating the lockboxes at this stage. Constraints (1) to (5) ensure that each region is assigned a lock box. Constraints (6) to (25) ensure that a region is only assigned to a lockbox location that is open. And, constraint (26) is a limitation on the number of lockbox locations that may be chosen. The right-hand-side of that constraint will be varied from 1 to 4 as the problem is solved 4 times. Shown below is the MILP model that is solved to find the optimal solution when up to 2 lockboxes may be used. CP - 42 Solutions to Case Problems INTEGER LINEAR PROGRAMMING PROBLEM MIN 48PNW+27PSW+112.5PCE+135PNE+60PSE+24SNW+40.5SSW+67.5SCE+108SNE+90SSE+48 ANW+5 4ASW+67.5ACE+81ANE+30ASE+48BNW+81BSW+90BCE+54BNE+45BSE S.T. 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13) 14) 15) 16) 17) 18) 19) 20) 21) 22) 23) 24) 25) 26) 1PNW+1SNW+1ANW+1BNW=1 +1PSW+1SSW+1ASW+1BSW=1 +1PCE+1SCE+1ACE+1BCE=1 +1PNE+1SNE+1ANE+1BNE=1 +1PSE+1SSE+1ASE+1BSE=1 1PNW-1P<0 +1PSW-1P<0 +1PCE-1P<0 +1PNE-1P<0 +1PSE-1P<0 +1SNW-1S<0 +1SSW-1S<0 +1SCE-1S<0 +1SNE-1S<0 +1SSE-1S<0 +1ANW-1A<0 +1ASW-1A<0 +1ACE-1A<0 +1ANE-1A<0 +1ASE-1A<0 +1BNW-1B<0 +1BSW-1B<0 +1BCE-1B<0 +1BNE-1B<0 +1BSE-1B<0 +1P+1S+1A+1B<2 OPTIMAL SOLUTION Objective Function Value = 231.000 CP - 43 Chapter 7 Variable -------------PNW PSW PCE PNE PSE SNW SSW SCE SNE SSE ANW ASW ACE ANE ASE BNW BSW BCE BNE BSE P S A B Value --------------0.000 0.000 0.000 0.000 0.000 1.000 1.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 0.000 1.000 0.000 1.000 Constraint -------------1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Slack/Surplus --------------0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 1.000 0.000 0.000 0.000 CP - 44 Solutions to Case Problems With 2 lockbox locations, the lost interest income is $231,000. Resolving with values of 1, 3, and 4 on the right-hand-side of constraint 26 provides the following: No. of Lockboxes 1 2 3 4 Locations Atlanta Salt Lake City & Boston Salt Lake City, Atlanta & Boston Salt Lake City, Phoenix, Atlanta & Boston Lost Interest Income ($) 280,500 231,000 216,000 202,500 From this information, it is clear that if the fixed cost per lockbox location is anywhere between $20,000 and $30,000 two locations (Salt Lake City & Boston) should be chosen. Mr. Wolff can be told that he only needs to collect information and negotiate on the cost to choose a site in those 2 locations. Case Problem 3: Production Scheduling with Changeover Costs A mixed integer programming model can be used advantageously to assist in developing recommendations. We describe such a model here; it has 48 decision variables and 64 constraints. We show here how to use Microsoft Excel to formulate and solve the problem. The spreadsheet at the end shows how we set up the problem and the optimal solution. We describe the model now. Variables There are variables for production, inventory, setup, and changeover in each week. Pi = number of P-Heads produced in week i Hi = number of H-Heads produced in week i IPi = number of P-Heads in inventory at end of week i IHi = number of H-Heads in inventory at end of week i SPi = 1 if line is setup for P in week i, 0 if setup for H Changei = 1 if a changeover occurs at the beginning of week i, 0 otherwise Constraints There are constraints for production capacity, inventory balance, maintenance of safety stock, and enforcement of changeovers. Also, Excel requires that you identify the 0-1 (binary) variables in the Solver dialog box. The constraints as specified in the Excel Solver dialog box are as follows (references are to cells of the spreadsheet): B20:C27 ≤ B34:C41 production capacity, or nonnegativity of slack G20:G27 ≥ H34:H41 G20:G27 ≥ I34:I41 forces Changei to 1 when a changeover occurs D20:E27 ≥ D6:E13 ending inventory ≥ safety stock D34:E41 = B6:C13 beg. inv. + production - end inv. = demand F18:F25 = Bin setup variables must be binary CP - 45 Chapter 7 Note that even though the Changei variable must also be integer it is not necessary to require it because minimization will never let it be any bigger than it has to be. And, the second set of constraints force it to be ≥ 1 whenever the setup variable changes from 1 to 0 or from 0 to 1. Objective We want to minimize total cost which is represented by cell J23 in the spreadsheet. It is the sum of production cost, inventory cost, and changeover cost. The Spreadsheet The first 14 rows of the spreadsheet contain the data for the problem; information on demand, safety stock, various costs and beginning inventories are given. Cells B20:G27, as shown contain the optimal solution to the problem. Before solving, those cells were empty. The spreadsheet formulation and solution are shown. A 1 2 3 B C D Product Demand P H P E 4 5 Week 6 7 8 1 2 3 55 55 44 38 38 30 44 35.2 0 9 10 4 5 6 7 8 9 0 45 45 36 35 35 0 48 48 58 57 58 36 36 28.8 28 28 11 12 13 14 F G H I J Production Scheduling Safety Stock P H H Production Cost 225 310 30.4 24 0 Max Weekly Rate Changeover Cost Weekly Inv. Rate 100 500 0.00375 80 500 0.00375 38.4 38.4 46.4 45.6 46.4 Weekly Inv. Cost Beginning Inv. 0.84375 125 1.1625 143 15 16 17 18 19 Model Week P H Inv. P Inv. H Setup P Changeover 20 1 18.00 0.00 88.00 105.00 1.00 0.00 Prod. Cost 117374 21 22 23 2 3 4 100.00 100.00 0.00 0.00 0.00 1.40 133.00 189.00 189.00 67.00 37.00 38.40 1.00 1.00 0.00 0.00 0.00 1.00 Inv. Cost Changeover Cost Min Total Cost 1280.35125 500 119154.3513 24 25 5 6 0.00 0.00 48.00 56.00 144.00 99.00 38.40 46.40 0.00 0.00 0.00 0.00 26 27 7 8 0.00 0.00 57.20 57.80 63.00 28.00 45.60 46.40 0.00 0.00 0.00 0.00 28 29 30 31 32 33 Inventory Balance Production Capacity P H 100 100 Beginning Inv. + Prod. - Ending Inv. P H 0 55 0 55 34 35 Week 1 2 36 37 38 3 4 5 100 0 0 0 80 80 39 40 6 7 0 0 41 8 0 Changeover Def. To P if 1 To H if 1 0.00 0.00 0.00 0.00 38 38 Week 1 2 44 0 45 30 0 48 3 4 5 0.00 -1.00 0.00 0.00 1.00 0.00 80 80 45 36 48 58 6 7 0.00 0.00 0.00 0.00 80 35 57 8 0.00 0.00 CP - 46 Solutions to Case Problems Solution Comments The spreadsheet contains the optimal solution. The minimum total cost is $119,154.35. The components of that cost are production: $117,374, inventory: $1280.35 and changeover: $500. From cells F20:F27 we see that the line will be setup to produce P-Heads in weeks 1-3 and H-Heads in weeks 4-8. Cell G23 shows that there will be a changeover from producing P-Heads to H-Heads at the beginning of week 4. By adjusting the data for this problem (e.g. beginning inventories and the various costs) a number of variations of this problem can be created with the same basic model. Also, one might want to vary the safety stock requirements and the number of weeks in the planning horizon to create other variations of the problem. CP - 47 Chapter 8 Nonlinear Optimization Models Case Problem: Portfolio Optimization with Transaction Costs 1. If $41,268.51are spent purchasing the Intermediate-Term Bond fund, and the transaction cost is 1 percent (i.e. one cent on the dollar), then the transaction fee is .01 ($41, 268.51) = $412.6851 2. The total dollar amount spent on transaction costs is $1090.311. 3. After rebalancing, Ms. Delgado has $100,000 - $1090.311 = $98,909.689 4. To calculate the expected amount in the Intermediate-Term bond fund at year end, figure out the year-end amount in each scenario and then weight each scenario by a factor of 1/5 = .2. Since Ms. Delgado started with $51,268.51 after rebalancing, this gives $51,268.51 + (.2)(.1764)( $51,268.51) + (.2)(.0325)( $51,268.51) + (.2)(.0751)( $51,268.51) - (.2)(.0133)( $51,268.51) +(.2)(.0736)( $51,268.51) = $51,268.51 + $3530.35 = $54,798.86 5. From the LINGO solution we see that Ms. Delgado can expect an average return of $10,000 on her investment. After rebalancing (see question 3) Ms. Delgado has $98,909.689 in her portfolio to start the year, so she is getting a return of greater than 10 percent on the amount of money she starts with after rebalancing! However, this is not what she wants or expects. Since she starts the year with $98,909.689 after rebalancing and earns (in expectation) $10,000 she can expect $98,909.689 + $10,000 = $108,909.689 in her portfolio. However, this does not give a return of 10 percent on her original portfolio of $100,000 which is what she expected. In order to end up with a return of at least 10 percent on her original portfolio she must have at least $110,000.00 in her portfolio at the end of the year. 6. Here is the formulation that provides an expected value at least $110,000 at year end. MIN = (1/5)*((R1 - RBAR)^2 + (R2 - RBAR)^2 + (R3 - RBAR)^2 + (R4 RBAR)^2 + (R5 - RBAR)^2); ! THE SCENARIO RETURNS; 1.1006*FS + 1.1764*IB + 1.3241*LG + 1.3236*LV + 1.3344*SG + 1.2456*SV = R1; 1.1312*FS + 1.0325*IB + 1.1871*LG + 1.2061*LV + 1.1940*SG + 1.2532*SV = R2; 1.1347*FS + 1.0751*IB + 1.3328*LG + 1.1293*LV + 1.0385*SG + (1.0670)*SV = R3; 1.4542*FS + (1 - .0133)*IB + 1.4146*LG + 1.0706*LV + 1.5868*SG + 1.0543*SV = R4; (1-.2193)*FS + 1.0736*IB + (1- .2326)*LG + (1- .0537)*LV + (1.0902)*SG + 1.1731*SV = R5; ! PORTFOLIO AVERAGE RETURN; (1/5)*(R1 + R2 + R3+ R4 + R5) = RBAR; ! UNITY CONSTRAINT; FS + IB + LG + LV + SG + SV + TRANS_COST = 100000; CP - 48 Solutions to Case Problems RBAR > RMIN; RMIN = 10000 + 100000; ! DEFINE BUY AND SELL QUANTITIES; FS = FS_START + FS_BUY - FS_SELL; IB = IB_START + IB_BUY - IB_SELL; LG = LG_START + LG_BUY - LG_SELL; LV = LV_START + LV_BUY - LV_SELL; SG = SG_START + SG_BUY - SG_SELL; SV = SV_START + SV_BUY - SV_SELL; ! DEFINE TOTAL TRANSACTION COSTS; TRANS_COST = TRANS_FEE*(FS_BUY + FS_SELL + IB_BUY + IB_SELL + LG_BUY + LG_SELL + LV_BUY + LV_SELL + SG_BUY + SG_SELL + SV_BUY + SV_SELL); FS_START = 10000; IB_START = 10000; LG_START = 10000; LV_START = 40000; SG_START = 10000; SV_START = 20000; TRANS_FEE = 0.01; @FREE(R1); @FREE(R2); @FREE(R3); @FREE(R4); @FREE(R5); CP - 49 Chapter 6 7. Here is the solution for the formulation in Question 7. Local optimal solution found. Objective value: Total solver iterations: Variable R1 RBAR R2 R3 R4 R5 FS IB LG LV SG SV TRANS_COST RMIN FS_START FS_BUY FS_SELL IB_START IB_BUY IB_SELL LG_START LG_BUY LG_SELL LV_START LV_BUY LV_SELL SG_START SG_BUY SG_SELL SV_START SV_BUY SV_SELL TRANS_FEE 0.3380209E+08 46 Value 119713.7 110000.0 112235.3 105190.3 109673.8 103186.9 10000.00 44769.45 10870.66 0.000000 719.8605 32664.18 975.8443 110000.0 10000.00 0.000000 0.000000 10000.00 34769.45 0.000000 10000.00 870.6613 0.000000 40000.00 0.000000 40000.00 10000.00 0.000000 9280.140 20000.00 12664.18 0.000000 0.1000000E-01 CP - 50 Reduced Cost 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 327.7706 0.000000 0.000000 0.000000 0.000000 0.000000 41.03809 106.0720 0.000000 0.000000 147.1100 0.000000 0.000000 147.1100 0.000000 147.1100 0.000000 0.000000 147.1100 0.000000 0.000000 0.000000 147.1100 0.000000 Chapter 9 Project Scheduling: PERT/CPM Case Problem: R.C. Coleman 1. R.C. Coleman's Project Network D A Start C B E F G Activity A B C D E F G H I J K Activity A B C D E F G H I J K I Earliest Start 0 0 9 13 13 23 13 29 29 35 39 K H J Expected Time 6 9 4 12 10 6 8 6 7 4 4 Latest Start 3 0 9 17 13 23 21 29 32 35 39 Earliest Finish 6 9 13 25 23 29 21 35 36 39 43 CP - 51 Finish Variance 0.44 2.78 0.44 7.11 1.00 0.44 7.11 0.44 2.78 0.11 0.44 Latest Finish 9 9 13 29 23 29 29 35 39 39 43 Slack 3 0 0 4 0 0 8 0 3 0 0 Critical Activity Yes Yes Yes Yes Yes Yes Yes Chapter 9 The expected project completion time is 43 weeks. The critical path activities are B-C-E-F-H-J-K. The variance of the critical path is 5.67. z= 40 − 43 5.67 = −1.26 Area = 0.3962 P(T ≤ 40) = 0.5000 - 0.3962 = 0.1038 Given the above calculations, we can conclude that there is about a 10% chance that the project can be completed in 40 weeks or less. Coleman should consider crashing project activities. 2. 20% 80% Planned project completion time Desires 40-week completion time For 80% chance, z = +0.84 Thus 40 − E (T ) 5.67 = 0.84 Solve for E(T) = 38 weeks. R.C. Coleman should crash activities to reduce the expected project completion time to 38 weeks. 3. In this section, we will use expected activity times as normal times and use a linear programming model based on expected times to make the crashing decisions. Let xi = the completion time for activity i yi = the amount of crash time for activity i CP - 52 Solutions to Case Problems Min 450yA + 400yB + 600yC + 300yD + 1000yE + 550yF + 750yG + 700yH + 800yI + 400yJ + 500yK s.t. xA + y A ≥ 6 xK ≤ 38 xB + yB ≥ 9 yA ≤ 2 xC + yC – xA ≥ 4 yB ≤ 2 xC + yC – xB ≥ 4 yC ≤ 2 xD + yD – xC ≥ 12 yD ≤ 4 xE + yE – xC ≥ 10 yE ≤ 3 xF + yF – xE ≥ 6 yF ≤ 2 xG + yG – xC ≥ 8 yG ≤ 3 xH + yH – xF ≥ 6 yH ≤ 2 xH + yH – xG ≥ 6 yI ≤ 3 xI + y I – xD ≥ 7 yJ ≤ 1 xI + y I – xF ≥ 7 yK ≤ 1 xJ + yJ – xH ≥ 4 All xi,yi ≥ 0 xK + yK – xI ≥ 4 xK + yK – xJ ≥ 4 The optimal crashing decisions are as follows: Crash Activity B F J K Weeks 2 1 1 1 Total CP - 53 Cost 800 550 400 500 2250 Chapter 9 A revised activity schedule based on these crashing decisions is as follows: Activity A B C D E F G H I J K Earliest Start 0 0 7 11 11 21 11 26 26 32 35 Latest Start 1 0 7 14 11 21 18 26 28 32 35 Earliest Finish 6 7 11 23 21 26 19 32 33 35 38 Latest Finish 7 7 11 26 21 26 26 32 35 35 38 Slack 1 0 0 3 0 0 7 0 2 0 0 Critical Activity Yes Yes Yes Yes Yes Yes Yes The student should comment on the fact that the crashing decisions may alter the variance in the project completion time. By defining revised optimistic, most probable, and pessimistic times for crashed activities B, F, J, and K, a revised variance in the project completion time can be found. Using this result, a revised probability of a 40-week completion time can be computed. CP - 54 Chapter 10 Inventory Models Case Problem 1: Wagner Fabricating Company 1. 2. Holding Cost Cost of capital Taxes/Insurance (24,000/600,000) Shrinkage (9,000/600,000) Warehouse overhead (15,000/600,000) Annual rate Ordering Cost 2 hours at $28.00 Other expenses (2,375/125) Cost per order 3. 14.0% 4.0% 1.5% 2.5% 22.0% $56.00 19.00 $75.00 Set-up Cost 8 Hours at $50.00 $400 per set-up 4. & 5. a. Order from Supplier - EOQ model Ch = IC = 0.22 ($18.00) = $3.96 Q* = 2 DCo = Ch 2(3200)75 = 348.16 units 3.96 Number of orders = D/Q = 9.19/year Cycle time = 250(Q) / D = 250(348.16) / 3200 = 27.2 days Reorder Point: P(Stockout) = 1 / 9.19 = 0.1088 Stockout 64 CP - 55 z = 1.24 Chapter 10 r = 64 + 1.24(10) = 76.4 Safety stock = 76.4 - 64 = 12.4 Maximum inventory = Q + 12.4 = 360.56 Average inventory = Q/2 + 12.4 = 186.48 Annual holding cost = 186.48(3.96) = $738.46 Annual ordering cost = 9.19(75) = $689.35 Purchase cost = 3200($18) = $57,600 Total annual cost = $59,027.81 b. Manufacture - Production lot size model Ch = IC = 0.22($17.00) = $3.74 P = 1000(12) = 12,000/year Note: The five-month capacity of 5,000 units is sufficient to handle annual demand of 3,200 units. Q* = 2 DCo = (1 − D / P )Ch 2(3200)(400) = 966.13 (1 − 3200 /1200)3.74 Number of production runs = D/Q = 3.31/year Cycle Time = 250(Q) / D = 250(966.13) / 3200 = 75.48 days Reorder point: P(Stockout) = 1 / 3.31 = 0.3021 Stockout 128 r = 128 + 0.52(20) = 138.4 Safety stock = 138.4 - 128 = 10.4 Maximum inventory = (1 - 3200/12000)966.13 + 10.4 = 718.89 Annual holding cost = (354.25 + 10.4)(3.74) = $1363.79 Annual set up cost = 3.31(400) = $1363.79 Manufacturing cost = 3200($17) = $54,400 Total Annual Cost = $57,088.67 CP - 56 z = 0.52 Solutions to Case Problems 6. Recommend manufacturing the part Savings: $59,027.81 - 57,088.67 = $1,939.14 (3.3%) Case Problem 2: River City Fire Department 1. During a three-week scheduling period, a unit is scheduled seven days and must have at least 186 firefighters on duty each day. Thus, each unit must provide staffing for 7 × 186 = 1302 firefighter days During a three-week scheduling period, each firefighter provides six days of service. Thus, the base number of firefighters per unit must be 1302 = 217 firefighters 6 2. The single-period inventory model can be used to determine the number of additional firefighters to be added to the unit to cover daily absences. Let Q* = minimum cost number of additional firefighters If Demand < Q*, Q* has overestimated demand. The cost of overestimating demand is the daily wage rate (d) for a firefighter. co = d If Demand > Q*, Q* has underestimated demand and overtime is needed. The cost of underestimating demand is the overtime wage (1.55d) minus the wage rate (d) that would have been required if we had hired enough additional firefighters. cu = 1.55d - d = 0.55d Using the single period inventory model, P(Demand < Q*) = cu 0.55d 0.55 = = = 0.3548 cu + co 0.55 + d 1.55 Using the normal distribution, area = 0.5000 - 0.3548 = 0.1452. z = -0.37 Q* = µ + zσ = 20 − 0.37(5) = 18.14 Use 18 additional firefighters. CP - 57 Chapter 10 3. From part 2, this is the probability demand will exceed 18.14 is 1 - 0.3548 = 0.6452. This is an acceptable approximation of the probability overtime will be needed. A more complete analysis would include the fact that we have rounded the number of firefighters needed to 18. Then allowing for continuity correction, overtime would only be needed if demand exceeded 18.5 firefighters. Thus a revised estimate of the probability of overtime is: z= 18.5 − 20 = −.30 Area = .1179 5 P(Overtime) = .5000 + .1179 = .6179 4. Firefighters for each unit 217 + 18 additional = 235 Firefighters for the department (3 units) 3 × 235 = 705 firefighters CP - 58 Chapter 11 Waiting Line Models Case Problem 1: Regional Airlines The analysis that follows is based upon the assumptions of Poisson arrivals and exponential service times. With one call every 3.75 minutes, we have an average arrival rate of λ = 60/3.75 = 16 calls per hour Similarly, with an average service time of 3 minutes, we have a service rate of µ = 60/3 = 20 calls per hour 1. Current System with no waiting allowed With 1 reservation agent: P0 = (λ / µ )0 / 0! k ∑ (λ / µ ) i = (16 / 20)0 / 0! = .5556 (16 / 20)0 / 0!+ (16 / 20)1 /1! = (16 / 20)1 /1! = .4444 (16 / 20)0 / 0!+ (16 / 20)1 /1! / i! i =0 P1 = (λ / µ )1 /1! k ∑ (λ / µ ) 1 / i! i =1 The probability a caller will get a busy signal and be blocked is P1 = .4444. With 2 reservation agents: P0 = (λ / µ )0 / 0! k ∑ (λ / µ ) i = (16 / 20)0 / 0! = .4717 (16 / 20) / 0!+ (16 / 20)1 /1!+ (16 / 20)2 / 2! = (16 / 20)1 /1! = .3774 (16 / 20) / 0!+ (16 / 20)1 /1!+ (16 / 20)2 / 2! = (16 / 20) 2 / 2! = .1509 (16 / 20) / 0!+ (16 / 20)1 /1!+ (16 / 20)2 / 2! / i! 0 i =0 P1 = (λ / µ )1 /1! k ∑ (λ / µ ) i / i! 0 i=0 P2 = (λ / µ ) 2 / 2! k ∑ (λ / µ ) i / i! 0 i=0 The probability a caller will get a busy signal and be blocked is P2 = .1509. Conclusion: The current system will answer approximately 85% of the calls immediately if two agents are employed. But, 15% of callers will receive a busy signal and be turned away. The ones turned away may not call back. CP - 59 Chapter 11 2. Expanded System with waiting allowed With 1 reservation agent:  λ 16 = 0.20 P0 = 1 −  = 1 − µ 20 ï£ ï£¸ Lq = λ2 162 = = 3.2 µ ( µ − λ ) 20(20 − 16) λ 16 = 3.2 + =4 20 µ L = Lq + Wq + Lq = λ 1 W = Wq + Pw = 3.2 = 0.20 hours (12 minutes) 16 µ = 0.20 + 1 = 0.25 hours (15 minutes) 20 λ 16 = = 0.80 µ 20 Operating the expanded reservation service with only one reservation agent appears unacceptable. With 80% of incoming calls waiting (Pw = 0.80) and an average waiting time of 12 minutes (Wq = 12), the company clearly needs to consider using two or more agents. Since Regional's management team agreed that an acceptable service goal was to immediately answer and process at least 85% of the incoming calls, the probability of waiting must be 15% or less. Computing Pw for k = 2 agents and k = 3 agents provides the following. Pw = 1 λ k! µ k kµ P 0 kµ - λ For k = 2 Pw = 1 16 2! 20 2 2(20) (0.4286) = 0.2286 2(20) - 16 3 3(20) (0.4472) = 0.0520 3(20) - 16 For k = 3 Pw = 1 16 3! 20 Based on the value of Pw, 3 reservation agents will be required to meet the service goal of answering 85% of the calls answered immediately. Other operating characteristics of the system with 3 reservation agents are as follows: P0 = 0.4472 CP - 60 Solutions to Case Problems Lq = 0.0189 L = 0.8189 Wq = 0.0012 hours = 0.07 minutes W = 0.0512 hours = 3.07 minutes 3. The analysis in parts (1) and (2) will provide discussion as to which is the best system. If meeting the service goal of answering 85% of the calls immediately is the only consideration, the current system is preferred. Only 2 reservation agents are needed. But, the drawback is that 15% of the callers will be turned away. The callers will either call back later or, perhaps, become lost business. If the expanded system is employed, 3 reservation agents will be needed to meet the service goal of answering 85% of the calls immediately. But, the advantage is that all calls will be answered and the average waiting time is only .0012 hours, or approximately 4 seconds. An extra $20 per hour will be required with the third agent, but better overall service is provided and no customer has to call back. Further analysis of the system with 3 reservation agents will provide additional information for management. For example, the probability of n units in the system is as follows: Units in the System 0 P0 = .4472 1 P1 = (λ / µ )1 (16 / 20)1 P0 = (0.4472) = 0.3577 1! 1! 2 P2 = (λ / µ ) 2 (16 / 20) 2 (0.4472) = 0.1431 P0 = 2! 2! Thus, the probability that there are three or more customers in the system: 1 – (0.4472 + 0.3577 + 0.1431) = 0.0520. The third agent would be busy when there are three or more customers in the system. The probability the third agent would be needed is 0.0520. This is equivalent to saying the agent would be utilized 5.2% of the time. During a one hour period, this would be 0.0520(60 minutes) = 3.12 minutes. Management should be advised that the third agent will be underutilized. Is it worth $20 an hour to have an agent who needs to be available for 3.12 minutes? On an 8-hour basis, this is a cost of $160 for approximately 25 minutes of work. This analysis suggests that management should reconsider the system with two reservation agents. The operating characterizes of this system are as follows: P0 = 0.4286 Lq = 0.1524 L = 0.9524 Wq = 0.0095 hours = 34 seconds CP - 61 Chapter 11 W = 0.0595 hours = 3.57 minutes Pw = 0.2286 The operating characteristics of this 2-agent system are still very good. With Pw = 0.2286, 1 - Pw = 0.7714, or approximately 77% of the calling customers will receive immediate service. But this is only 8% less than the target service goal of 85%. More importantly, the average waiting time is only 34 seconds, which is not unreasonable with today’s telephone systems. This solution saves $20 per hour, or $160 per day, over the system with the underutilized third agent. Look for some good discussion of the pros and cons of the alternative systems in this case. Each of three systems may be considered the best depending upon the criteria used by the analyst. A recommendation that provides management with the key summary statistics for all three systems and the advantages of each may be the best possible write-up for this case. 4. We would need to know the average arrival rate for each hourly period throughout the day. An analysis similar to the one above would determine the recommended number of reservation agents each hour. This information could then be used to develop full-time and part-time shift schedules which would meet the service goals. Case Problem 2: Office Equipment, Inc. 1. λ = 1 call/50 hours = 0.02 calls per hour 2. Mean service time = travel time + repair time = 1 + 1.5 = 2.5 hours µ = 1 / 2.5 hours = 0.4 customers per hour 3. The travel time is 1 hour. While this is considered part of the service time it actually means that the customer will be waiting during the first hour of the service time. Thus, travel time must be added to the time spent in line as predicted model in order to determine the total customer waiting time. 4. Using output from The Management Scientist, we have the following: Probability that no customers are in the system Average number of customers waiting Average number of customers in the system Average time a customer spends in the waiting line Average time until the machine is back in operation Probability of a wait more than one hour Hours a week the technician is not on service calls (0.5380) x 40 hours = 21.5 hours Total cost per hour for the service operation 0.5380 0.2972 0.7593 1.6082 hours* 4.1082 hours 0.4620 $155.93 *The average time a customer spends in the waiting line is 1.6082 hours. This is the average time for the service technician to complete all previous service call commitments and be ready to travel to the new customer. Since the average travel time is 1 hour for the service technician to reach the new customer's office, the total customer waiting time is 1.6082 + 1 = 2.6082 hours. Thus, the one technician is able to meet the company's 3-hour service guideline. The total cost is $155.93 per hour. Note that the waiting line model indicates the probability that a customer has to wait is 0.4620. Since all customers wait an average of 1-hour of travel time whenever the service technician is free, this probability is actually the probability that a customer will have to wait more than 1-hour for a service technician to arrive. CP - 62 Solutions to Case Problems 5. If the company continues to use one technician when the customer base expands to 20 customers, the average time in the waiting line will increase to 6.9454 hours. With an average travel time of 1 hour, the average total waiting time will be 6.9454 + 1 = 7.9454 hours. The total cost will be $397.78 per hour. This average total waiting time is too long and a second technician is definitely necessary. Using output from The Management Scientist, two service technicians provide the following: Probability that no customers are in the system 0.3525 Average number of customers in the waiting line 0.2104 Average number of customers in the system 1.1527 Average time a customer spends in the waiting line 0.5581 hours* Average time until the machine is back in operation 3.0581 hours Probability of a wait more than one hour 0.2949 Hours a week the technicians are not on service calls P0 = 0.3525 (0.3525) x 2 technicians x 40 hours = 28.2 hours P1 = 0.3525 (0.3525) x 1 technician x 40 hours = 14.1 hours Total = 42.3 hours Total cost per hour of service operation $275.27 *The average time a customer spends in the waiting line is 0.5581 hours. This is the average time for the service technician to complete all previous service call commitments and be ready to travel to the new customer. Since the average travel time is 1-hour for the service technician to reach the new customer's office, the total customer waiting time is 0.5581 + 1 = 1.5581 hours. Thus, two technicians are needed to meet the company's 3-hour service guideline when the company reaches 20 customers. The total cost is $275.27 per hour. 6. A comparison of two and three technicians with 30 customers shows that the average total waiting time with two technicians will be 2.6895 hours and the average total waiting time with three technicians will be 1.2626 hours. The hourly cost with two technicians is $391.94 and the hourly cost with three technicians is $397.08. While three technicians provide a smaller waiting time, two technicians are able to meet the 3-hour service guideline for a total lower cost. Thus, the company should continue to use two technicians when the customer base expands to 30 customers. Using output from The Management Scientist, two service technicians provide the following: Probability that no customers are in the system 0.1760 Average number of customers in the waiting line 0.9353 Average number of customers in the system 2.3194 Average time a customer spends in the waiting line 1.6895 hours* Average time until the machine is back in operation 4.1895 hours Probability of a wait more than one hour 0.5600 Hours a week the technicians are not on service calls P0 = 0.1760 (0.1760) x 2 technicians x 40 hours = 14.08 hours P1 = 0.2640 (0.2640) x 1 technician x 40 hours = 10.56 hours Total = 24.64 hours Total cost per hour of service operation $391.94 *The average time a customer spends in the waiting line is 1.6895 hours. While the average travel time is 1-hour for the service technician to reach the new customer's office, the average total customer waiting time is 1.6895 + 1 = 2.6895 hours. 7. The OEI planning committee’s proposal anticipated that three technicians would be needed at a total cost of $397.08 per hour. Thus, the recommendation to stay with two technicians has as annual savings of (397.08 – 391.94) x 8 hours/day x 250 days/year = $10,280. CP - 63 Chapter 12 Simulation Case Problem 1: Tri-State Corporation With the specific financial analysis data input into cells D3:D8, the formulas used to develop the portfolio projection worksheet are shown below. The rows are copied to extend the worksheet to the desired 30-year projection. A 10 11 12 13 14 15 16 Year 1 2 3 4 5 1. B Age =D3 =B12+1 =B13+1 =B14+1 =B15+1 C D Beginning Portfolio =D5 =G12 =G13 =G14 =G15 E Salary =D4 =(1+$D$6)*D12 =(1+$D$6)*D13 =(1+$D$6)*D14 =(1+$D$6)*D15 New Investment =$D$7*D12 =$D$7*D13 =$D$7*D14 =$D$7*D15 =$D$7*D16 F Portfolio Earnings =$D$8*(C12+0.5*E12) =$D$8*(C13+0.5*E13) =$D$8*(C14+0.5*E14) =$D$8*(C15+0.5*E15) =$D$8*(C16+0.5*E16) G Ending Portfolio =C12+E12+F12 =C13+E13+F13 =C14+E14+F14 =C15+E15+F15 =C16+E16+F16 Increasing the annual investment rate in cell D7 will generate the following 30-year portfolio projections: Projected Portfolio $ 721,667 815,397 909,127 1,002,857 Rate 5% 6% 7% 8% A 1% increase in the annual investment rate increases the projected 30-year portfolio by $93,730. The annual investment rate would have to be increased to 8% in order to achieve the $1,000,000 goal. 2. The simulation worksheet that we developed placed the simulated annual salary growth rate in column D and the simulated annual portfolio growth rate in column G. The revised data input section and the cell formulas used to develop the simulation worksheet are as follows. Data Input A 1 2 3 4 5 6 7 8 9 10 11 12 B C D Financial Analysis - Portfolio Projection Age Current Salary Current Portfolio Annual Salary Growth Rate Minimum Rate Maximum Rate Annual Investment Rate Annual Portfolio Growth Rate Mean Rate Standard Deviation CP - 64 25 $34,000 $14,500 0% 10% 8% 10% 5% Solutions to Case Problems Formula Worksheet – Columns D to I 14 15 16 17 18 19 20 D Salary Growth % E =RAND()*($D$8-$D$7) =RAND()*($D$8-$D$7) =RAND()*($D$8-$D$7) =RAND()*($D$8-$D$7) Salary =D4 =(1+D17)*E16 =(1+D18)*E17 =(1+D19)*E18 =(1+D20)*E19 F New Investment =$D$9*E16 =$D$9*E17 =$D$9*E18 =$D$9*E19 =$D$9*E20 G Portfolio Growth % =NORMINV(RAND(),$D$11,$D$12) =NORMINV(RAND(),$D$11,$D$12) =NORMINV(RAND(),$D$11,$D$12) =NORMINV(RAND(),$D$11,$D$12) =NORMINV(RAND(),$D$11,$D$12) H Portfolio Earnings =G16*(C16+0.5*F16) =G17*(C17+0.5*F17) =G18*(C18+0.5*F18) =G19*(C19+0.5*F19) =G20*(C20+0.5*F20) I Ending Portfolio =C16+F16+H16 =C17+F17+H17 =C18+F18+H18 =C19+F19+H19 =C20+F20+H20 The simulated 30-year portfolio amounts will vary considerably from trial to trial. The uncertainty associated with the annual salary growth rate and the annual portfolio growth rate will show that there is uncertainty in reaching the $1,000,000 even if the annual investment rate is increased to 8%. A few simulation trials will point out that the 30-year portfolio variability and the uncertainty of reaching the $1,000,000 goal. However, repeating the simulation numerous times will be necessary to provide an objective basis for estimating the probability of reaching $1,000,000. We preformed the simulation of 1000 trials. The probability of reaching $1,000,000 was estimated to be 0.48. Thus, the simulation conclusion is that there is less than a 50% chance of reaching $1,000,000 even if an 8% investment rate is used. During the 1000 trials, the 30-year portfolio varied from $550,000 to $1,900,000. There was a 30% chance that the portfolio would not reach $900,000. 3. The simulation model suggests additional strategies should be considered to obtain a reasonably high probability of reaching the $1,000,000 portfolio goal. Increasing the annual investment rate to 9% or 10% may be worth considering. If this rate is getting too high, then extending the 30-year period by adding years may be the best strategy for reaching the $1,000,000 goal. 4. The longer 35-year period is definitely a good idea. Expanding the simulation spreadsheet by five years will show that almost every simulated 35-year portfolio exceeded $1,000,000. 1000 simulated trials with the 35-year period showed better than a 99% chance that the portfolio would exceed $1,000,000. In fact, the extra five years increased the expected portfolio from $1,002,857 to $1,697,622, or almost $700,000. This result demonstrates the long-term advantages of investing. 5. The simulation worksheet can be used for any employee. The employee may enter his/her age, current salary, current portfolio, and any assumptions he/she cares to make about salary growth rate, investment rate, and portfolio growth rate. These assumptions in cells D3:D8 would be different for each employee and rows would be added to the worksheet to reflect the number of years appropriate for the employee. By varying the inputs and projecting the ultimate portfolio value, the employees should be able to make investment observations such as the following: • • • • Begin early with an investment program. The more years the better the ending portfolio value. Make new contributions to an investment program at the highest possible rate. Increasing the rate 1% or 2% will have a significant impact on the long-term portfolio. Resist the temptation to use assets in investment programs for short-term personal expenditures. Early withdraws can be shown to have a major impact the long-term portfolio value. Be patient. Benefits of investment programs may not be significant for the first 5 to 10 years. It is really in the later years where a consistent investment program pays its biggest dividends. CP - 65 Chapter 12 Case Problem 2: Harbor Dunes Golf Course The Crystal Ball simulation worksheet that we used is as follows: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 A B C Harbor Dunes Golf Course Simulation - Option 1 Available Tee Times Regular Green Fee Replay Green Fee Cart Fee D E 20 $160 $25 $20 Number of Advance Reservations 8 9 10 11 12 13 14 15 16 17 18 19 20 Probability 0.01 0.04 0.06 0.08 0.10 0.11 0.12 0.15 0.10 0.09 0.07 0.05 0.02 B C Option 1: $25 Green Fee + Cart Fee Number of Replay Requests Probability 0 0.01 1 0.03 2 0.05 3 0.05 4 0.11 5 0.15 6 0.17 7 0.15 8 0.13 9 0.09 10 0.06 Cell Formulas: 25 26 27 28 29 30 31 32 33 34 35 36 37 1. A Crystal Ball Model D E Number of Advance Reservations Number of Times Available Number of Requests for a Replay Number of Replays Crystal Ball Custom Distribution with Data B11:C23 =B3-C27 Crystal Ball Custom Distribution with Data D11:E21 =MIN(C28,C29) Revenue from Advance Reservations Revenue from Replays =4*(B4+B6)*C27 =4*(B5+B6)*C30 Total Revenue Forecast Cell =C32+C33 Selected statistical results: CP - 66 Solutions to Case Problems Simulation results for various. Below is one set of possible results: Statistics Mean Median Standard Deviation Minimum Maximum Option 1 $11,062 $11,160 $ 1,802 $ 5,940 $14,400 Option 2 $11,173 $11,360 $ 1,838 $ 5,760 $14,400 Using the mean, the options are similar, but Option 2 is preferred with a daily revenue advantage of $11,173 - $11,062 = $111. 2. Go with Option 2: The $50 per replay option. 3. Without the replay option, Harbor Dunes reported $10,240 daily revenue. Thus the Option 2 replay policy is estimated to generate an additional revenue of $11,173 - $10,240 = $933 per day. Over a 90day spring season, the estimated revenue increase is 90 * $933 = $83,970. The replay policy is definitely worthwhile. 4. One suggestion is that Harbor Dunes wait until mid-morning to determine the afternoon replay option. If by mid-morning, the afternoon tee time reservations are relatively low, Harbor Dunes may want to offer the Option 1 replay policy which generates more demand for the afternoon tee times. However, if by mid-morning, the afternoon tee time reservations were relatively high, Harbor Dunes would want to continue the Option 2 replay policy. Simulation runs could help determine a level of advance afternoon tee time reservations that would indicate whether to implement the Option 1 or Option 2 replay policy. CP - 67 Chapter 12 Case Problem 3: County Beverage Drive-Thru Excel worksheets patterned after those used for the ATM one- and two- channel simulations in Figure 15.15 and 15.17 may be used to solve this case problem. We recommend using one workbook with three worksheets, one for each of the following Drive-Thru designs: 1 channel with 1 clerk, 1 channel with 2 clerks and 2 channels with 2 clerks. The 1 Channel with 1 Clerk Design Use the format of the Hammondsport 1 ATM simulation model and set up the data for the 1 channel with 1 clerk design as follows: A 1 B C D County Beverage Drive-Thru with One Channel One Clerk 2 3 4 Interarrival times (Exponential Distribution) Mean 6 5 6 7 8 9 10 11 12 13 14 15 16 Service Time Distribution Lower Random No. 0.00 0.24 0.44 0.59 0.73 0.85 0.93 0.98 Upper Random No. 0.24 0.44 0.59 0.73 0.85 0.93 0.98 1.00 Service Time 2 3 4 5 6 7 8 9 The interarrival time for customer 1 would be provided by the cell formula =-$B$4*LN(RAND()) The service time for customer 1 would be provided by the cell formula =VLOOKUP(RAND(),$A$9:$C$16,3) The summary statistics formulas should be provided at the end of the simulation data. The =COUNTIF function can be used to count the number of customers who wait more than 6 minutes and more than 10 minutes. CP - 68 Solutions to Case Problems The 1 Channel with 2 Clerks Design Use the format of the Hammondsport 1 ATM simulation model and set up the data for the 1 channel with 2 clerks design as follows: A 1 B C D County Beverage Drive-Thru with One Channel Two Clerks 2 3 4 Interarrival times (Exponential Distribution) Mean 6 5 6 7 8 9 10 11 12 13 Service Time Distribution Lower Random No. 0.00 0.20 0.55 0.85 0.95 Upper Random No. 0.20 0.55 0.85 0.95 1.00 Service Time 1 2 3 4 5 The service time for customer 1 would be provided by the cell formula =VLOOKUP(RAND(),$A$9:$C$13,3) The summary statistics formulas should be provided at the end of the simulation data. The =COUNTIF function can be used to count the number of customers who wait more than 6 minutes and more than 10 minutes. The 2 Channels with 2 Clerks Design Use the format of the Hammondsport 2 ATMs simulation model and set up the data as previously shown for the 1 channel with 1 clerk design. Customer interarrival times and service times are the same as shown for the 1 channel with 1 clerk design. This part of the case is optional in that significant spreadsheet modeling skills are required to duplicate the 2-channel simulation model. Selected cell formulas area as follows: Cell I16 Cell J16 Cell I17 Cell J17 = = = = G16 0 IF(I16=MIN(I16,J16),G17,I16) IF(J16=MIN(I16,J16),G17,J16) Cells I17 and J17 can be copied to fill columns I and J. The summary statistics formulas should be provided at the end of the simulation data. The =COUNTIF function can be used to count the number of customers who wait more than 6 minutes and more than 10 minutes. Each design was tested with a 1000 customer simulation run. Data on the first 100 customers was discarded and summary statistics collected for a total of 900 customers. CP - 69 Chapter 12 Simulation results will vary but the approximate results are as follows: Characteristic Number Waiting Probability of Waiting Average Waiting Time Maximum Waiting Time Utilization of Drive Thru Number Waiting > 6 Minutes Probability Waiting > 6 Minutes Number Waiting > 10 Minutes Probability Waiting > 10 Minutes 1 Channel 1 Clerk 640 0.71 6.1 37.8 0.72 322 0.36 191 0.21 1 Channel 2 Clerks 377 0.42 1.0 11.7 0.42 26 0.03 4 0.01 2 Channels 2 Clerks 167 0.19 0.4 9.3 0.36 7 0.01 1 0.00 The 1 channel with 1 clerk system appears unacceptable. The mean waited time is over 6 minutes which exceeds the company guideline of 1.5 minutes. In addition, over 300 customers waiting over 6 minutes and almost 200 customers waited over 10 minutes. The company must do something to improve the service characteristics of its drive-thru operation or face the loss of substantial business. The 1 channel with 2 clerks system appears to be the best design., The mean waiting time of approximately 1 minute is with the company’s guideline of 1.5 minutes. Relatively few customers experienced the 6 to 10 minute waiting times. The performance of the 2 channel with 2 clerks system is the best overall, but the added cost may not justify the expansion to the two channel operation. CP - 70 Chapter 13 Decision Analysis Case Problem 1: Property Purchase Strategy The decision tree for the Oceanview decision problem is shown below. Note that the final outcome of whether the zoning change is approved or rejected by the voters occurs of Oceanview decides to bid and then wins the bid. Otherwise, the outcome of the vote on the zoning change does not enter into the problem. Payoff* Highest Bid Bid Predicts Approval Zoning Approved 1,985,000 Zoning Rejected -515,000 9 6 Not Highest Bid 3 -15,000 Do Not Bid Market Research -15,000 Highest Bid 2 Bid Predicts Rejection Zoning Approved 1,985,000 Zoning Rejected -515,000 10 7 Not Highest Bid 4 -15,000 Do Not Bid 1 -15,000 Highest Bid Bid No M arket Research Zoning Approved 2,000,000 Zoning Rejected -500,000 11 8 Not Highest Bid 5 Do Not Bid 0 0 * Payoff values are the profit computed from the given revenue, property cost, construction expenses, property cost deposit and marketing research cost given in the problem. CP - 71 Chapter 13 The decision tree with branch probabilities and expected values is as follows: Approved .6585 Highest Bid .2 EV = 1,131,250 Rejected .3415 Bid Predicts Approval .41 9 6 Not Highest Bid .8 3 -15,000 Approved .0508 2 Highest Bid .2 EV = 78,993 10 7 -15,000 Do Not Bid 1 -15,000 Approved .3 Highest Bid .2 No M arket Research 11 8 -500,000 EV = 50,000 Not Highest Bid .8 5 2,000,000 EV = 250,000 Rejected .7 Bid -515,000 EV = -89,600 Not Highest Bid .8 4 1,985,000 EV = -388,000 Rejected .9492 Bid Predicts Rejection .59 -515,000 EV = 214,250 Do Not Bid Market Research 1,985,000 Do Not Bid 0 0 If the market research is not conducted, Oceanview should go ahead and bid. The expected value of this action is $50,000 as shown at node 8 of the decision tree. However, the analysis shows that the recommended strategy is to conduct the market research first. the value of the market research exceeds its cost. The optimal strategy is as follows: If the market research predicts approval of the zoning change, bid. If the market research predicts rejection of the zoning change, do not bid. The expected value of this strategy, including the cost of conducting the market research, is $78,993. The market research has a value of $78,993 - $50,000 = $28,993 over and above the cost of the research only. CP - 72 Solutions to Case Problems Case Problem 2: Lawsuit Defense Strategy 1. Decision Tree Allied Accepts John's Offer of $750,000 $750 John Accepts Allied's Counteroffer 0.10 $400 Allied Loses Damages of $1,500,000 0.3 $1500 1 Allied Counteroffers with $400,000 John Rejects Allied's Allied Loses Damages Counteroffer of $750,000 2 3 0.4 0.5 Allied W ins No Damages 0.2 Allied Accepts John's Counteroffer $750 $0 $600 Allied Loses Damages of $1,500,000 0.3 John Counteroffers with $600,000 4 0.5 Allied Rejects John's Counteroffer Allied Loses Damages of $750,000 5 0.5 Allied W ins No Damages 0.20 To find the optimal decision strategy, we must fold the tree back computing expected values at the chance nodes and choosing the least cost alternative at the decision nodes. Shown below are the values computed at each node. Node 1 2 3 4 5 Value 670 670 825 600 825 CP - 73 $1500 $750 $0 Chapter 13 Shown below is the decision tree with all nonoptimal branches eliminated. John Accepts Allied's Counteroffer 0.10 $400 Allied Loses Damages of $1,500,000 0.3 $1500 1 Allied Counteroffers with $400,000 Allied Loses Damages John Rejects Allied's of $750,000 Counteroffer 2 3 0.4 0.5 Allied W ins No Damages 0.2 Allied Accepts John's Counteroffer $750 $0 $600 John Counteroffers with $600,000 4 0.5 2. Allied should not accept John's offer to settle for $750,000. The strategy associated with a counteroffer of $400,000 has an expected value of $670,000. 3. If John accepts Allied's counteroffer of $400,000, no further action is required. If John rejects Allied's counteroffer and elects to have a jury decide the settlement amount, Allied must prepare for a trial. If John counteroffers with $600,000, Allied should accept John's counteroffer. 4. To develop the risk profile for the optimal strategy, we simply multiple the branch probabilities on all paths to the end points of the decision tree. The risk profile is given below in tabular form. Settlement Amount ($1000s) 0 400 600 750 1500 Probability 0.08 0.10 0.50 0.20 0.12 1.00 CP - 74 Chapter 14 Multicriteria Decision Problems Case Problem: EZ Trailers, Inc. Let x11 x12 x21 x22 s11 s12 s21 s22 = = = = = = = = number of EZ-190 trailers produced in March number of EZ-190 trailers produced in April number of EZ-250 trailers produced in March numb r of EZ-250 trailers produced in April EZ-190 ending inventory in March EZ-190 ending inventory in April EZ-250 ending inventory in March EZ-250 ending inventory in April P1 Goal: Meet demand for the EZ-250: March: April: 300 + x21 - s21 - d1+ + d1− = 1000 (1) s21 + x22 - s22 - d 2+ + d 2− = 1200 (2) P2 Goal: Meet demand for the EZ-190: March: April: 200 + x11 - s11 - d 3+ + d 3− = 800 (3) s11 + x12 - s12 - d 4+ + d 4− = 600 (4) P3 Goal: Limit labor fluctuations from month to month to at most 1000: March: 5300 ≤ 4 x11 + 6 x21 ≤ 7300 4 x11 + 6 x21 - d 5+ + d 5− = 5300 + 6 (5) − 6 4 x11 + 6 x21 - d + d = 7300 (6) April: (4 x11 + 6 x21) - 1000 ≤ 4 x12 + 6 x22 ≤ (4 x11 + 6 x21) + 1000 4 x12 + 6 x22 = [(4 x11+ 6 x21) - 1000] + d 7+ + d 7− + 7 − 7 4 x12 + 6 x22 - 4 x11 - 6 x21 - d + d = -1000 or (7) 4 x12 + 6 x22 = [(4 x11 + 6 x21) + 1000] + d 8+ + d8− or 4 x12 + 6 x22 - 4 x11 - 6 x21 - d 8+ + d8− = 1000 CP - 75 (8) Chapter 14 The complete goal programming model is ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Min P1 d1− + P1 d 2− + P2 d3− + P2 d 4− + P3 d5− + P3 d 6+ + P3 d 7− + P3 d8+ s.t. x 21 – s21 – d1+ + d –1 = 800 + – x22 + s 21 – s 22 – d 2 + d 2 = 1200 + – + – x11 – s 11 – d3 + d 3 = 600 x12 + s 11 – s 12 – d4 + d 4 = 600 + – + – 4 x11 + 6 x21 – d 5 + d 5 = 5300 4 x11 + 6 x21 – d 6 + d 6 = 7300 + – –4 x11 + 4 x12 – 6 x21 +6x22 – d 7 + d 7 = – 1000 –4 x11 + 4 x12 – 6 x21 +6x 22 – d +8 + d –8 = 1000 all variables ≥ 0 Information Needed for the Managerial Report 1. EZ-190 EZ-250 2. March 775 800 No changes since the ending inventories for the optimal production schedule are as follows: EZ-190 EZ-250 3. April 425 1200 March 175 0 April 0 0 The following constraints must be added to the model: April ending inventory: s12 ≥ 100, s22 ≥ 100 Maximum storage of 300 units in each month: s11 ≤ 300, s12 ≤ 300, s21 ≤ 300, s22 ≤ 300 The new optimal production schedule is as follows: EZ-190 EZ-250 March 900 800 April 400 1300 March 300 0 April 100 100 The corresponding ending inventories are: EZ-190 EZ-250 CP - 76 Solutions to Case Problems 4. The new optimal production schedule is: EZ-190 EZ-250 March 625 800 April 275 1200 March 25 0 April 0 0 The corresponding ending inventories are: EZ-190 EZ-250 CP - 77 Chapter 15 Forecasting Case Problem 1: Forecasting Sales 1. Month 1 corresponds to January for year 1, month 2 corresponds to February for year 1, and so on. A graph of the time series is shown below: 300 250 Sales 200 150 100 50 0 5 10 15 20 25 Month 2. Analysis of seasonality: Month January February March April May June July August September October November December Seasonal-Irregular Component Values 1.445 1.441 1.301 1.297 1.344 1.343 1.047 1.034 1.044 1.054 .779 .801 .882 .834 .857 .848 .618 .638 .725 .675 .843 .862 1.137 1.180 Seasonal Index 1.44 1.30 1.34 1.04 1.05 .80 .83 .85 .63 .70 .85 1.16 CP - 78 30 35 40 Chapter 15 The deseasonalized time series is shown below: t 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Deseasonalized Sales 168.06 180.77 173.13 171.15 175.24 175.00 174.70 178.82 174.60 185.71 178.82 177.59 182.64 183.08 184.33 185.58 183.81 186.25 Deseasonalized Sales 189.16 189.41 193.65 185.71 196.47 198.28 195.83 196.15 197.76 197.12 200.00 200.00 200.00 204.71 200.00 211.43 203.53 202.59 t 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 The trend line fitted to the deseasonalized time series is T t = 169.499 + 1.02 t 3. Sales forecasts Forecast for Year 4 Using T t = 169.499 + 1.02 t Month January February March April May June July August September October November December 4. Trend Forecast 207.239 208.259 209.279 210.299 211.319 212.339 213.359 214.379 215.399 216.419 217.439 218.459 Seasonal Index 1.44 1.30 1.34 1.04 1.05 .80 .83 .85 .63 .70 .85 1.16 Monthly Forecast 298.424 270.737 280.434 218.711 221.885 169.871 177.088 182.222 135.701 151.493 184.823 253.194 Forecast error = $295,000 - $298,424 = -$3,424 The forecast we developed over predicted by $3,424; this represents a very small error. 5. The analysis can be easily updated each month, especially if a computer software package is used to perform the analysis. CP - 79 Solutions to Case Problems Case Problem 2: Forecasting Lost Sales 1. The data used for the forecast is the Carlson sales data for the 48 months preceding the storm. Using the trend and seasonal method, the seasonal indexes and forecasts of sales assuming the hurricane had not occurred are as follows: Month January February March April May June July August September October November December 2. Month September October November December Forecast ($ million) 2.16 2.54 3.06 4.60 The data used for this forecast is the total sales for the 48 months preceding the storm for all department sores in the county. Using the trend and seasonal method, the seasonal indexes and forecasts of county-wide department store sales assuming the hurricane had not occurred are as follows: Month January February March April May June July August September October November December 3. Seasonal Index 0.957 0.819 0.907 0.929 1.011 0.937 0.936 0.974 0.797 0.936 1.119 1.677 Seasonal Index 0.773 0.813 0.976 0.935 0.989 0.924 0.901 1.017 0.861 0.907 1.141 1.763 Month September October November December Forecast ($ million) 50.55 53.20 66.78 103.11 By comparing the forecast of county-wide department store sales with actual sales, one can determine whether or not there are excess storm-related sales. We have computed a "lift factor" as the ratio of actual sales to forecast sales as a measure of the magnitude of excess sales. Forecast Sales ($ million) 50.55 53.20 66.78 103.11 273.64 Actual Sales ($ million) 69.0 75.0 85.2 121.8 351.0 Lift Factor 1.365 1.410 1.276 1.181 1.283 From the analysis a strong case can be made for excess storm related sales. For each month, actual sales exceed the forecast of what sales would have been without the hurricane. For the 4-month total, actual sales exceeded the forecast by 28.3%. CP - 80 Chapter 15 The explanation for the increase is that people had to replace real and personal property damaged by the storm. In addition, the additional construction workers, the disaster relief teams, and so on, created additional commercial activity in the area. 4. One approach would be to use the forecast of what sales would have been without the hurricane and then multiply by the lift factor to account for the excess storm-related sales. Such an estimate of lost sales is developed below: Forecast ($ million) Lift Factor 2.16 2.54 3.06 4.60 1.365 1.410 1.276 1.181 Lost Sales ($ million) Total 2.948 3.581 3.905 5.433 15.867 Based on this analysis, Carlson Department Stores can make a case to the insurance company for a business interruption claim of $15,867,000. Another approach would be to use the 48 months of historical data to compute a market share for Carlson. That is, compute Carlson’s sales as a fraction of county-wide department store sales. Then you could develop a forecast of Carlson’s market share for September through December. Finally, an estimate of lost sales for each of the four months can be obtained by multiplying the forecasts of market share by the actual department store sales. CP - 81 Chapter 16 Markov Processes Case Problem: Dealer’s Absorbing State Probabilities in Black Jack 1. The first step is to create the transition probability matrix for the dealer’s hand. The states are defined by the value of the dealer’s hand after the up card is dealt, and the values the dealer’s hand may assume after the down card is revealed, and after taking hits. The absorbing states are 17, 18, 19, 20, 21, and bust. According to the house rules, once the dealer’s hand takes on one of these values she quits taking hits and either pays or collects the bets that are on the table. The states S12, S13, S14, S15, and S16 represent what are called soft hands; they include an ace that may be played as 1 or 11. There are no S17, S18, S19, S20, and S21 states because the dealer plays these the same as hard hands of the same value. There are 27 states so, due to space restrictions, the transition matrix that is shown below is printed in 3 parts. All of the rows and a subset of the columns are shown in each part. A S12 S13 S14 S15 S16 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Bust A 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 S12 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 S13 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 S14 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 S15 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 S16 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 CP - 82 2 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 3 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 4 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 5 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 6 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 Solutions to Case Problems A S12 S13 S14 S15 S16 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Bust 7 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 8 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 9 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 10 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 11 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 12 13 14 15 16 0.3077 0.0769 0.0769 0.0769 0.0769 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 0.3077 0.0769 0.0769 0.0769 0.0000 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0 0 0 0 0 0 0.3077 0.0769 0.0769 0.0000 0.0000 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0 0 0 0 0 0 0.3077 0.0769 0.0000 0.0000 0.0000 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0 0 0 0 0 0 0.3077 0.0000 0.0000 0.0000 0.0000 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0 0 0 0 0 0 CP - 83 17 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0769 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 1 0 0 0 0 0 Chapter 16 A S12 S13 S14 S15 S16 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Bust 18 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0 1 0 0 0 0 19 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0 0 1 0 0 0 20 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0769 0 0 0 1 0 0 21 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0769 0.3077 0.0769 0.0769 0.0769 0.0769 0.0769 0 0 0 0 1 0 Bust 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.3077 0.3846 0.4615 0.5385 0.6154 0 0 0 0 0 1 CP - 84 Solutions to Case Problems The Q matrix is defined as the first 21 rows (through state 16) and the first 21 columns (through state 16) of the transition matrix. We must now compute I – Q and (I – Q)-1. We did this using Excel and Excel’s matrix inversion procedure described in the chapter appendix. To find the absorption probabilities we multiply (I – Q)-1 times the R matrix. The R matrix is a 21 x 6 matrix. It is identified as the first 21 rows and the last 6 columns of the transition matrix. We show (I – Q)-1R below. Probability of Dealer’s Finishing State Given Starting State A S12 S13 S14 S15 S16 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 0.1308 0.1510 0.1455 0.1400 0.1346 0.1292 0.1398 0.1350 0.1305 0.1223 0.1654 0.3686 0.1286 0.1200 0.1114 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 18 0.1308 0.1510 0.1455 0.1400 0.1346 0.1292 0.1349 0.1305 0.1259 0.1223 0.1063 0.1378 0.3593 0.1200 0.1114 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 19 0.1308 0.1510 0.1455 0.1400 0.1346 0.1292 0.1297 0.1256 0.1214 0.1177 0.1063 0.0786 0.1286 0.3508 0.1114 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 20 0.1308 0.1510 0.1455 0.1400 0.1346 0.1292 0.1240 0.1203 0.1165 0.1131 0.1017 0.0786 0.0694 0.1200 0.3422 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 21 0.3616 0.1510 0.1455 0.1400 0.1346 0.1292 0.1180 0.1147 0.1112 0.1082 0.0972 0.0741 0.0694 0.0608 0.1114 0.3422 0.1035 0.0961 0.0892 0.0828 0.0769 Bust 0.1153 0.2450 0.2725 0.3000 0.3272 0.3541 0.3536 0.3739 0.3945 0.4164 0.4232 0.2623 0.2447 0.2284 0.2121 0.2121 0.4827 0.5196 0.5539 0.5858 0.6154 This matrix shows the probabilities for the ending values of the dealer’s hand given any of the starting states identified by each row. For instance, if the dealer has a 6 as an up card, the probability of finishing with 17 is .1654, the probability of finishing with 18 is .1063 and so on. The probability of the dealer busting is .4232. For this reason, black jack players recommend not taking a hit with a hand of 12 or better. Why should the player take a chance of busting when the dealer has a high probability of busting? CP - 85 Chapter 16 2. For this situation another row and column must be added to the transition matrix to account for the soft 17 state, S17. When the dealer has to stay on S17, there was no reason to differentiate soft 17 and hard 17. So only the state, 17 was used. For space reasons, we do not show the new transition matrix here. But, after computing (I – Q)-1R, the absorbing state probabilities are given below. Probability of Dealer's Finishing State Given Starting State A S12 S13 S14 S15 S16 S17 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 0.0575 0.0829 0.0823 0.0813 0.0801 0.0786 0.3422 0.1301 0.1263 0.1224 0.1184 0.1148 0.3686 0.1286 0.1200 0.1114 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 18 0.1432 0.1625 0.1562 0.1500 0.1438 0.1377 0.1114 0.1365 0.1320 0.1273 0.1229 0.1148 0.1378 0.3593 0.1200 0.1114 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 19 0.1432 0.1625 0.1562 0.1500 0.1438 0.1377 0.1114 0.1313 0.1271 0.1228 0.1184 0.1148 0.0786 0.1286 0.3508 0.1114 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 20 0.1432 0.1625 0.1562 0.1500 0.1438 0.1377 0.1114 0.1257 0.1218 0.1179 0.1138 0.1103 0.0786 0.0694 0.1200 0.3422 0.1114 0.1035 0.0961 0.0892 0.0828 0.0769 21 Bust 0.3740 0.1389 0.1625 0.2669 0.1562 0.2929 0.1500 0.3189 0.1438 0.3448 0.1377 0.3704 0.1114 0.2121 0.1196 0.3567 0.1162 0.3767 0.1126 0.3971 0.1089 0.4177 0.1057 0.4395 0.0741 0.2623 0.0694 0.2447 0.0608 0.2284 0.1114 0.2121 0.3422 0.2121 0.1035 0.4827 0.0961 0.5196 0.0892 0.5539 0.0828 0.5858 0.0769 0.6154 This matrix shows the probabilities for the ending values of the dealer’s hand given any of the starting states identified by each row for the case when the dealer hits soft 17. For instance, if the dealer has a 6 as an up card, the probability of finishing with 17 is .1148, the probability of finishing with 18 is .1148, and so on. The probability of the dealer busting is .4395. 3. Mathematicians have shown that when the dealer stays on soft 17 it is better for the player. We can provide the following argument using the finishing values for the dealer’s hand shown in part 2 above. Note that when the dealer gets S17 and hits it, the finishing values and probabilities for the dealer’s hand are Ending Value 17 18 19 20 21 Bust Probability .3422 .1114 .1114 .1114 .1114 .2121 CP - 86 Solutions to Case Problems If the player has 17, the probability of a tie is .3422, the probability of losing is 4(.1114) = .4456, and the probability of winning is .2121. So, when the player has 17, she will do better (a tie) if the dealer stays on S17. Similarly, if the player has 18, 19, 20, or 21, she will do much better if the dealer stays on S17. She will always win. But, if the player has 12, 13, 14, 15, or 16, she will be better off if the dealer hits soft 17. This is because when the dealer busts she will win (if the dealer does not hit S17 she loses). And, this will happen with probability .2121. But, the player will lose with probability 1 - .2121 = .7879 when the dealer hits S17. This small probability of the player winning with 12, 13, 14, 15, and 16 when the dealer hits S17 is not enough to offset the large increase in the probability of the player winning with 17, 18,19, 20, and 21 when the dealer stays on S17. CP - 87 Chapter 21 Dynamic Programming Case Problem: Process Design The optimal solution is to operate the heater at a temperature of 800°, use catalyst C2 for the reactor, and purchase separator S1. This results in a weekly profit of $9,580 calculated as follows. Weekly revenue (2700 lbs. x $6.00) Weekly costs $16,200 Payback on heater (12,000/100) Payback on reactor ( 50,000/100) Payback on separator (20,000/100) Operate heater Operate reactor Operate separator (2700 x 0.10) Raw material (4500 x 1.00) Net Weekly Profit 120 500 200 380 650 270 4,500 6,620 $ 9,580 Calculations Let stage 1 be separation, stage 2 reactor, and stage 3 heating. Stage 1 d1 x1 s1 s2 d1* f1(d1) 900 lb. $5,110 $5,170 s2 $5,170 1800 lb. 10,420 10,390 s1 10,420 2700 lb. 15,730 15,610 s1 15,730 Stage 2 d2 x2 c1 c2 d 2* f2(d2) x1 4500 lb., 700° $4,220 $9,270 c2 $9,270 1800 4500 lb., 800° 9,470 14,580 c1 14,580 2700 Stage 3 d3 x3 700° 800° d 3* f3(d3) x1 4500 lb. $4,370 $9,580 800° $9,580 (4500, 800°) CP - 88