ESD. 71. Flexibility Analysis in Ocean Freight Transportation System. Lita Das

Similar documents
THE VALUE OF DISCRETE-EVENT SIMULATION IN COMPUTER-AIDED PROCESS OPERATIONS

Basestock Model. Chapter 13

Aggregate Planning and S&OP

White Paper. Best Practices to Reduce International Freight Costs. A Guide for Global Logistics Managers

Forecasting Survey. How far into the future do you typically project when trying to forecast the health of your industry? less than 4 months 3%

Planning Optimized. Building a Sustainable Competitive Advantage WHITE PAPER

Scenario-based simulation of revenue loss at seismically damaged seaports

TRANSPORTATION PROBLEM AND VARIANTS

Antti Salonen KPP227 KPP227 1

Refuse Collections Division Solid Waste Services Department Anchorage: Performance. Value. Results.

Miami River Freight Improvement Plan Financial Management Number:

THE SUPPLY CHAIN OF FERTILIZER IN SOUTH VIETNAM A CASE STUDY

D DAVID PUBLISHING. Stacking Sequence of Marine Container Minimizing Space in Container Terminals. 1. Introduction. Ning Zhang and Yutaka Watanabe

Risk Analysis Overview

with Dr. Maria Rey-Marston Lecturer, Georgia Tech Supply Chain & Logistics Institute Managing Partner, MRM+ Partners LLP

Inventory Control for Highly Variable Demand

Predictive Safety s Fatigue Management System: Benefits Realization Report

New Specialty Crops for California

Whitepaper. Smarter Supply Chain Solutions

SAP Supply Chain Management

Assurance of Supply. How Confidence in a Reliable Supply Stream Creates Healthy Companies. A GT Nexus White Paper

SUSTAINABLE REVERSE LOGISTICS

Factors Affecting Transportation Decisions. Transportation in a Supply Chain. Transportation Modes. Road freight transport Europe

The Use of Simulation for Bulk Cargo Terminal Planning and Design

Planned Shortages with Back-Orders

Ammonia costs spike sharply higher Nitrogen prices disrupt calm in fertilizer market By Bryce Knorr, grain market analyst

TIME SAVINGS BENEFITS ASSESSMENT FOR SECURE BORDER TRADE PROGRAM PHASE II

DHL OCEAN CONNECT LCL KEEPING YOUR PROMISES AND DEADLINES

INCOTERMS UIA Annual Congress. International Sales Commission. Melina Llodra

Επιχειρησιακή Συνέχεια και Εφοδιαστική Αλυσίδα, 14 Οκτωβρίου Moving ahead in a changing environment

Administration Division Public Works Department Anchorage: Performance. Value. Results.

Make-to-Stock under Drum-Buffer-Rope and Buffer Management Methodology

Balancing Risk and Economics for Chemical Supply Chain Optimization under Uncertainty

EXW EX WORKS (... named place) FCA. FREE CARRIER (... named place)

Inventory systems for independent demand

THE PROJECT. Executive Summary. City of Industry. City of Diamond Bar. 57/60 Confluence.

Epicor Selection and Implementation

Simulation of Container Queues for Port Investment Decisions

Operations Management

Incoterms ICC Rules for the use of Domestic and International Trade Terms. Kenya Maritime Authority

USD per cubic meter (CBM) USD per consignment, per waybill

Chapter 4. Models for Known Demand

Instructions on Data Collection for the UNESCAP Time/Cost-Distance Methodology. - Basic Template Version-

Global Sourcing: What You Need to Know. Robi Bendorf, C.P.M., Consultant Bendorf & Associates, ( Monroeville, PA ,

Cross-Dock Modeling And Simulation Output Analysis

Dynamic Reallocation of Portfolio Funds

Justifying Simulation. Why use simulation? Accurate Depiction of Reality. Insightful system evaluations

1) Operating costs, such as fuel and labour. 2) Maintenance costs, such as overhaul of engines and spraying.

Implementation Status & Results Brazil Rio de Janeiro Mass Transit Project II (P111996)

Strategic inventory management through analytics

KPI ENCYCLOPEDIA. A Comprehensive Collection of KPI Definitions for PROCUREMENT

Analysis of Railway Fulfillment of Shipper Demand and Transit Times. Prepared for: Rail Freight Service Review

Intelligent Fulfillment

Empty Containers Repositioning in South African Seaports

Lehman Brothers T Conference San Francisco. Craig DeYoung, Vice President Investor Relations December 9, 2004

CHAPTER 6. Inventory Costing. Brief Questions Exercises Exercises 4, 5, 6, 7 3, 4, *14 3, 4, 5, 6, *12, *13 7, 8, 9, 10, 11, 12, 13

Whitepaper Series Cross-Docking Trends Report Secondary Packaging Outsourcing Report

Introduction to Transportation Systems

Empty Intermodal Container Management

Call Center Benchmark India

IMPORTING: COMMERCIAL INVOICES. Importing into the United States. The Import Commercial Invoice COMMERCIAL INVOICE

15th Annual Masters of Logistics Survey: Strategy Shift

Importing to the U.S.: Key requirements you need to be aware of

Dear Importer, BONDING: December Client Alert: New US Customs Compliance regulations for 2009

Inventory Management

Using Flowcasting to Predict Future Retail Supply Chain Constraints

Revenue for chemical manufacturers

I know that you all understand the critical importance of the freight transportation system

What is Waveless Processing and How Can It Optimize My Operation?

Coal, Carbon, and the Future of the Global Energy Mix

Manufacturing Efficiency Guide DBA Software Inc.

The Ultimate Guide to Performance Check-Ins

Call Center Benchmark

FINANCE AND STRATEGY PRACTICE CFO EXECUTIVE BOARD. Safeguarding Supply. Protecting the Enterprise from Unforeseen Supply Chain Risks

Q SOCIAL TRENDS REPORT

Topics in Supply Chain Management. Session 3. Fouad El Ouardighi BAR-ILAN UNIVERSITY. Department of Operations Management

Determination of Operational Parameters for an Efficient Container Service in the Port of Guaymas

Business Math Curriculum Guide Scranton School District Scranton, PA

The U.S. Over-Supply of Oil is Ending

Analysis of Demand Variability and Robustness in Strategic Transportation Planning

MRP Configuration - Adobe Interactive Forms (Japanese) - SCN Wiki

By: Adrian Chu, Department of Industrial & Systems Engineering, University of Washington, Seattle, Washington November 12, 2009.

Contents. Semiconductor DQ Monday Report Issue 47

COURSE LISTING. Courses Listed. 30 January 2018 (11:31 GMT) TM100 - SAP Transportation Management. SAP Transportation Management

A MANAGER S ROADMAP GUIDE FOR LATERAL TRANS-SHIPMENT IN SUPPLY CHAIN INVENTORY MANAGEMENT

Future Freight Flows NCHRP

Order Entry User Manual

MEDLINE HAWAIIAN ISLANDS SHIPPING POLICY AND DEDICATED CUSTOMER SERVICE

Signs align for corn profit hopes Short crop in Brazil could be fix the market needs By Bryce Knorr, senior grain market analyst

Developing and Delivering Complex Projects using Quantitative Risk Analysis Colin Cropley BE(Chem), PMP, Cert PRINCE2 Practitioner Managing Director,

Schedule 2 TECHNICAL INFORMATION

DYNAMIC FULFILLMENT (DF) The Answer To Omni-channel Retailing s Fulfillment Challenges. Who s This For?

GHANA. February 2015 CONTENTS. 1.Introduction Farm Gate price Data Collection in Ghana: Data Reporting... 3

Chapter 12 Inventory Management. Inventory Management

Justifying Advanced Finite Capacity Planning and Scheduling

Proceedings of the 2015 Winter Simulation Conference L. Yilmaz, W. K. V. Chan, I. Moon, T. M. K. Roeder, C. Macal, and M. D. Rossetti, eds.

Renewable Energy Powered Greenhouses A Concept Paper, March 2011

ANNUAL PLATTE RIVER SURFACE WATER FLOW SUMMARY

International Shipping & Customs Clearance Guidelines

MIT SCALE RESEARCH REPORT

Transcription:

Flexibility Analysis in Ocean Freight Transportation System Lita Das

Flexibility Analysis: Inventory Management of Ocean Freight System Abstract: Managing inventory in an ocean freight transportation network can be a very challenging task. In the case discussed in this report the decision maker is observing an uncertain demand, which is to be satisfied, by shipping units through ocean freight mode. Another major uncertainty that drives decisions in this system is lead time which is a sum of lead times across 4 different segments of the ocean freight system, namely port to port, destination port dwell, destination port to destination and unloading time at the destination port. The variability in the demand and lead-time force the decision maker to incur inventory carrying costs and high transits cost or switch to airfreight mode to avoid lost sales or invest in options contracts. The objective of this report is to conduct a evaluation for two different flexible cases against a base case scenario and compare the value in terms of annual inventory cost and annual transit cost. The base case and the flexible cases are analyzed using Monte Carlo simulation method by running the simulations in each case for a 1000 times. Finally a comparative analysis of the base case and flexible cases 1 & 2 is done on basis of a Value at Risk & Gain (VARG) curve and multidimensional criteria (expected, maximum and minimum costs), which are a result of the simulation model. The results conclude that the flexible cases prove to be a more favorable decision as compared to the base case in terms of annual inventory and transit cost. Results are tabulated below: (the values marked in yellow are the favorable outcomes) Value in million $ Base Case Flexible Case1 Expected Annual Inventory Cost 28.14 14.64 Minimum Annual Inventory Cost 23.65 12.17 Maximum Annual Inventory Cost 32.68 17.38 Standard Deviation of annual inventory cost 1.40 0.84 P 10 26.40 13.55 P 90 29.94 15.75 Value in million $ Base Case Flexible Case2 Value in million $ Expected Annual 28.14 18.01 Expected Inventory Cost Annual Transit Minimum Annual Inventory Cost Maximum Annual Inventory Cost Standard Deviation of Cost 23.65 15.74 Minimum Annual Transit Cost 32.68 19.90 Maximum Annual Transit Cost 1.40 0.61 Standard Deviation of Base Case Flexible Case2 286.22 210.80 266.03 195.64 301.93 227.85 5.02 4.79 2

annual inventory cost annual Transit cost P 10 26.40 17.25 P 10 253.84 195.64 P 90 29.94 18.79 P 90 301.93 227.85 Table of Contents Introduction...4 System Background...4 The System...4 Base Case & Flexible Cases...6 Structure of Analysis...6 Uncertain Factors Determination...7 Fixed Factors Determination... 11 Sensitivity Analysis... 12 Simulation... 15 Calculations... 17 Simulation Method...18 Simulation Model for base case...18 Inventory Cost as a Performance Measure... 21 Base Case... 21 Flexible Case... 23 Flexible Case1...24 Base Case vs Flexible Case1 *... 28 Flexible Case2...30 Sensitivity Analysis to determine maximum order quantity allowable...31 Base Case vs Flexible Case2... 36 Transit Cost as a Performance Measure... 38 Base Case... 38 Results of Simulation for Base Case...38 Flexible Case 2...39 Base Case vs Flexible Case2... 41 Conclusion... 42 Course Reflections and Lessons Learned... 44 References... 45 3

Introduction With the ever-increasing growth of international trade, it is very important to recognize uncertainties affecting the ocean freight system. Trends in ocean freight show that US ports handled 17% more cargo (in metric tons) in 2010 as compared to 2009. In the current market, decision makers,who can look out for opportunities in the risks and create flexible designs, which add value to the system, can survive in this global competitive ocean freight network. This application portfolio will discuss an ocean freight network that is affected by demand, lead-time and fuel cost uncertainties that encourage the decision maker to build flexibility into the system in order to gauge the benefits of the latter over the conventional base case. The performance metrics used are annual inventory cost and annual transit cost. Hence the cases with lower costs are more favorable over the other. System Background The System The System of interest is global ocean transportation network of leading manufacturing and logistics firms. Specifically it includes freight transportation of from China to Port of Los Angeles.The final destination of the freight is the distribution center in San Francisco. One of the major aspects of their operation is the global transportation of containers of product between suppliers, assembly plants, and customer locations. This network consists of the entire chain from the origin to the destination. A diagrammatic representation of the network is shown below: Origin Origin Port Ocean Transit Destination Port Destination Fig. 1:Diagrammatic representation of the System 4

The system includes the following: 1. Origin, Origin Port, Ocean Transit, Destination Port, Destination 2. Delivery Time Schedule up to the final destination 3. Ground Transportation between the origin/origin port & destination port/destination 4. Demand of the product being shipped and the urgency to ship the same The system excludes: 1. Type of product being shipped 2. Volume of shipments 3. Extent of involvement of the third party logistics firm that is responsible for handling operations from the origin to origin port before the shipment leaves the port The decision maker in this system is the party at the destination (San Francisco). They see a daily demand, which is uncertain, and are in contract with the supplier in China. The destination port is LA, USA. The shipment leaves a port in China and reaches the LA port after which it is transferred to San Francisco warehouse via a truck. There is no specific port in China that the shipment can originate from. It can start from Shanghai or Hong Kong or Shenzhen. The lead-time from the origin port to the destination is a sum of the following independent lead-times(lt): LT from Origin Port to destination Port LT from dwell (waiting time) at the destination port LT due to unloading at the destination port LT from inland transit from destination port to destination The party at the destination chooses to order a quantity,the current limit of which is set in the contract with the supplier. It is assumed that the contract allows for any number of orders during the time period of the contract (that is 1 year). The decision maker follows a (R, Q) order policy, where R is the reorder point and Q is the order quantity. According to this policy the decision maker orders Q units every time its inventory position goes below R units. It is also assumed that the inventory is monitored continuously that is every day of the year because demand is observed at the same frequency. The demand that is not met is all converted into lost sales (loss of customers for forever) which have a large penalty cost attached. The costs that the decision maker will incur are Annual Inventory Cost 1. Holding Cost This is the cost incurred because of excess inventory held. The decision maker is forced to hold excess inventory or safety stock because of uncertain demand and lead-time. 2. Shortage Cost- This is the cost incurred when customer demand is not met (lost sales). Again shortage cost is a result of demand and lead-time uncertainty. 5

3. Order Cost- This is a fixed cost incurred every time the decision maker orders from its supplier in China. Annual Transit Cost 1. Port to port transit cost This cost is the cost of travelling the oceanic distance between China and port of LA. This is affected by the frequency of orders and per ton bunker fuel (ship fuel) cost uncertainty. 2. Inland transit on truck to the destination- This is the cost of inland transit from port of LA to San Francisco, which is again affected by the frequency of orders and per gallon cost of truck fuel. 3. Storage Cost-A cost incurred to dwell at the destination port for more than the maximum allowable free dwelling days. Dwell time is often unplanned and seen due to congestion at the LA port on any day because of other shipments. Congestion can cause delays in customs clearance or unloading and loading operations from the ship to the truck before it is transferred to the destination. Base Case & Flexible Cases The system analysis will compare 2 flexible cases against a base case. The base case performance is evaluated using uncertainty in demand and lead-time. The flexible cases (1 & 2) are evaluated after building in decision rules into the base case simulation model. They are compared on the basis of calculation of the above-mentioned costs Adding to the base case strategy the: 1 st flexible case- Since the penalty cost incurred due to lost sales is very large as compared to the order cost, I choose to use the option of shipping the freight through air for the next time period with the penalty of an increased order cost (10 times more than that of base case!) if the lost sales in any time period is beyond a maximum lost sales allowable. Air transportation has a fixed lead-time facing the same demand uncertainty as the base case. Demand stays the same because it is not under the control of the decision maker. Details of this case are described in sections to follow. 2 nd flexible case- I could choose my contract with the supplier, to allow me to ship out a maximum of 155 units at any time which again comes with an increased order cost (set at $100 more than the base case). I choose to expand if I see lost sales >0 in the previous 2 time periods and increase my order quantity by 15 more units as compared to the last time period as long as it is less than 155 units.details of this case are described in sections to follow. Structure of Analysis In the following sections the uncertainties that affect the system as mentioned above are defined and reasoned why they are modeled using specific distributions. 6

The base case and the two flexible cases are then modeled in excel Monte Carlo simulation method to arrive at the annual inventory and transit costs which are used to compare the two scenarios. Finally the results are summarized with explanations on the observed trend seen in the VARG graph obtained from the simulation model. At this point I would like to justify the use of Monte Carlo simulation method for the purpose of evaluating the base case and the flexible cases in terms of the performance measures indicated above. Monte Carlo allows me to support a wide range of stochastic processes. Also, it does not restrict me to model only particular distributions of uncertainties into my system. Monte Carlo simulation also allows me to effectively incorporate decision rules along the simulated paths. Uncertain Factors Determination 1. Demand distribution: The distribution of customer demand is a major uncertainty factor that influences the decision to choose between the flexible and the base cases.in this case customer demand is observed everyday of the year. Approximation of the demand distribution can be a pretty difficult task when there is a lack of data. The data available with me does not give me information about the customer demand and for purposes of non-disclosure, information about the type of product is not available, hence eliminating any possibility of researching for realistic data, however the distribution of lead-time is found out to be normal. In this case the demand distribution is assumed to be normal, mainly because the calculations used below are derivable only with normal demand distributions. The demand has a mean of 110 units and a standard deviation of 20. The minimum limit is set at 0 units because demand cannot be negative at any time and is set for a maximum limit of 210, which is 5 standard deviations above the mean. The maximum limit on demand is set at very high limit from the mean to be highly conservative with my results by accounting for seasonality of the products if any. The units being sold are seasonal and 210 units of demand is probably seen during certain times of the year also called the peak seasons. 7

2. Lead Time Distribution Fig. 2:Demand distributionin a year Uncertain lead-time forces companies to make decisions before realization of demand and hence accounts for safety stock inventory levels that account for annual holding cost. More uncertain the lead-time more is the safety stock held in the warehouse. The data available with me for shipments from China to port of LAshow a distribution as indicated in the graph in Fig.3 below. Although, the real data shows that the lead times beyond 19 days are outliers, the case study assumes it to be normal between 8 and 34 days (considers the outliers too) as indicated in Fig.4. This would give us a more conservative result, since research on other data set sources indicate pretty high lead times (nearly 44 days) for LA shipments from China. In this case the minimum is set to 5 days instead of 8 days to account for - 3 standard deviations from the mean and is also a plausible data set point ascertained from real life cases of shipments from China to LA. The lead-time components of port to port (OP-DP), destination port (DP) to destination and dwell times are extracted from the data set available for shipments between China and LA port. However, unloading time at the destination port is assumed data and is made normal to match with the distributions of the above components of lead times and make calculations easier (it is easier to add 4 normal distributions instead of adding 3 normal and say may be one uniform distribution.) The total lead-time mean and standard deviation calculations are sum of the individual components standard deviation and mean because they are independent and are all normally distributed. The dwell time at the destination ports is affected by factors like loading onto the truck after unloading from the ship, custom clearance at the LA ports, congestion rate at the ports, service level of the freight forwarder trucking responsibilities from the destination port to the destination, number of service booths to help in custom clearance at the destination port and the unforeseen circumstances like natural calamities. The total transit time, which is 8

the sum of the above 4, components is therefore uncertain and is the total leadtime of the ocean transportation from China to LA. Fig. 3:Port to Port real lead time distribution Fig. 4(a):Port to Port modeled lead time distribution 9

Fig. 4(b): Total lead time modeled distribution Distribution Mean (days) Standard deviation (in days) OP-DP Normal 14 3 DP-Destination Normal 1.9 0.3 Unloading Normal 0.4 0.1 Dwell time @ Normal 5.4 1.8 destination 3. Fuel cost This case study is concerned with two types of fuel for the two legs of freight transportation. During the oceanic transportation segment, the ships use bunker fuel. The second segment, which is the inland transfer from the destination port to destination by truck, is the fuel used in cargo trucks. The fuel cost data for bunker fuel is obtained from historical data (for the first quarter of 2011). The maximum and minimum limits are the limits for the first quarter year 2011 and bunker fuel cost for any day is modeled as a random number between the set limits. The cost for truck fuel is equal to the average price of fuel in the state of California for the months Jan to Nov 2011. The price for December is assumed to be the same as that of November. 10

Fixed Factors Determination The performance evaluation of the oceanic freight transportation base and flexible cases also involves a number of fixed factors as follows: 4. Holding cost factor: It is the carrying charge, the cost of having one dollar of the item tied up in inventory for a unit time interval. ($/$/time). This is expressed as a percentage of per unit cost of an item. 5. Per Unit cost: This is the cost of a unit item, which is variable in the sense that it the total cost depends on the number of units bought. It is fixed on a per unit basis. 6. Order Cost: This cost is a fixed cost per order irrespective of the number of units being shipped. This includes the cost of labor and miscellaneous factors that are involved in shipping. In my case I have set the order cost for air to be 10 times more than that of ocean freight, which is used while evaluating the first flexible case. In case of the second flexible case where the contract allows setting a maximum limit on the number of items that the shipper can ship, the ordering cost is set at $100 more than that of the base case wherein the option is not available. 7. Average fuel consumption: This is the fixed average fuel consumption in terms of tons of bunker fuel a day for the ship and miles per gallon in case of the truck. Fig. 5:Bunker Fuel consumption vs. speed (knots) It is known that the size of containers used for ocean freight in this case is 8000 TEU and it is assumed that the ships go on the normal speed of 22 knots. Hence the graph suggests that the bunker fuel consumption in 160 tons/day. 8. Storage cost and maximum allowable free storage days at the destination port: 11

When the cargo reaches the destination port, it is allowed to dwell at the port for a maximum allowable of 4 days for free after which is charged a storage cost of $500 every day beyond 4 days. The cost of storage is obtained by research on the storage cost in port of LA. This is used to calculate the storage cost, which is a part of the total transit cost. 9. Maximum lost sales allowable: The maximum number of lost sales allowable is fixed at 110 units for the company (or the decision maker). This is a constraint used to evaluate the flexible case of shipping more units when the lost sales (or customers) are above 110 units. It is set at 110 units because it is equal to the mean of demand and thus the logic is such that lost sales beyond average of the demand observed can prove a loss for the company. 10. Lost Sales Cost: This is the penalty cost incurred by the decision maker when he loses the sale of a unit. This is set at $1000/unit, which is usually calculated by the marketing and finance department of the organization.calculating the cost of lost sales can be a pretty difficult task. Different companies follow different calculation procedures and depend on factors like estimating the type of customer lost and the effect on company profits on losing them forever. 11. Lead Time for air freight transportation: Freight transportation through air is usually not variable and is an expedited shipping option. This lead-time is assumed to be equal to 3 days in this case for China to LA shipments. It is set at 3 days after a little research on airfreight lead times between China and US shipments. This lead-time includes the total number of days until the final destination (San Francisco in this case). 12. Destination Port to Destination Distance: This is the road distance between Port of Los Angeles to the distribution center in San Francisco, which is the final destination of the shipments from China. 13. Safety Stock Factor: This is the normal distribution service factor based on desired service level. In this case the safety cost factor is 1.64, which is based on a desired customer service level of 95% as set by the decision maker. 14. Maximum Order Quantity is the maximum number of units allowable as per the contract between the shipper (the decision maker) and the supplier. In other words the shipper can ship in any number of units less than or equal to the maximum order quantity. This limit is used to exercise the decision rule of the second flexible case. Sensitivity Analysis Since the input variables are uncertain, sensitivity analysis was conducted to find out if the performance measures are affectedon changing the numerical values assigned to the factors were changed to other plausible values. Tornado diagram was used to understand 12

the importance of the effect of the uncertain factors. The minimum and the maximum values of a particular factor are plugged into the base model to find the changes in the annual inventoryand transit cost. While doing so, all other factors, besides the one whose effect is being estimated, are kept constant. For instance while finding the effect of demand on the annual inventory cost, total lead-time was set at a constant, (at the mean of the total lead time), while finding the cost for a minimum of 0 demand and 210 units. The tornado diagram is as below. Parameter Unit Low parameter High parameter Sensitivity value value Demand Units 0 210 21768.57% Lead Time Days 6.2 49 380.79% Lost sales cost $/unit 1000 1500 49.96% per unit Order Cost $ 100 700 0.2% Holding cost % 8 16 0.15% factor Fig. 6(a):Tornado Diagram for the uncertain factor determination for annual inventory cost The above diagram suggests that demand,lead-time and lost sales cost per unit have a pretty significant impact on the annual inventory cost. Clearly, demand and lead-time are important uncertainties to be kept in mind while modeling the simulation model for inventory cost. At the same time order cost and holding cost factors have small affects on the inventory cost as compared to demand and lead-time changes. Hence they can be fixed at certain values with no distribution in the simulation model. Historical data is 13

used to set the holding cost factor to 16%, which is generally used by logistics firms that handle large supply chains. However, for the lost sales parameter, it is unlikely that a firm will choose a distribution for the same. It is because this factor is based on setting at a reasonable limit based on past experience of losing customers due to unavailability of stock at hand. But the main take away from the graph with respect to the lost sales parameter is that flexible decision rules should be modeled around lost sales per unit because it has a considerable impact on the performance measure. Similarly, the tornado diagram for annual transit cost gives us an idea of the parameters that have considerable effect on the performance measure. Parameter Unit Low parameter value High parameter value Sensitivity Demand Units 0 210 704.55% Lead Time Days 6.2 49 255.53% Bunker Fuel Cost $/tons 200 500 73.18% Trucking fuel cost Storage Cost @ destination port $/gallon 3.36 40 0.92% $ 100 500 2.16% Fig. 6(b):Tornado Diagram for the uncertain factor determination for annual transit cost Again, the considerable effects of demand and lead-time are evident from the sensitivity analysis shown above. As in the case of annual inventory cost, transit cost is also affected 14

by uncertain lead-time and demand. Bunker fuel cost has a significant impact too and hence is given a distribution such that it generated random numbers between a set limit, which are obtained from historical data as indicated before. It is important to note the effect of trucking fuel on the transit cost, the maximum limit that I have used for sensitivity analysis is highly unrealistic (40 $/gallon) but it is meant to show the fact that the effects of trucking fuel cost are not drastic, although the inland transportation cost constitutes a significant part of the annual transit cost. In order to be as close to real life as possible I have set the trucking fuel cost to be the average of the trucking fuel for the month for the state of California (the destination state) for the year 2011, as described before. Change in storage cost per day beyond the allowable number of free days again does not have a significant impact on the transit cost but historical data suggests that storage cost is set at $500 in port of LA. Simulation Simulation is used to evaluate the performance of the base case and the two flexible cases. The first flexible case and the base case are compared on the basis of the annual inventory costs. The case with the lower annual inventory costs and lower annual transit costs is better than the other. The simulation model will include the uncertain and the fixed factors explained above to arrive at the annual inventory and transit costs. The input/entries for the simulation model are described below: Factor Demand Origin Destination Port Lead Time Destination port to destination lead time Destination port dwell time Unloading time Lost Sales Cost Holding cost factor Fixed/ Variabl e Varia ble Varia ble Varia ble Varia ble Varia ble Fixed Fixed Distributio n Mean StdD ev. Minim um Maxim um Unit Normal 110 20 0 210 Unit s Normal 14 3 5 34 Days Normal 1.9 0.3 1 3 Days Normal 5.4 1.8 0.1 11 Days Normal 0.4 0.1 0.1 1.0 Days $/uni t $/uni t/day Value (if fixed) 1000 16% 15

Per unit cost Fixed $ 50 Order cost Fixed $ 500 for ocean freight Order cost Fixed $ 5000 for air freight Average Fixed Tons 160 bunker fuel consumption /day Average fuel consumption by truck Fixed mpg 25 Bunker fuel Varia Random 200 500 $/ton cost Storage cost at the destination port Number of days allowed to store free of cost Maximum lost sales allowable Total lead time through air transportatio n Maximum Order Qty Destination port to Destination distance Notes ble Fixed $/da y Fixed Days 4 Fixed Unit s Fixed Days 3 Fixed Fixed Unit s Mile s There can be other factors like exploring the possibility of transferring shipping responsibility to the freight forwarders (party responsible for shipping) that can be 500 110 155 350 16

included in the evaluation of the annual inventory cost and annual transit cost. However, this analysis will be a stepping stone for more detailed levels in the future long term project. Since contractual terms for the maximum allowable order quantity and maximum allowable lost sales vary according to different organizations (or decision makers), I have assumed their values for the current decision maker and also assumed that they serve the demand 365 days a year. Calculations Before explaining the calculation of the annual inventory and the transit costs, I will include the calculations for the factors used to calculate the former. They are listed below. 1. Order Quantity (Q)- Starting order quantity is the Economic order quantity (EOQ). EOQ is often the starting point of order quantity in case of uncertain demand and lead-time scenarios because it is proven to minimize the inventory cost. Where: A = Order Cost E(D) = Expectation of the Demand (Mean) v = per unit cost r = holding cost as % of per unit cost 2. Reorder point(r)- is the point when an order is placed by the shipper (in this case the decision maker), if the inventory position is below this point. The formula below is again proven to be an optimized reorder point in terms of inventory cost. In case of uncertain demand and fixed lead time, as in the case of airfreight mode for the first flexible case Where: 17

E(L) = expected lead time =sum of mean of individual lead times because they are independent of each other E(D) = Expected Demand = Variance of Demand = Variance of Lead time = sum of variance of individual paths because they are independent of each other = Safety stock factor 3. Per Unit Holding Cost (HC)- is the per unit cost of holding an unit in inventory which accounts for the holding cost for the annual inventory cost. Where: r = holding cost per unit item per unit time as a percentage of the cost of one item (v) v = cost of per unit item 4. Short Cost (SC)- is the per unit cost of losing the sale of the items, which accounts for the shortage costs for the annual inventory costs. where: l = cost per unit lost sale (here it is $1000) endinv = ending inventory evaluated at the end of every day Simulation Method The following will describe the steps used to set up the base and flexible cases along with the evaluation of the performance metrics-annual Inventory Cost and Annual Transit Cost. The two cases are compared on basis of VARG curves, which are a result of simulation. Simulation Model for base case 1. The order quantity (Q) and reorder point(r) is calculated using the formula mentioned above in calculation section. 2. The initial inventory is set to be equal to the order quantity. It is assumed that an initial inventory equal to the order quantity is passed on to the beginning of the current time period. The values for the base case are as follows: 18

No. of units Order Qty 117 Reorder point 3337 Initial inventory 117 For Day 1: 3. Generate lead-time for the 4 segments of lead time, namely Origin port to destination port (OP-DP LT), destination port dwell (DPD), destination port to destination (DP-D LT) and unloading at destination port(udp) lead times using the mean and standard deviation as stated in the entries / input section using the following excel command: MIN(MAX((NORMINV(RAND(),Mean,Std.Dev)),Minimum limit),maximum limit) Total Lead time = (OP-DP LT) + DPD + (DP-D LT) + (UDP) It is important to note here that the lead-time segments are independent of each other and can be added for the mean and standard deviation of the total lead-time because they are all normally distributed and independent. 4. Beginning inventory is equal to ending inventory of the previous day 5. Generate Demand for the day using the following excel statement: MIN(MAX((NORMINV(RAND(),Mean,Std.Dev)),Minimum limit),maximum limit) Where the mean, standard deviation and the limits are 110, 20, 0 and 210 respectively. 6. Orders Received is a binary variable, which can either be true or false. Orders received are True if there are any orders to be received this current day. We know that orders are to be received if the orders are placed on any day before the current day and the time when the order is due of any of the previous days matches the day number of the current day, using the following statement: NOT(ISERROR(MATCH(Current day, Time due on day 1:Time due (current day- 1),1))) For instance, Orders received for Day 1 is set to False. Orders received for day 2 is True if Orders were placed on Day 1 and the total lead-time for day 1 is equal to 1 day. 19

7. Units Received is equal to the order quantity if there are any orders received this current day. (Or Orders received is True) 8. Ending Inventory is equal to MAX(0,Beginning Inventory Demand for the day + Units Received) 9. Lost Sales is equal to IF(Demand>BeginningInv + UnitsRecd,IF(Beginning Inv+unitsRecd>=0,Demand-Beg. Inv- Units Recd,Demand),0) 10. Orders Placed is again a binary variable, which can be either True or False. Orders Placed is true if : Ending Inventory < = Reorder Point 11. Time the order is Due is IF(Orders are placed,currentday+total lead time of the current day+1,) 12. Holding Cost is equal to MAX(0,Ending Inventory*holding cost per unit item) 13. Order Cost is equal to If (Orders are placed, Ordering Cost, 0) 14. Short Cost is equal to : Lost Sales * per unit cost of lost sales 15. Total Inventory Cost for the day is equal to Holding Cost for the day + Short Cost for the day + Lost Sales Cost 16. Generate Bunker fuel cost for the day RANDBETWEEN (minimum fuel cost, maximum fuel cost) 17. Trucking fuel cost is known from historical data for 2011. The average cost per Month is used as tabulated below: Cost( $/gall on) Ja n 3.3 56 Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec 3.49 1 3.963 4.2 4.213 3.95 3 3.82 4 3.785 3.946 3.86 5 3.834 3.834 18. Port to Port Transit Cost 20

If(Orders are placed, (OP-DP LT * number of tons of bunker fuel consumed/day * fuel cost for the day),0) 19. Destination Port to Destination Transit Cost is If(Orders are placed, (Destination port to destination distance (in miles)/fuel consumption by truck(milespergallon) * cost of fuel per gallon),0) 20. Storage Cost at the destination port If (dwell time > maximum number of allowable free days, cost of storage per day* (dwell time- maximum number of allowable free days),0) 21. Total transit cost is equal to Port to port transit cost + Destination port to destination transit cost + Storage Cost 22. The above steps are repeated 365 times for annual inventory and transit costs, which is the sum of inventory and transit cost per day for the entire year. Inventory Cost as a Performance Measure Base Case The base case considers that the decision maker is observing an uncertain demand each day and is also forced to deal with uncertain lead time factors leading to lost sales and safety stock inventory levels. It has no option of increasing the order quantity or choosing an expedited shipping method like airfreight transportation to avoid lost sales for the day. The decision maker incurs the total inventory and total transit cost as per the contractual terms. The calculations for the annual inventory and transit costs are done as described in the Calculations section above. The annual inventory cost is affected by uncertainty in demand and lead-time. Monte Carlo Simulation is run for 1000 times, results of which are summarized below. 21

Fig. 7 :Sample Base Case spreadsheet with the annual inventory cost Results of Simulation for Base Case Value in million $ Expected Annual Inventory Cost 28.14 Minimum Annual Inventory Cost 23.65 Maximum Annual Inventory Cost 32.68 Standard Deviation of annual 1.40 inventory cost 22

Fig. 8:VARG curve for Base Case Fig. 9:Histogram for Base Case Flexible Case I will be analyzing two flexible cases with different decision rules. The ocean freight transportation has potential for a number of flexible cases; the two mentioned below are the cases, which I think are suitable to be implemented if I were the decision maker in real life. The two flexible cases are called flexible Case1 and flexible Case 2. It is 23

important to note there that the two flexible cases are compared against the base case but not compared with each other. Flexible Case1 Transit time, demand variability and the associated unreliability impose the highest penalty in form of service failures leading to short sales. In real life, as companies keep increasing their sourcing from countries like China, length of supply chains keep increasing and problems associated with them keep getting worse. One coping mechanism is to use airfreight, which usually comes at an added ordering cost, in this case it is 10 times more than the base case ordering cost. Since it comes with an added cost, if lost sales keep recurring the decision maker opts for airfreight. The specification of the circumstances in which airfreight is chosen is under the decision maker s control. The advantage of using airfreight is certain (or in some real life cases a very low variability) lead-time. It is assumed to be a constant of 3 days as obtained from research. However, since it comes with an added cost,the purpose of the simulation is used to decide if a trade off is advantageous to the decision maker economically. The simulation model is designed to take maximum benefit of the flexible design by reducing the risk of uncertain lead-time. Demand uncertainty is not under the control of the decision maker and hence he chooses to concentrate on uncertain lead-time. In my case the decision maker choses to use air freight transportation if the lost sales in any time period is greater than maximum allowable lost sales in the particular time period. It is important to set a maximum limit on the lost sales allowable because it comes with a large penalty cost and also affects the image of the company in the present and future time periods. In this case the decision maker choses to set the maximum allowable lost sales in any time period to be 110 units. This decision for shifting to air mode once made, is, not a one time decision, implying that the decision maker can choose to go back to ocean freight transportation when lost sales are observed to be less than the maximum allowable lost sales in the following time periods after the decision is made to use air mode. Similarly, it is assumed that the decision maker can choose airfreight mode any number of times in the time period in question (1 year). For the purposes of simplistic simulation model, I have assumed that the decision maker decides to ship in all the units using airfreight in the anticipation of lost sales in day x to be more than the maximum allowable lost sales if the lost sales in day (x-1) is greater than maximum allowable lost sales. In other words, since the decision maker is seeing an uncertain demand, he will anticipate that the demand observed the following day would lead to lost sales more than maximum allowable if a similar pattern is seen the day before. The inventory cost for the flexible case on any day is the inventory cost due to airfreight transportation mode if airfreight is chosen for the current day or is equal to the inventory cost due to ocean freight if the freight is shipped via ocean. A binary variable is used to indicate if implementing the decision rule as stated above uses airfreight. The annual 24

inventory cost is then added for the 365 days in a year and compared against that of base case, which uses only ocean freight transportation. The Decision rule used in the simulation model is as stated below: For time t and maximum lost sales allowable =110: If lost sales in t >110 If No Air = True If Yes Ship order using airfreight Inventory cost (t) = air freight inventory cost Inventory cost (t) = ocean freight inventory cost 25

Fig. 10(a): Sample Flexible Case 1 spreadsheet with the annual inventory cost based on airfreight or ocean mode for every day 26

Fig. 10(b):Sample Flexible Case 1 spreadsheet with the annual inventory cost based on airfreight or ocean mode for every day Fig 10(a) can be used to demonstrate the implementation of the decision rule for flexible case1. The binary variable that we are concerned with is the column named Air (Y/N). For the day 1 (first row), where the value of lost sales with the base case is 0units, exercising the decision rule results tell the decision maker not to use the air option because the lost sales units is less than maximum allowable lost sales (0<110). However, day 7 (row 7) sets the binary variable to True. It is because the number of lost sales units is greater than maximum allowable units (164>110 units). Results of Simulation for Flexible Case 1 Value in million $ Expected Annual Inventory Cost 14.64 Minimum Annual Inventory Cost 12.17 Maximum Annual Inventory Cost 17.38 Standard Deviation of annual 0.84 inventory cost 27

Fig. 11 :VARG curve for Flexible Case 1 Fig. 12 :Histogram for Flexible Case 1 Base Case vs Flexible Case1 * 28

Value in million $ Base Case Flexible Case1 Expected Annual 28.14 14.64 Inventory Cost Minimum Annual 23.65 12.17 Inventory Cost Maximum Annual 32.68 17.38 Inventory Cost Standard Deviation of 1.40 0.84 annual inventory cost P 10 26.40 13.55 P 90 29.94 15.75 *(the values marked in yellow are the favorable outcomes) Fig. 13: VARG curve for Flexible case1 and Base case. The penalty cost that I incur due to lost sales is 1000 per unit whereas the ordering cost for airfreight shipment is $5000(is10 times more than ocean shipping). In the base case cost due to lost sales account for a large percentage of the total inventory cost, which are due touncertain demand and uncertain lead-time. On getting the air option I make my system flexible and hence the total inventory costs reduce because shortage cost reduces much more as compared to increase in the ordering cost and holding cost. Numbers from one simulation indicates that while the holding and ordering costs increase, the shortage 29

cost in the flexible option is 97% less than that of base case, which is about 28 million USD. This improvement in shortage cost outweighs increase in cost due to holding cost and ordering cost. The simulation results show that the flexible case is stochastically dominant and is preferred over the base case always. This flexible design reduces the downside risk by reducing / eliminating the lost sales and also takes advantage of the opportunities (in this case reduced holding cost because of certain lead time factor instances as compared to always procuring unitsvia highly uncertain ocean transportation). The histograms of the flexible case and the base cases show that the base case inventory costs have more tendencies to be towards higher costsas compared to the flexible case. (Base case between 26.81-29.97; flexible case between 14.00-15.56). Since our objective is to minimize the costs flexible case looks better than the base case, on comparing the above histograms. Flexible Case2 Due to highly uncertain demand and lead time pattern in this case the decision maker can use real options contract for very long supply chain from China to USA like the one being discussed here.it helps the decision maker hedge the risk of lost sales due to uncertainties. The decision maker (the shipper) will have the right but not an obligation to ship in a maximum of 155 units. Sensitivity analysis is used to arrive at the maximum allowable order quantity in terms of its sensitivity to the inventory cost. The results are described in the following section. Exercising the right allows the decision maker to expand based on a certain decision rule. In this case the decision rule is based on observing lost sales over 2 consecutive days before the current day and the current order quantity is less than the maximum order quantity allowable under the contract, the decision maker has the option of increasing the order quantity by 15 units (or expanding ). The advantage of this option is reducing the lost sales to the maximum extent possible, sometimes even to 0. However, it comes with a cost of having the option, which is $100 more than the base case in terms of the ordering cost. The decision maker has to trade off between the two options to settle on the base or the flexible case. The decision maker is allowed to expand any number of times as long as the condition of maximum allowable order quantity is satisfied, that is the order quantity after expanding is less than or equal to the maximum allowable order quantity under the options contract. If the option is exercised, the order quantity remains the same for the rest of the following days until the option is exercised again. In other words, the simulation model is set such that the order quantity in day x is greater than or equal to the order quantity in day (x-1). The inventory costs are calculated using the same calculations as described before. 30

Annual Inventory costs and annual transit costs of the base case and the flexible cases are now compared. Sensitivity Analysis to determine maximum order quantity allowable Sensitivity analysis was done to determine the maximum order quantity allowable since there is no data available on the same. Assuming 165 units as the maximum allowable order quantity, the flexible and the base cases were analyzed. Although it gave the desired results (flexible case 2 being stochastically dominant), changing the order quantity limit to 155 units gave even better results. Also it was interesting to note that the lost sales with 155 units were less than setting the limit anywhere between 130-145 units. So 155 units gave a lower cost and lower lost sales on average. In real life, managers like the decision maker in the current case, base their analysis on the inventory cost. Hence the difference between average annual inventory cost obtained from simulation of the base case and the flexible case 2 suggested maximum cost difference (maximum value) for a maximum order quantity of 155 units. In the figure below a positive cost difference implies annual inventory cost of flexible case is better(lower) than the base case. Fig. 14: Sensitivity Analysis to determine maximum order quantity The analysis shows that beyond 170 units the flexible case 2 starts performing worse than the base case because at this point the holding cost overcomes the shortage and ordering cost.it is because increasing the order quantity reduces the lost sales, constantly reducing 31

shortage cost and thus for every order cycle a lot more units are ordered than required. A point after which the lost sales goes to 0, shortage costs no longer have a bearing on the total annual inventory cost and hence ordering just the optimum amount enough to counter the lost sales is the answer. For dayt, order quantity Q and maximum order quantity allowable under the options contract =155: If lost sales in (t 2) AND (t 1)>0 AND Q(t 1) <155 If No Expand If Yes If Q(t 1) < = 155 If Yes If No Q(t) = Q(t 1) Q(t) = Q(t 1)+ 15 Ship current order using new order quantity Calculate New inventory Cost Inventory cost (t) = base case order quantity inventory cost 32

Fig. 15: Sample Flexible Case spreadsheet with the annual inventory cost In order to demonstrate how the flexibility rule works, we could take a look at snapshot of the spreadsheet for the flexible case above (Fig.15). Under the column named expand, the 4 th day suggests that the decision maker should expand. It is because the lost sales column (column to the left) indicates a lost sale of 81 and 75 units for day 2 and 3 (2 time periods before 4) which are greater than 0 units. However, for day 2, the model suggests that there should be no expansion because day 1 fortunately did not see any lost sale and hence only one day between day 1 and 2 saw a positive lost sale unit. Also, after the order quantity (column to the right of expand) reaches 143 units from the starting point of 128 units, the model restricts any more expansion even though the model suggests expansion because 143 +15 =158 units is greater than 155 units, which is the maximum number of units allowable under contract between the decision maker and the shipper. Results of Simulation for Flexible Case 2 With maximum order quantity = 155 units 33

Value in million $ Expected Annual Inventory Cost 18.01 Minimum Annual Inventory Cost 15.74 Maximum Annual Inventory Cost 19.90 Standard Deviation of annual 0.61 inventory cost Fig. 16(a1): VARG curve for Flexible case2 (155 units) Fig. 16(a2): Histogram for the annual inventory costs in flexible case2 (155 units) 34

For the purposes of comparison and to understand the results of the sensitivity analysis the inventory cost corresponding to 165 units are tabulated below: Value in million $ Expected Annual Inventory Cost 25.09 Minimum Annual Inventory Cost 23.25 Maximum Annual Inventory Cost 26.96 Standard Deviation of annual 0.62 inventory cost Fig. 16(b1): VARG curve for Flexible case2 35

Fig. 16(b2): Histogram for the annual inventory costs in flexible case2 Base Case vs Flexible Case2 Value in million $ Base Case Flexible Case2 Expected Annual 28.14 18.01 Inventory Cost Minimum Annual 23.65 15.74 Inventory Cost Maximum Annual 32.68 19.90 Inventory Cost Standard Deviation of 1.40 0.61 annual inventory cost P 10 26.40 17.25 P 90 29.94 18.79 36

Fig. 17: VARG curve for Flexible case2 and Base case. Flexibility of having the option of expanding or increasing the order quantity based on a decision rule of lost sales comes with a cost premium. However the results above justify the added cost by benefitting in inventory costs (and later as we will see in transit costs too). The cost of lost sales of one unit is much more ($1000) as compared to added costs of $100. Although the costs are not directly comparable, the scale difference speaks a lot about the effect of lost sales on the cost. Hence the flexibility rule can be seen as an insurance against lost sales outcomes. In case of the flexible case, increasing the order quantity in light of uncertain demand and lead-time reduces the shortage cost part of the annual inventory cost. This benefit again outweighed the increase in cost of ordering. Also, increasing the order quantity reduces the number of instances of orders placed. On comparing the results, the flexible option is a more favorable decision when compared to the base case. On comparing the expected annual inventory cost, the company reduces the cost by about 10 million USD by exercising the flexibility of increasing the order quantity. VARG curve comparison of the base case and the flexible case analysis of extreme values can reveal interesting results to the decision maker. It also shows that the curve for the base case reaches higher values of cost faster than the flexible case, thus proving that the flexible case is better than the base case. The VARG curve shows that the flexible case is better than the base case under all circumstances and is stochastically dominant (since the X axis is the cost the graph to the left is better performing than the one on the 37

right). Also the flexible has a lower annual inventory cost or greater value (lower inventory cost) at the P 90 & P 10. Other parameters like the maximum cost and the standard deviation reiterates the fact that the flexible case is better than the base case and helps the decision maker to be completely sure when investing in the option. Transit Cost as a Performance Measure Base Case Transportation costs, I believe, is an important measure for the performance of the ocean transportation system. This is the responsibility of the shipper if it owns the ships that are used to ship the products. In many other cases the shipper transfers the responsibility to the freight forwarder who owns the ships and the shipper pays a transportation cost which is added to the fixed order cost. In the case being discussed here, the former case is used which implies that the decision maker is responsible for the transit cost. This is turn implies that the decision maker is affected by uncertainties in fuel cost. The transit cost as discussed earlier constitutes three parts, oceanic transit fuel cost, in land transit fuel and the storage cost at the destination port because of the dwell time exceeding 4 days, which is the maximum number of days allowed to be stored for free in the destination port. Fuel used in ships or bunker fuel is modeled to be random numbers between 200 and 500 USD/ton and the cost for truck fuel is set as the average price for each month. Similar to the simulation for annual inventory cost, 1000 simulations are run for the base case and the flexible case to compare the annual transit cost. Calculations for the cost are described in the section above. Again the case with the lower transit cost is valued superior to the other. Results of Simulation for Base Case Value in million $ Expected Annual Transit Cost 286.22 Minimum Annual Transit Cost 266.03 Maximum Annual Transit Cost 301.93 Standard Deviation of annual Transit 5.02 cost 38

Fig. 18: VARG curve for Base Casefor annual transit cost Flexible Case 2 Fig. 19: Histogram for Base Casefor annual transit cost Besides affecting the annual inventory cost, the real options flexibility in case of the second flexible case has effects on the transit cost. The change in order quantity changes the number of orders that are placed and hence this frequency is responsible for a change in the annual transit cost as compared to the base case. This is known to indirectly affect the storage costs at the destination port because the lead-time, which is different for different days when the order is placed, leads to a different storage cost. The same logic applies to all the 3 parts of the annual transit cost. The simulation decision rule is the same as that described in the section for decision rule for inventory costs for flexible case 2. The transit costs are updated similar to the inventory costs. The transit costs are independent of the number of units being shipped. 39

I have not compared flexible case 1 and base case transit cost because of the unavailability of information on the airfreight fuel costs. Results of Simulation for Flexible Case 2 Value in million $ Expected Annual Transit Cost 210.80 Minimum Annual Transit Cost 195.64 Maximum Annual Transit Cost 227.85 Standard Deviation of annual Transit 4.79 cost Fig. 20 : VARG curve for Flexible Case2 for annual transit cost 40

Fig. 21 : Histogram for Flexible Case2 for annual transit cost Base Case vs Flexible Case2 Value in million $ Base Case Flexible Case2 Expected Annual Transit 286.22 210.80 Cost Minimum Annual Transit 266.03 195.64 Cost Maximum Annual Transit Cost 301.93 227.85 Standard Deviation of 5.02 4.79 annual Transit cost P 10 253.84 195.64 P 90 301.93 227.85 41