A Cross-Docking Approach for Farfetch Global Delivery

FACULDADE DE ENGENHARIA DA UNIVERSIDADE DO PORTO A Cross-Docking Approach for Farfetch Global Delivery Filipa Coelho Nunes Mestrado Integrado em Engenharia Eletrotécnica e de Computadores Supervisor: Pedro João Co-supervisor: António Almeida Pedro Borges July 25, 2018 c Filipa Coelho Nunes, 2018 Resumo AFarfetch não possui qualquer stock de artigos, pois tudo o que vende na sua plataforma online vem de parceiros. Com este tipo de negócio, quando uma encomenda é composta por artigos vendidos por parceiros diferentes, a empresa é incapaz de enviar os artigos todos juntos para o cliente. Este tipo de envio afeta a satisfação do cliente e a empresa constatou que precisa de uma forma de resolver este problema. O presente projeto pretende estudar uma estratégia logística chamada de cross-docking. Esta estratégia logística é algo novo para a retalhista, e é importante identificar quais as variáveis com mais impacto nesta estratégia, construir um modelo que permita perceber qual o impacto no processo de envio de encomendas ao usar cross-docking, e quais são as principais limitações da Farftech em relação a implementação desta estratégia logística. Neste projeto são estudadas duas situações diferentes, onde aplicar a estratégia referida pode trazer vantagens. A primeira situação foca-se em fazer uma análise de como tem de ser o sistema de processamento das encomendas de forma a suportar cross-docking e uma simulação, capaz de expressar o que acontece dentro de um cross-dock, alimentada com dados reais, fornecidos pela empresa, e que se foca no mercado Chinês, e que tem como objetivo ganhar conhecimento sobre as operações involvidas em crossdock e sobre os aspetos físicos do edifício. A segunda situação consiste numa simulação, com o objetivo de dar uma perspetiva global ao cross-docking, olhando para o processo de entrega de uma encomenda a um cliente, desde o momento em que o artigo saí da loja até chegar ao seu destino final, usando, também, dados reais. A simulação do segundo caso representa o fluxo de encomendas entre a Europa e os Estados Unidos da América e representa dois cross-docks, um na Europa e outro nos EUA, sendo que é neste segundo onde acontece a agregação das encomendas. Cada caso tem os seus resultados, criados depois de correr a simulação. Para o primeiro caso, a variável usada para analisar o impacto da estratégia logística é o tempo que as encomendas passam dentro do edifício. Depois de inúmeras simulações com diferentes configurações, os resultados obtidos são satisfatórios, pois o tempo adicionado, devido à passagem das encomendas pelo cross-dock, para a maioria das encomendas, não é suficiente para baixar a qualidade geral do serviço de envio, mas é preciso estudar todo o processo para estabelecer o que acontece quando todo o processo de shipping é tido em conta. O segundo caso, não só olha para o tempo que as encomendas passam dentro dos edifícios, como também usa o tempo em trânsito de cada encomenda para avaliar o impacto de aplicar cross-docking nas operações da Farfetch. O tempo que as encomendas passam nos edifícios é bastante satisfatório e é melhor que aquele obtido no primeiro caso. Para analisar o tempo em trânsito, a companhia forneceu os tempos usuais desta variável para cada estado Americano e para todas as rotas entre Estados Unidos e todos os países Europeus. Os valores obtidos de tempo em trânsito, na simulação, são mais baixos quando comparados com aqueles que foram fornecidos pela Farfetch provando que, mesmo que mais variáveis sejam adicionadas ao modelo, como tempo de desalfandegamento, usar cross-docking nunca terá um impacto capaz de baixar a satisfação do cliente no que diz respeito ao tempo que este espera para receber os produtos. i ii Abstract Farfetch does not hold stock, everything that its sold on farftech.com comes from partners. With this type of business, when an order is composed of articles sold by different boutiques, Farfetch is unable to send all the items together to the client. The method of shipping impacts the client’s satisfaction and the company realized that it needs to have a way to solve this problem. The current project seeks to study a logistic strategy called cross-docking. Cross-docking is something new for the company, and is important to identify the variables with the most impact in cross-docking, build a model that can be used to understand the impact of this strategy in the shipping process, and what are Farfetch’s limitations on implementing this strategy. The project studies two different situations were cross-docking can bring advantages. The first situation focuses on doing a To-Be analysis of the company’s order processing system and a simulation, capable of expressing what happens in the cross-dock with the aim of gaining knowledge on the operations involved in cross-docking and the physical aspects of the facility, that runs with real data provided by the e-commerce company and looks at the Chinese market. The second situation consists of a simulation with the aim of giving a global perspective to cross-docking, looking to the process of delivering an order to a client, since the moment of the article’s departure from the boutique until arriving at its final destination, also running with real data. The simulation for the second case represents the flow of orders between Europe and the United States of America and has two facilities, one in Europe and another in the USA, where the aggregation of the orders happens. Each case has its results, created after running the simulation. For the first case, the variable used to analyze the impact of cross-docking is the time span of the boxes inside the cross-dock. After running the simulation with different configurations, the results obtained were satisfactory and the time added due to going through the facility, for the majority of the boxes, is not high enough to decrease the overall quality of the shipping service, but there is a need to study the whole process to see what happens when the whole shipping process is taken into account. The second case, apart from the time span in the cross-dock, also uses the time in transit of each box to evaluate the impact of applying cross-docking in Farfetch’s operations. The time span of the boxes inside the facilities is very satisfactory and is even better than the one obtained in the first case. To analyze the time in transit, the company provided the usual time in transit for each American state and for the routes between European countries and the USA. The values for this variable, given by the simulation, are lower when compared to what was given by Farfetch proving that, even if more variables are added to the model, like clearance times, cross-docking will never have an impact capable of decreasing the satisfaction of the client regarding the time that it has to wait to receive its products. iii iv Acknowledgements The conclusion of an academic course marks the end of a long journey, in my case, started five years ago, and I need to show my gratitude towards the people that marked this journey. First, and because she was always with me, I need to thank my mom for the love, infinite support and for always believing in me. To my brother and my father, thank you for the affection and for always caring. To all the friends that I made in FEUP, thank you for everything we shared since we meet, and to my other group of friends, I need to show my gratitude for always having patience and for the comprehension. I need to give a special thanks to José Pedro Gomes, because, apart from being a great friend, he was my partner in every assignment, where we had the chance to choose who do we wanted to work with, and I could not ask for a better partner. To Engineer Pedro João, thank you for the guidance and accessibility demonstrated during the realization of this project. I was lucky enough to spend my last months, while a student, in Farfetch and I need to thank everyone that I had the pleasure to meet during this months, but I need to enhance my gratitude towards some people. To Sílvia Rocha, thank you for the kindness and for the concern with my project. To Pedro Borges, thank you for the interest, the opinions about my work and the sympathy. Lastly, but not least, I need to thank António Almeida, for all the ideas, the feedback, the availability, and, again, the sympathy and the interest demonstrated in my work. I appreciate everything that you did during this past months and I cannot thank all of you enough for the help, that was beyond what I was expecting. Filipa Coelho Nunes v vi "Work hard in silence. Let your success be your noise." Frank Ocean vii xiv LIST OF TABLES Abbreviations B2B Business to Business B2C Business to Consumer C2C Customer to Customer EDD Estimated Delivery Date FA Firefly Algorithm GLNPSO Particle Swarm Optimization with Multiple Social Learning Structures HSA Hybrid Simulated Annealing IXD Inbound Cross Dock JiT Just-in-Time LTL Less-than-Truckload OMS Farfetch Order Management Service OVRP Open Vehicle Routing Problem OVRPCD Open Vehicle Routing Problem with Cross-docking PDP Pick-up-and-Delivery Problems Post-C Post-distribution Cross-docking Pre-C Pre-distribution Cross-docking PSO Particle Swarm Optimization SA Simulated Annealing SIT Atlas’ label SHIPS_INDIRECTLY_TO TL Truck-Load TPL Third Party Logistic VRPCD Vehicle Routing with Cross-Docking WMS Warehouse Management System xv Chapter 1 Introduction 1.1 The Company Farfetch is an e-commerce platform funded by a Portuguese entrepreneur, José Neves, in 2008. The idea of the company emerged during a trip to Paris Fashion Week when José realized that independent high-fashion boutiques needed a place in the online market. Neves was involved in the world of fashion before Farfetch. His grandfather owned a shoe factory and, in the mid-1990s, José launched the shoe design business Swear. Later in 2001, he created B Store, a company dedicated to selling a range of up-and-coming designer labels in a physical store. José was trying to wholesale the B store brand during the trip to Paris and realized that boutiques owners were going through a rough moment and did not have the expertise on how to enter the online market. They knew that could not rely on local shoppers because their businesses targeted a highly specific customer segment and, with his idea, Neves wanted to give the boutiques support to enter the e-commerce market, but, at the same time, allow them to retain their values and beliefs. The first round of funding only came on July 9 of 2010, with Advent Partners investing $4.5M in the e-commerce platform and, until this point, José was funding his company with personal funds. The investment helped expand the business market to Europe, North America, and to Brazil. Apart from Advent Partners, as of May 31 of 2018, Farfetch has eight more lead investors, Index Ventures, Condé Nast, Vitruvian Partners, DST Global, Eurazeo, IDG Capital Partners and JD.com. In total, the company of José Neves has eight rounds of fundings and sixteen investors. The second round of investment was on January 15 of 2012, and injected $18M in the ecommerce platform, with Index Ventures being the lead investor of this round. Condé Nast came has an investor on the third round of funding, that resulted in $20M invested in the Portuguese’s company on March 3 of 2013. The next round of funding came on May 1 of 2014, with Vitruvian Partners entering as an investor, and resulted in $66M injected in the fashion business, and on March 4 of 2015 DST Global led the fifth round, that brought $86M to Farfetch. 1 2Introduction On May 4 of 2016, two new investors came to the online platform, Eurazeo and IDG Capital Partners. On this date, two rounds of funding happened, the sixth had the two investors mentioned as lead investors and brought $110M to the company, and the seventh round only had Eurazeo as an investor and brought a total amount of $20M into the business. The last round of investment came on June 21 of 2017, with JD.com as the only investor, and resulted on $397M injected into the company of the Portuguese entrepreneur. Apart from farfetch.com, Farfetch owns Browns, a London Boutique, acquired in May of 2015. Before the purchase, Browns was already a Farfetch partner, but with the acquisition of the first physical, comes the opportunity to innovate the physical shopping experience. In 2015, is announced the project named Store of the Future and, in 2017, José Neves unveil some details about it. The project aims to give an augmented reality to the client and completely change the way of purchasing in physical stores, and it is planned to launch later in 2018, in Browns and Thom Browne, a store in New York. Farfetch also provides a service called Black & White, that allows other brands to use features of farfetch.com platform on their websites. Brands that want a Black & White service can choose the design of their website but can trust their day-to-day operations to Farfetch. The company also has a unique delivery service called F90. F90 allows a customer to buy and in 1 hour and 30 minutes later, the package arrives at the destination that the client chose at the moment of the purchase. This service is exclusive to Gucci purchases and is available in 10 cities across the globe, London, Paris, Madrid, Milan, New York City, Los Angeles, Miami, Dubai, Tokyo, and São Paulo. On the operational side, Farfetch has offices in eleven cities across the world, London, New York, Los Angeles, Porto, Guimarães, Lisbon, São Paulo, Shanghai, Moscow, Hong Kong and Tokyo, with over 1500 employees, and sells products from over 700 partners, including boutiques, dispersed all over the globe, and brands. Farfetch is a multicultural company and assents on the values of Be Human, Be Brilliant, Todos Juntos, Be Revolutionary, Think Global and Amaze Customers day to day. 1.2 Objectives Farfetch has a very particular business model because it does not keep stock of merchandise, everything sold on farfetch.com comes from partners and, although this constitutes a main advantage to the company because it allows offering the clients a vast range of articles, it is also one of the principal disadvantages. It is a disadvantage because Farfetch partners are around the globe, for example, one client that lives in Portugal buys two articles, one sent from a boutique in Italy, and the second one leaves from a British partner. In this circumstances, the customer receives two packages and, most of the times, they will arrive at different times, which impacts the client satisfaction. 1.3 Methodology 3 To increase customer satisfaction, Farfetch needs to be able to consolidate the packages, without decreasing the quality of the overall service, which means that the consolidation of the order cannot have a great impact on the shipping’s price, and in the delivery time, and this need originated the theme of this dissertation. The principal goal of this research project is to understand in which situations a cross-docking approach can increase the client’s satisfaction and what impact will this logistic strategy have in Farfetch and in the shipping service. In order to reach the main objective, is important to understand the requirements to implement a cross-docking solution and what are the factors that need to be evaluated to correctly analyze the problem of Farfetch’s shipping operations and, to achieve the principal objective, the project will focus on: •Understanding the operations involved in cross-docking: Inside the cross-dock, the principal operation is the aggregation of products. To avoid rising the shipping’s costs and negatively impact the overall shipping process, in the client’s perspective, the Estimated Delivery Date (EDD) of each product needs to be analyzed and needs to be established a maximum value for the gap, between the arrival of the products, that allows having aggregation. Another case of interest is the place where the aggregation should happen, in the case of having multiple facilities. To gain knowledge about cross-docking is crucial to do a literature review on what is already done. •Identifying and analyzing the variables with impact on a possible cross-docking strategy: The first step to start this project is to identify what are the variables with possible impact on a cross-docking strategy. For instance, the number of packages that will arrive at the cross-dock in each day will influence the capacity of the cross-dock. •Designing a cross-docking strategy: After the analyzes mentioned above, there is enough data to start building a strategy to implement a cross-docking approach in Farfetch’s warehouses and, to evaluate the influence of the developed solution, there is a need to develop a model capable of simulating this new strategy. •Understanding Farfetch’s limitations: Is imperative to understand Farfetch’s current platforms and what are the limitations on a possible cross-dock implementation. 1.3 Methodology Due to the complexity of the project, at the beginning of it, a plan was made, including all the project’s tasks. This plan was done recurring to a Gantt chart, Annex A. Table 1.1 records the start date and the ending day of the tasks and, for each task, the table shows if it was finished on time or not. Comparing the chart and the table allows checking the evaluation of the project’s state and detect possible delays. The first task of the project was a two weeks Induction Program. During these two weeks, the principal goal was to get to know the company, how it works, how do the different teams interact, 4Introduction who does what, essentially the induction aims to give a global knowledge to help kick off the actual project. The second task was researching about cross-docking, to gain knowledge on what is already done, what has an impact to implement this strategy, and who uses this logistic strategy. The third task was the refinement of the project’s objectives, to clarify what this thesis would focus. The fourth task consisted of studying the company and understanding what are the main obstacles when considering the implementation of cross-docking. The next two tasks are more complex and consist on developing a program capable of simulate cross-docking, recurring to the software called AnyLogic. Lastly, the earlier mentioned tasks happen in parallel with the writing of this thesis. Table 1.1: Tasks Diagram Task Planned Duration (days) Start Date End Date Completed on Time? Induction Program 10 Feb 5th Feb 16th Yes State of the Art 15 Feb 12th Feb 28th Yes Project’s Objectives Refinement 5 Feb 16th Feb 23rd Yes Understand Farfetch’s Position 10 Feb 26th Mar 9th Yes Implementation of the Chinese Case 30 Mar 5th Apr 13th Yes Implementation of the Transatlantic Bridge Case 35 Apr 16th May 11th Yes Writing the Dissertation 85 Mar 5th June 6th Yes 1.4 Outline Aside from the present chapter, Introduction, this thesis has four more chapters. Chapter 2, presents the literature review about cross-docking, reviewing works made about this topic, and some companies that apply this strategy in their operations. Chapter 3is about the current situation of Farfetch regarding cross-docking and identifies what a cross-docking implementation will affect. Chapter 4contains all the information about what was implemented and studied during the realization of the dissertation. The last chapter, Chapter 5, is the conclusion of this thesis, showing a perspective for future works and the main conclusion of the dissertation. Chapter 2 State of the Art 2.1 E-commerce E-commerce refers to buying and selling products on an online platform. This type of transition can be of three different categories: •Business to Business (B2B): this type of transition happens when the two parties involved are businesses. •Business to Consumer (B2C): when a business sells online to a third-party client. •Customer to Customer (C2C): this type of e-commerce happens in auctions websites, like eBay, where a private client sells to another third-party consumer. With e-commerce, anyone can buy something without leaving home and has a vast range of products at disposal. Another benefit is the 24-hour availability because with online shopping there is no closing time, which provides flexibility for clients that do not have the time or do not like to go to physical stores. The international reach given to brands or stores is another benefit that e-commerce gives, for the reason that even if a company does not have stores in a particular country, customers from that country can still buy products through the company’s website. On the other hand, doing shopping online makes the shopper buy before actually holding the products, which can be a huge downside, apart from the fact that they have to wait to receive the products. One specific sector of e-commerce is the online luxury market, that is slowing expanding to this type of sales method. Not only the big brands are starting to have their websites, but online platforms, like, Farfetch, Net-a-Porter, and Mytheresa are rising in this market, with their platforms, selling a very diversified range of products from brands, and, in Farfetch’s case, from boutiques dispersed all around the globe. 5 6State of the Art 2.2 Overview of a Cross-docking Approach Cross-docking is a logistic strategy used by many companies nowadays in different industries, based on the idea of having a warehouse or distribution center, called cross-dock, that receives packages from multiple inbound vehicles, sorts, and consolidates everything inside the cross-dock before loading the boxes into the outgoing trucks without the need of storing them. This approach creates a Just-in-Time (JiT) shipping process, that aims to increase efficiency and decrease waste by receiving materials only when they are needed, which eliminates the need of having stock. With cross-docking, instead of shipping small orders, that most of the times, do not occupy the entire cargo area of a trailer, the packages are aggregated into one large lot with the objective of fulfilling the cargo area of a truck (Vasiljevic, Stepanovic, and Manojlovic, 2013, [10]). 2.2.1 Warehousing versus Cross-docking Warehousing has four main steps, receiving, storage, order picking, and shipping (Bartholdi and Gue, 2004, [34]). If the storage of an article occurs, then an operator needs to unload the packages from the incoming vehicles, stored them, and, in the moment of loading, pick them up from the storage, and only after this, the loading of the items into the truck happens. These movements are demanding labor wise and generate handling costs. The storage of items also creates expenses, defined as holding inventory costs, because demands physical space to hold the items and, if an object spends too much time in the warehouse, it can lose value on the market. In a cross-docking approach, represented in Figure 2.1, the storage, and order-picking steps do not exist, because the storage of packages does not happen, in fact, a box should never spend more than 24 hours on the cross-dock. This type of approach can eliminate, or drastically reduce, the costs mentioned above, and with the aggregation of different orders and the creation of a Truck-Load (TL) shipping method, the reduction of costs can be even lower. The operations inside the cross-dock include the unloading of the packages, temporary storage, that may or may not occur, depending on the cross-docking type, discussed in Subsection 2.2.3, but if the storage happens, then this step is followed by picking, product’s preparation and loading the outbound vehicle. 2.2.2 Direct-shipping versus Cross-docking With direct-shipping, the packages ordered are delivered directly to the clients without using intermediate facilities. Usually, each vehicle has a route designated to it and performs one or multiple pickups and deliveries along the path. This strategy needs to program courses which ensures that the vehicle picks up the package before making the delivery, and assumes that all vehicles return to a central depot after making the deliveries. These problems are known as Pickup-and-Delivery Problems (PDP) (Savelsbergh and Sol, 1995, [36]). Tarantilis (2013, [4]) study cross-docking strategies problems know as Problems of Vehicle Routing with Cross-Docking (VRPCD). These two types of approaches always consider that vehicles leave from a central depot, which implies 2.2 Overview of a Cross-docking Approach 7 Figure 2.1: A Cross-Docking Approach in Khalili-Damghani et al., “A customized genetic algorithm for solving multi-period cross-dock truck scheduling problems”, [1], 2017 that the company has its fleet of trucks to make the deliveries. But having a fleet of vehicles can raise the expenses supported by the company, which can be a disadvantage, so the company may hire a Third Party Logistic (TPL) company to make the deliveries. In these cases, the vehicle routing problem is called Open Vehicle Routing Problem (OVRP), and in a cross-docking approach is designated as Open Vehicle Routing Problem with Cross-Docking (OVRPCD). In the problems mentioned, the vehicles do not leave a central depot instead they all converge to the same point, the cross-dock, on schedules times, but the pickup and deliveries happen at different times. With cross-docking instead of having the trucks delivering their packages to the same stores, the vehicles go the cross-dock, drop all the cargo there, and then the items undergo operations of sorting and aggregation. After these operations, the trucks leave to make the deliveries, but with the guarantee that they will not make the same routes, which lowers the expenses related with shipping when compared with the option of having trucks delivering only one type of product. Nikolopoulou et al. (2017, [26]) used a local-search meta-heuristic algorithm as an optimization framework to compare the PDP and VRPCD models. After various computational experiments, the conclusion reached was that, for a customer located in the same geographic area as the suppliers, the direct-shipping yield fewer costs than deliveries made with cross-docking. On the other hand, when the distance between the pair supplier-customer is abysmal or when the relations between suppliers and customers are many-to-many, cross-docking is capable of reducing costs and, if the location of the facilities is in central positions, the expenses can be even more reduced. Yu, Jewpanya, and Redi (2016, [28]) considered an OVRPCD and developed a Simulated Annealing (SA) algorithm to solve it. The algorithm uses a constructive heuristic and a local search to improve the quality of the solution found by the SA algorithm. They considered a problem where a retailer supplies a single product to multiple stores in a city. The algorithm determines the optimal number of vehicles needed and their routes. 14 State of the Art in 1988, discount department stores, with Sam Walton, Walmart’s founder, opening the first one in 1962, and grocery stores, called Neighborhood Market, opening for the first time in 1998, and this Neighborhood Market was designed to provide small communities a pharmacy, affordable groceries, and merchandise. Cross-docking is vital in the supply chain of Walmart because is used to replenish the inventory of the hypermarkets and allows the elimination of extra storage, transferring all the items directly from the inbound to the outbound trucks. The strategy has proved to improve efficiency, minimize transportation time and helped reduce prices. This approach was implemented in the early 1980s and became quite important because Walmart started to purchasing the items directly from the manufacturer and distributing the products with its fleet of vehicles. As of now, Walmart has more than 150 distribution centers across the United States of America. The network of cross-docks ships general merchandise, dry groceries, perishable groceries, along with other specialty categories daily. Six of the distribution centers are disaster distribution centers, strategically located across the country and stocked to provide rapid response to struggling communities in the event of natural disasters and with the launch of e-commerce, Walmart now also has fulfillment centers, strategically located in the United States of America. Each fulfillment center has unique characteristics based on its location. Suppliers and manufacturers within Walmart’s supply chain synchronize their demand projections, inventory forecasting, and stock replenishment. This integration between the entire supply chain helps everyone knowing the whole process and helps them behave, almost, like a single company. Over the past years, the company has become the world’s largest retailer and one of the most powerful and, in 1989, was named the Retailer of the Decade, with estimated distribution costs at 1.7% of its cost of sales, while its competitors reached values like 5%. And this was only possible because Walmart was one of the first companies to realize the advantages of cross-docking and was successful in its implementation. 2.4.2 Amazon Amazon is an American e-commerce platform funded in 1994 and, only in 2015, Amazon surpassed Walmart as the most valuable retailer in the United States. In amazon.com, a customer can buy books, music, movies, toys, housewares, electronics and many more types of products. The company also makes the famous e-book reader, Kindle, and since its launch, Amazon became a major force in the book market. As of now, when a client buys more than one item in the platform, an option to consolidate the packages is available, as long as all the orders are sold or fulfilled by amazon.com, and reduces the costs of shipping. Items that ship within 3 days or less of placing the order are sent together and, items that take longer than 3 days to ship are sent together based on their availability. A diverse type of facilities owned by Amazon uses cross-docking. To receive products from overseas to the U.S., Amazon has facilities called Inbound Cross Dock (IXD). As of now, Amazon 2.4 Case Studies of Successful Cross-docking Implementations 15 owns eight IXD in the United States of America, but the construction of one more, in the first quarter of 2019, is scheduled to happen. Outside America, the numbers are not completely available, but, in Germany, there is information about two facilities, and in the United Kingdom, one will be built. In this cross-docks, the products are received, broke into small bulks and temporarily stored until the fulfillment centers ask for the products. When this time comes, the merchandise is aggregated into truckloads and transported to the fulfillment centers. The IXD facilities are located near the major U.S. ports to minimize inbound ground transportations costs from the port to the facility. Apart from this facilities, Amazon also has 39 sorting centers, with the plan to build one more in the future, known as cross-docks, just in the United States of America. In these buildings, the packages are sorted by zip codes and, after the sorting, delivered to post offices responsible for each zip code. The boxes are then delivered by the United States Postal Service, which allows taking volume from UPS and FedEx. This system was introduced in the U.S. in 2014 and helped Amazon taking control over its outbound shipping costs. The number of facilities is not completely available, especially outside of U.S., but in France, Amazon has one sorting center, in Germany, in the future, the company wants to have one, and in India, there is information about Amazon having 25 sorting centers. 2.4.3 Target Target Corporation was funded in 1902 and is the second biggest discount store retail, losing only to Walmart. Just like Amazon, Target has multiple facilities with different purposes and applies cross-docking in two different type of facilities. When Target needs a smaller shipment, of less than truckload volumes, the shipment is processed in a domestic consolidation point, that act as a cross-dock. In the present moment, Target has a network of seven of these facilities, all in the United States of America, that are run by third-party logistics companies. In these facilities, the pallets with the products are unloaded and consolidated to enable the transfer of these products to Target’s regional distribution centers. The company has 26 regional general merchandise distribution centers in the United States. These facilities also use cross-docking as logistic strategy and operate in an exclusive-mode, having one side of the dock dedicated to inbound trucks, while the other side is exclusive to outbound vehicles. The operations involved in the cross-docking process are, mostly, done in an automated way, where the packages, pre-labeled boxes, are unloaded from the trucks directly to conveyor belts, and go directly to the outbound dock door, allowing the reduction of costs with material handling. 2.4.4 Sonae Sonae is a Portuguese retail company funded in 1959 by Afonso Pinto de Magalhães but only started to grow during the 1980s, after Belmiro de Azevedo assumed control of the company. 16 State of the Art Sonae’s brands are the favorites among the Portuguese consumers. In 2015, the company had a project to optimize the routes network and, one of the aspects covered was the implementation of cross-docking. According to the company website, the project was born with the need of simplifying the distribution design and reduce the logistics costs and, in order to achieve this, the principal focus of the project was to reduce the number of trips, and the distance traveled. The redesign of the distribution model was achieved through multi-load planning and crossdocking operations, revising the truck’s rental mode and the type of vehicles in the fleet. The company declared more than 1.2 millions euros globally gained during the year and the reduction of costs, in distribution centers, was more than 1.5 million euros. Apart from this, the number of trips in that year reduce 16.4%, and the cargo space of the vehicles occupied increased by 7%. 2.5 Conclusion E-commerce is a growing business, but to stay ahead of the competitors, Farfetch needs to identify opportunities to differentiate from its business rivals, and one of this opportunities is studying the implementation of a cross-docking strategy. Inside a cross-dock, only three operations need to happen, and they are the unloading and the loading of the trucks, and the operation of consolidation. Concerning the facility, the most important parameter is the number of inbound and outbound doors, related to the number of packages that the cross-dock will receive in a day, the location of the building, and the shape of it. Most of the articles available solve a simplified version of the truck scheduled problem in cross-docking. They consider a one-door problem and a one-to-one relation between client and supplier. In the real world, cross-docks have more than one door on the inbound and outgoing side of the facility, and a relation many-to-many is highly probable to happen, especially on an e-commerce platform like Farfetch, but heuristic or meta-heuristic algorithms, like tabu search, genetic algorithms or particle swarm optimization, are generally used to solve problems related with cross-docking. After reading the case studies mention in Section 2.4 and, taking in account, that the first three companies are among the most powerful companies in the U.S.A., this indicates that they are an example to follow when it comes to supply chain optimization, that, in these cases, includes cross-docking implementation. Sonae, the Portuguese group, has proven that the implementation of cross-docking, improves the overall efficiency of the supply chain, and can reduce the costs of products’ transportation. None of the companies are in the same area of business as Farfetch, but the one with most similarities is Amazon, and if cross-docking works so well in Amazon, the fourth most valuable public company in the world, Farfetch needs to look to this case of success and analyze the possibility of implementing cross-docking. Chapter 3 Farfetch Current Situation Cross-docking is a new strategy for Farfetch because currently, the e-commerce platform does not apply cross-docking in the shipping process, so is necessary to analyze its system structure to understand what this new strategy would impact. Figure 3.1 represents the system breakdown structure, and the next subsections of this thesis explore each component. Apart from identifying each part of the system, is also essential to understand the connection between them, and, in order to do this, was drawn an activity diagram, Figure 3.2, that allows getting the full vision of the order processing where a cross-docking strategy will have an impact. Figure 3.1: Farfetch Order Processing System Breakdown Structure After this analysis is right to conclude that a cross-docking implementation would affect the Checkout process, that sends messages to Routing, being that this last one consumes data from Atlas. After the client confirms the order, the Order Management Service (OMS) receives information about that order and starts a process with the final objective of shipping the article. 17 18 Farfetch Current Situation Figure 3.2: Farfetch Order Processing Interactions 3.1 Checkout Checkout is the section in farfetch.com where a client can confirm the purchase of the desired items and, this action, marks the beginning of the order processing by Farfetch. Currently, three major steps constitute Checkout: •Shipping; •Payment; •Review. The first step in Checkout is the shipping tab, where the user needs to choose the delivering place of the order and the destination can either be an address or a Click & Collect point. The succeeding step is the selection of payment method, that varies accordingly to the destination address. Depending on the method of payment chosen, the following steps can diversify. For instance, if a client selected payment with credit card, then information about the card will be asked, but if a user selects to pay with PayPal, Farfetch’s Checkout only needs to know what is the email of the client’s PayPal account. After this, the remaining step is the review, where the customer chooses the desired shipping method. The most common method of shipping is standard or express service, but two other services, Same Day delivery and F90, this last already overviewed in Section 1.1, may be available. 3.2 Routing 19 The Same Day delivery availability depends on destination city, and is available only in New York, Madrid, Milan, Paris, London, Barcelona, Los Angeles, Miami, and Rome, and with the boutique that sells the item ordered by the client. Only a couple of boutiques in each city are suitable to do services of Same Day deliveries, and, with this type of shipping service, the client receives the order in the same day of placing it. To be able to decide which shipping services are available, Checkout needs to communicate with the Routing service, because this service is responsible for deciding the shipping methods available at the moment of placing an order. The Routing service is discussed in detail in Section 3.2. The shipping costs are related to the shipping service selected, and different services can have different couriers assigned, being UPS and DHL the couriers most used. For each shipping service, the client has information about the shipping cost and the EDD. If the client is purchasing more than one item, and the articles come from different boutiques, then Checkout shows information about the client receiving two or more packages. When a customer confirms the order, all the items inside the client’s virtual basket will have a common identifier, designated portal id, associated with the client’s order. Each article in the basket will then have a specific identifier, called boutique order, that distinguish the different articles of the same order. 3.2 Routing As told in the previous section, the Routing service communicates with Checkout providing information about the available shipping services to a specific order, and is responsible for selecting the stock point to fulfill an order. A stock point is a local where a partner has stock of products. If the partner is a boutique, most of the times, the stock point is the boutique. This type of partners are small businesses and keep all the items in the store, but if the partner is a brand, most of the times, there are multiple stock points capable of fulfilling an order. In order to determine the best stock point capable of fulfilling a customer request, Routing takes into account information like the stock depth of each fitting stock point, the capacity of satisfying a whole order, and the commission earned by Farfetch with the sale of a certain item in each stock point. The stock point owner defines this commission with Farfetch and, aside from what was already mention, Routing also takes into consideration the preference set by partners from where to ship items and the destination of the order. Apart from communicating with Checkout, Routing consumes data from another service, called Atlas, discussed in the next section of this dissertation. 20 Farfetch Current Situation 3.3 Atlas Atlas is a centralized service with information regarding shipping’s routes, storing information about the expenses associated with each route, the courier assigned to it, and the exclusivity of the route. This service receives requests from the Routing service and filters the fitting routes. After finding all these routes, Atlas returns them to Routing, even if a route leads to a stock point that does not have stock to fulfill the order. Atlas consumes this information from a database of graphs, called Neo4J. In this database, a node represents a location identified by its zip code or the location’s country name, and the label SHIPS_TO is used to identify the routes between two nodes. Apart from these nodes, there are nodes labeled as All World Nodes that represent a place connected to the rest of the world, in order to lower the complexity of finding a fitting route. The availability of services like F90 and Same Day is done recurring to nodes identified by zip code, but this type of nodes prove to be a barrier in the countries that do not have zip codes. 3.4 Order Management Service OMS is responsible for processing an order after it enters Farfetch’s system and has six steps, represented in Figure 3.3, and is necessary to understand this process to identify in which steps a cross-docking strategy will have an impact. Figure 3.3: Farfetch Order Processing Steps If a store desires to sell an article on farfetch.com, the boutique has to agree with Farfetch requirements, because the e-commerce company decides the article’s display style on the platform. If the boutiques agrees, then it needs to send all the items to one of Farfetch’s Production offices. In these offices, a photographer from the eTailer company takes pictures of the items and, after this, the object is ready to be listed on farfetch.com. Lastly, Farfetch sends the article back to the partner. This whole operation happens to all the articles listed on the website. 3.4 Order Management Service 21 Step 1, Check Stock, is a responsibility of the boutique chosen by Checkout, and consists of checking if the partner has stock to fulfill the order. Cases of no stock can happen when Checkout chooses a boutique, but this boutique does not have stock, that may happen, due to an offline sale. An offline transaction occurs when a partner sells the article in the store and, after this transaction, it needs to update the stock in a Farfetch application, given to all the partners, used to synchronize the stock between them and farfetch.com. In the case of no stock, the customer receives information about the situation and, sometimes a similar article is suggested, and the customer accepts this suggestion, but, if no similar articles are fitting of being a suggestion or the client does not accept it, the customer receives a refund. Farfetch is responsible for the execution of Step 2, Approve Payment, that occurs at the same time as the previous step. In this step, Farfetch’s fraud team checks if the payment is fraudulent or not, which can be an automatic process or not. In order to accelerate the time span of this step, although its time span is much lower than the time span of Step 1, the e-commerce platform has a group of lists in the database, about clients that previously made purchases in the platform. The likelihood of being fraudulent is based on the placement of each client in a list. If the client was ever fraudulent than, for that moment forward, all the orders placed by this client are automatically declined, but in the cases that Farfetch does not have any information about the client, a manual process begins and consists in investigating the client to determinate the probability of being a fraudulent payment. The Step 3, Decide Packing, only beings after the approval of the payment and confirmation of stock to fulfill the order, given by the boutique. Each item has a suggested package, decided when the products are in the Production office, but the boutique can choose another package, not attending to the suggestion. The packages are given to the boutiques by Farfetch, and each partner needs to manage the stock of it in the store, and, when it is low, the boutique oughts to require more. After packaging the order, the e-commerce platform has the duty of creating the shipping label, Step 4, that, most of the times, is an automatic process. When the client did not provide the full information or correct one, the information needs to be edited manually and, only after that, the boutique can print the label, and the order is ready to be sent. The next step, Step 5, Send Parcel, begins automatically after Step 4 and, if the boutique has daily pickups scheduled, then the package is sent in the next pickup, but if not, the scheduling of pickup with the courier needs to occur. The last step, Step 6, In Transit, is a responsibility of the courier and precedes the reception of the package by the client. When the courier processes the package, the client receives an email with shipping’s confirmation and information about tracking. 22 Farfetch Current Situation 3.5 Shipping For shipping an order to the client, Farfetch only does direct routes, which means that the order goes straight from the boutique to the customer, without intermediate stops. A client who purchases multiple articles receives them one by one, with the exception of articles that belong to the same boutique. Indirect routes, made in cross-docking approaches, in the eTailer platform only happen in returns, and this approach is called multi-leg logic. A multi-leg logic happens when a package suffers a stop, or more, before arriving at its destination, and in Farfetch’s returns originated outside of Europe, that are going back to a European boutique, all the packages need to re-enter Europe through the United Kingdom. These orders suffer a stop in London’s warehouse, Norsk, and after begin received, they go back to the boutique that sold the article. In this approach, the shipping label received by the client in the article’s package, has the address of Norsk and corresponds to the first leg. The shipping label for the second leg is printed in the English warehouse and has the address of the boutique. The indirect approach allows lowering the taxes of re-entering the European space, that is proven to be inferior when this re-entering happens through the United Kingdom. Choosing the courier for the deliveries is also an important part of the shipping process. UPS and DHL are couriers generally used, but when the market demands it, the utilization of local couriers to deliver the packages can happen. For instance, in Brazil, the courier used is Correios, the Brazilian post, and for domestic orders in China, Farfetch uses SFExpress to delivery the packages to Chinese clients. Chapter 4 Proposed Solutions At the beginning of the project, it was clear that increasing the customer satisfaction had to be studied in different situations, and, in order to do this, were identified two application cases. In order to improve the client’s satisfaction, the objective is to aggregate the maximum amount of articles of the same order, reducing the number of packages that a customer receives. The first application case looks to the Chinese market and, in this case, the aim is to gain knowledge on how cross-docking works and what needs to change in Farfetch’s Product to support this type of strategy. The second case is a transatlantic bridge between Europe and USA and, this case, is an extension of the first case. With the simulation for this case, the aim is to analyze the whole shipping process, since the moment of leaving the boutique until reaching the client, and see how this strategy affects the lead time of the orders. 4.1 Chinese Case The solution for the Chinese market case is divided into two big groups, one being the To-Be analysis of Farfetch’s order processing system, and the other a simulation of the cross-dock, to gain knowledge on what happens inside the physical installation. 4.1.1 To-Be Analysis Checkout With cross-docking implemented in Farfetch, the client would have a new choice in Checkout, the option to consolidate the order. The presentation of this option would be a check-box and, if the client clicked on it, then a boolean variable, indicating if a client wants to aggregate the articles or not, would be set to true. If the customer selects the cross-docking option, services like Same Day delivery or F90 are disabled, and the client can only select standard or express delivery. Besides this, near the checkbox, information about the number of packages that the client will receive and the EDD for each 23 30 Proposed Solutions facility. The modification of the box appearance eases the work of the operators that know in which state the package is, without using the dedicated application, as mentioned previously in the loading operation. When the box arrives at the cross-dock, the box is closed and has printed in the top the portal id. After moving the package to the loading racks, the agent can assume two different presentations, if the box will suffer any operation of consolidation then the package appears has an open box that has the portal id printed in one of the sides of the package. After the consolidation, or if no aggregation happens, the box, again, appears as a closed package, but with a paper on top of it, representing the shipping label for the second leg, and the portal id. For the Boxes a state chart was elaborated, Figure 4.2, in order to choose different courses for each box in the system. Not all the boxes that go through the cross-dock are suitable of suffering aggregation, either because the order only has one item, or because the articles arrive with more than one day in between, and, for these cases, is important to let the package leave immediately after arriving at the cross-dock. The initial state of the chart is active when the package is waiting for being unloaded. After being placed in the unloading rack, a message is sent and fires the transition, putting the agent’s state chart in the next state, moveToDockDoor, that will remain active until the placement of the package in a loading rack, and, in this step, is chosen the path for the box. If the package will suffer an operation of aggregation, then the agent receives a message to wait for aggregation and goes to the state waitForAggregation, but if the box will not suffer consolidation, then the message received is that the box is ready to load, and the state active is the waitForTruck state. When the chart reaches this state, happens the marking of a package as ready to load, and, when a loading truck arrives, the last state is activated, and the loading starts. The Forklift represents the operator responsible for the operations that happen in the crossdock. Each forklift only has two variables, one used to mark it as occupied, inUse, and another to assign it to an unloading or loading truck. The Forklift agent is placed in the cross-dock in a specific place, in order to control where the forklifts are when they are not in use. The Load dock and the Unload dock agents represent the outbound dock and the inbound dock, respectively. Each one of these agents has parameters related to simulation’s configurations, like the storage rack, the place where the packages are stored, assigned to it, the node where the unload or load happens, and the number of the dock. Apart from this, the two mentioned agents also have the spot where the truck should dock and where it should turn to start approaching the dock, and, for the loading dock, an extra parameter is used to identify where the operation of consolidation takes place. Apart from these parameters, the two agents mentioned have a group of variables in common giving information about the occupancy of the dock, that refers to the existence of a truck suffering any operation in that dock, the forklift assigned to do the operations of loading or unloading, and the space available for storing more packages. Besides the two already mentioned, the agent of the load dock also has a vector of strings with the cities assigned to that dock. Like mentioned in Chapter 2is a good practice to have cities assigned to each dock, and that principle was applied in the simulation. 4.1 Chinese Case 31 Figure 4.2: State Chart for the Boxes agent The Loading Truck and the Unloading Truck agents represent the vehicles that come to the cross-dock. These agents have a variable indicating the number of packages already processed, meaning the number of packages already loaded or unloaded, respectively. Apart from this, the two agents have a variable to store information about the dock where the parking of the vehicles will happen, and a variable about the capacity of the truck, that needs a further explanation because of Farfetch’s business model. Farfetch does not have a fleet of vehicles, so the shipping process is all made by external companies, which means that it does not control how the courier will make the pickup of the merchandise in the boutiques. Due to this fact, the information about the capacity of the trucks was not available, and, to the simulation, the capacity of the vehicles was crucial. The way for working around this problem was by calculating the capacity of the vehicles, but in different ways to the agents mentioned. For the Loading Truck is assumed almost infinite capacity, meaning that each incoming truck carries everything marked as ready to load in the dock assigned to it. The capacity of the Unloading Truck is based on the number of packages received in each day and is calculated by dividing this value by 24, that represents the 24 hours available in a day to do the operation of unloading. If the amount of packages received in a day is lower than 24, then only one agent is created and is assumed that it brings all the articles. 32 Proposed Solutions Apart from these parameters, each of these agents has a function responsible for incrementing the number of processed packages, that returns true when this number equals the capacity of the agent, and this marks the ending of the process of unloading or loading. Lastly, in order to run the simulation, the program needs to have a top-level agent that can connect all the other agents, named Main. All the previous agents, excluding the Forklift, have a population located in the main agent, and the parameters of the other agents are defined in this agent. This agent also has parameters that, in this case, are user’s inputs, and are the total number of loading and unloading docks, and the number of loading docks that store only articles going for Hong Kong. Apart from these parameters, the user also needs to choose the number of operators of the cross-dock, the speed of the trucks that arrive at the cross-dock and the time needed for using the dedicated application, but Table 4.1 lists all the variables able for configuration. Table 4.1: Configurable parameters in the simulation for the Chinese Case Study Parameters configurable At the beginning of the simulation Through coding Number of loading docks Maximum and minimum value of loading docks allowed Number of unloading docks Maximum and minimum value of unloading docks allowed Number of loading docks to Hong Kong Maximum and minimum number of loading docks to Hong Kong allowed Number of operators Maximum and minimum number of operators allowed Truck speed Maximum and minimum value allowed for the truck’s speed Time needed for using the mobile application Maximum and minimum time to use the application allowed Loading Schedule Time between arrivals to enable consolidation Simulation’s Logic Four major groups compose the simulation’s logic. The first group is the logic for the unloading process, followed by the logic of moving the unloaded packages from the racks in the inbound dock to the storage in the load docks. After this, is evaluated if the article will suffer consolidation, if not the box moves to the loading logic block, but if it will suffer consolidation, then the agent, representing the item, goes to the logic block of the aggregation. 4.1 Chinese Case 33 Apart from these four groups, were define blocks of resources. The resources identified were the loading docks, the unloading docks, the operators responsible for the aggregation of the packages, the workers in charge of doing the loading of the trucks and then a group of operators responsible for all the others operations previously mentioned. Each resource has a capacity established, and these capacities are the input parameters of the main agent. The resource pool of the loading dock has a capacity equivalent to the parameter of the number of loading docks, and the resource pool for the unloading docks has a capacity equivalent to the parameter of the number of unloading docks. For this pool, each resource created is added to the correspondent population, load docks, and unloading docks, respectively, and they are configured as static resources, meaning that these resources do not move. The forklifts deserve special attention because the resource pool for the forklifts is divided based on their location in the cross-dock. For aggregation, exist two pools of resources, one in the right side of the cross-dock and one in the left side. This division also happens for the forklifts responsible for loading the trucks and the group that does the remaining operations. The number of forklifts allowed to do consolidation operations is stable and is always two, because in the loading docks, as previous mention, the cells can hold more than one item and work with cell depth was more challenging than what was expected. Due to this difficulty, only one aggregation can happen at each moment in each side of the cross-dock, because the aggregation’s logic is divided in two based on the number of loading docks, but this will be explained with detail later in this section. Is necessary to enhance that the dock has 20 entries, a value given by Farfetch based on a warehouse that is planned to be used by the company, but the simulation may not need to use them all, and because of this, is crucial to explain the actual capacity of each resource pool. The capacity of the forklifts, in the left side, responsible for the operation of loading is always 10, because the minimum value of loading docks, in this simulation, is 11, so the simulation always uses the left half of the cross-dock’s entries. The capacity of the forklifts on the right side is the result of subtracting 10 from the number of loading docks. Lastly, the pool of operators assigned to do the operation of unloading and moving the boxes has the remaining forklifts. The capacity of this pool is the result of the difference between the number of operators, a user’s input, and the capacity of the resources pools explained above. This capacity is split by the forklifts in the left side and on the right side, and Equation 4.1 shows the formula used to obtain the capacity of the pool in the left side. The objective of the equation is to set the number of forklifts available for each unloading dock, and then multiply this value by the number of docks served by the pool of forklifts in the left side. The number of docks served by this pool is always 10, because, in order to run, the simulation needs to have, at least, 12 unloading docks. The capacity of the pool of forklifts in the right side follows the same logic and Equation 4.2 shows the formula used. Number o f Forkli fts −Number o f Loading Docks −2 Number o f Unloading Docks ×10 (4.1) 34 Proposed Solutions Number o f Forkli fts −Number o f Loading Docks −2 Number o f Unloading Docks ×(Number o f Loading Docks −10 ) (4.2) These resources have a configuration of moving resources and is necessary to define their speed, set to 5 km/h. The value for the speed was chosen based on the recommended safety speed for forklifts that move inside warehouses. Starting with the logic of the unloading process, the arrival of the first truck triggers the activation of a one-time occurrence function responsible for the attribution of destinations to the load docks and of a function responsible for doing initializations of vectors. The amount of docks to Hong Kong is an input of the user, with a minimum set to five docks, because, after analyzing the data, was possible to conclude that 53.57% of the orders made during November of 2017 have as destination Hong Kong. This volume of orders to Hong Kong happens because of the difference in the laws of importing between China and this city, and due to this, people often tend to ship articles that come from outside of China to Hong Kong and, after arriving there, they ship the items to China. At midnight of each day, a function is responsible for checking, from the database with data about the orders, which orders have the EDD equal to the present day and for storing in an array the portal id, boutique id, country and city of destination for the orders that will come to the crossdock in that day. Apart from this, the function also sets the total number of boxes that will come to the cross-dock. The arrival of the first vehicle in each day sets the calculating of the number of trucks that will arrive in that day and is calculated based on the total number of packages, arriving in this day, divided by the capacity of the unloading truck. If the value obtained multiplied by the agent capacity does not equal the number of packages, then it means that the calculation produced a non-integer value, and the truck that arrives first, will have its capacity adjusted to compensate the difference between the total packages received and the multiplication mention. After arriving in the dock, the truck needs to seize a dock where the unloading can happen. Each time that a vehicle seizes a dock from the unloading docks’ pool of resources, this dock is marked as occupied, until the vehicle that seized it leaves the cross-dock. If a truck comes into the cross-dock, and no dock is available, it waits until a dock is available. Following this task, the vehicle parks in front of the entrance to start the unloading process, and, when the vehicle stops, an operator is chosen to do this operation, based on the number of the dock. If the number of the dock divided by two is lower or equal than half of the maximum amount of unloading docks in the cross-dock, always considered as twenty because this is the maximum amount of entries that can be used, then the forklift chosen is from the resource pool on the left, but if not, then the forklift chosen comes from the pool in the right. As soon as the worker starts moving, the operator is marked as busy. Following the unloading of one package, happens the creation of an agent in 4.1 Chinese Case 35 the population of boxes. The moment of creation sets the portal id, boutique id, country and city of destination, of the agent and, after setting this information, a counter used to know which row of the array, containing the information mentioned, needs to be checked, increments. The truck leaves the dock when all the packages that it brought are unloaded and, when it leaves, the truck’s seized dock is released and marked as free, and the truck is deleted from the system. The unloading of a box triggers a function responsible for associating the number of boxes to be aggregated with the one that just arrived. This function checks all the packages with the same portal id coming in the current and following day. The base for the time gap between arrivals is the 24-hour rule, but is important to understand that the EDD provided by Farfetch defines the moment of arrival. This value does not give an estimation in terms of hours, which causes, for some packages, spending more than 24 hours inside the cross-dock, depending on the hour of arrival of the trucks, that is random. For instance, if a box arrives in the 5th of November, then the packages to be aggregated with this one, need to come during that day and during the 6th. If the first box arrives at 8:00 and the last one, already arriving in the following day, arrives at 14:00, then it is cleared that the total time span inside of the cross-dock is larger than 24 hours. So, for this case, the limit time span inside of the building is set to 48 hours. Apart from this, is important to understand that because of considering the EDD as the date of arrival in the cross-dock and only considering orders placed in November, the first and the last days of the month have lower volume than normal. No orders can arrive on the 1st of November, and in the second day of this month only five come to the dock, because, during this first days, the clients are making the orders. The simulation finishes in the 27th of December, but only two boxes arrived in this day, as result of the last order considerate being placed in the 30th of November and the larger part of the packages arrived before the 7th of December. Aside from what was mentioned, the function triggered by the unloading of a package, also verifies which load dock is assigned to the box’s destination and reserves space for the package in the load rack, in accordance to what was already mentioned in the case of already having a box with the same portal id inside the cross-dock. If the article is the first one of the order arriving or if no consolidation will happen, the space reserved is the nearest to the dock door, on the lowest level of the rack, and on the right side of the rack. This preference is set to speed up the process of loading because the closest the boxes are to the entrance of the dock, fewer steps the operator need to give, and placing the boxes in the lowest level has the same purpose, considering that with this, the worker does not need to lift the forklift, which reduces the time to take the packages from the rack. Following the unloading of a box, starts the second major step, the moving of the package, occurring in parallel with the unloading of a vehicle. For each article, one free agent from the forklifts pool is selected, based on the same principle of choosing the forklift responsible for the unloading of the trucks. An explanation about the operation of moving the boxes is given in one of the previous subsections, and the simulation replicates the process described. To simulate the use of the mobile application, the agent stops moving and stays in front of the racks. 36 Proposed Solutions If the box is going to suffer consolidation, in the simulation, the agent waits in a block, that resembles a queue, but with a functionality that allows releasing agents in a custom way. Is important to explain that when the agent enters the logic of consolidation, the block where the agent waits depends on the number of the agent’s load dock. If the number of the load dock is lower or equal to ten, half of the maximum number of loading docks, then the agent will wait in a block that leads to the box being consolidated by the forklift in the left side responsible for this operation. If the number of the load dock is greater than ten, the agent will wait in a block that leads to the operator on the right side. Each time a box enters this block, a function checks if the packages that already are inside the block have the same portal id and batch size, and for each found, it increments a counter. After checking all the agents in the block, if the counter matches the desired size of boxes to aggregate, the packages are released, and the operation of consolidation starts. The simulation follows what was explained previously about the operation of aggregation. In the simulation, the variable itemsInside of the mother-box, after the consolidation, equals the batch size, and, for the remaining packages, the variable is zero. At the end of each consolidation, a function checks which boxes have the number of items inside equal to zero and deletes them from the population of boxes. The only remaining group is the loading process and is crucial to understand that the arrival of load trucks to the cross-dock follows a schedule. This schedule defines at which hour the loading trucks come to the facility and the number of pickups that happen in a day. When the moment of a pickup arrives, a function checks which dock has packages marked as ready to load. If it has, then the dock is marked as can be seized and a counter is incremented, to know how many trucks will come into the cross-dock. If the dock does not have packages ready to load, then no truck comes to that dock because the dock is not marked as can be seized, and the counter does not increment in these cases. The number of pickups is programmable before running the simulation, as well as the moment of occurrence. The process of entering the dock is similar to the one explained for the unloading process, with the difference that only loading docks, marked as can be seized, are available for seizing. When the truck parks in front of the dock, a function responsible for counting the number of boxes marked as ready to load in the dock seized by the vehicle, that now has the status of occupied, defines the number of packages that the truck will have to carry away. The parking of the truck triggers the beginning of the loading process that follows the process described in the Subsection 4.1.2. The loading truck leaves when all packages are loaded and, when it exists the cross-dock, the assigned loading dock is also released and marked as free. Lastly, the vehicles and all the articles that it carries are deleted from the system. Results The simulation has minimum requirements to run, and the results mention below are obtained with the parameters set to these minimums. There are 11 loading docks, 5 of them dedicated to Hong 4.1 Chinese Case 37 Kong, 12 unloading docks, the time dedicated to using the warehouse application is 35 seconds, the speed of the trucks is set to 25km/h and exist 107 operators inside the cross-dock. The number of operators was set to avoid having wait time for lack of operators, despite the simulation running with less but at the cost of increasing the average time inside the facility. For this case the schedule has three pickups, the first one happens at midnight, the second occurs at 8:00, and the last moment of loading is at 16:00. To evaluate the impact of cross-docking in Farftech, aside from the programming of the operations inside the cross-dock, were elaborated some statistics, that can be seen in Annex C. First, there are two graphs made at the start of the simulation that remain static during it. The graphs are both pie charts, and one of them shows the number of orders with more than one item and the number of orders with only one item. This graph shows that 35.9% of the orders made by Chinese customers, during the month of November, can be suitable for aggregation, a value superior to what the company was expecting. This graph does not take into account the 24-hour rule, and the percentage of 35.9% represents a total of 12553 orders, from the total of 34979 made during that month, and represents a total of 36819 articles. The other pie chart takes into consideration the 12553 orders that can suffer consolidation and divides all the orders based on the number of articles that each order has. There are nine partitions that go from orders with two articles to orders with ten or more. This graph shows that 56.9% of these orders have two items, which represents a total of 7144 orders, but the number of orders with ten or more items, representing 1.2% of the total orders that can suffer aggregation, which is equivalent to 154 parcels, surpasses the number of orders with eight and nine articles. The pie charts mentioned are in the view area of the statistics, alongside two others that have the input data altered during the simulation. One of this non-static graphics is a pie chart, just like the static pie charts, but the other is a stack chart. The pie chart shows the number of packages aggregated, to understand the impact of the rule of the 24-hour, because this rule causes the nonaggregation of packages that, without it, can suffer consolidation. At the end of the simulation, the graph showed that from the 36819 items that could suffer aggregation, only 22.4% did not go through this operation, which means that 28555 articles suffer an operation of consolidation. The stack chart has the same propose as the pie chart, but shows the number of items sent in each package that leaves the cross-dock. This graph has the nine partitions previous mentioned and allows comparing the two graphs with this division. Taking, for example, the case of orders with two items, at the end of the simulation, 7032 packages exited the cross-dock with that amount of items inside, which represents a difference of 112 boxes between what could be and what was actually aggregated, after considering the rule of the 24-hour. For the case of boxes with ten or more articles, only 56 packages left with this amount of items inside, less 98 boxes when compared with the static chart. Aside from these four graphics, three more can be seen in the 3D view area, that change during the simulation and, one of them, is a time plot that shows the average time to unload and to load a truck. In this graphic the unit of measure is minutes and, by the end of the simulation, the average 38 Proposed Solutions time to unload a truck is 75 minutes, while the time to load a truck is around 20 minutes. The second one is also a time plot, that uses hours as the unit of measurement, and shows the average time that a box spends inside of the dock and, for the case of the three daily pickups, the average time inside the dock is 9.2 hours. The average time inside the dock is mainly related to the gap between each pickup. For instance, if two articles from the same order arrive at the cross-dock with a space of one hour, where the first arrives at 8:30, and the other arrives at 9:30. If the boxes were aggregated and shipped immediately, then the time inside the dock would be around one hour, because the operation of aggregation does not take long. In the case of three pickups, the next load truck only arrives at 16:00, which causes the box to spend 6.3 hours inside the cross-dock. In order to confirm the relation between the time spent in the dock and the number of pickups, this last variable was incremented to four. The time gap between the pickups, for this case, was reduced to six hours, instead of eight, meaning that the first pickup happens at midnight, the second at 6:00, the third at 12:00, and the last occurs at 18:00. After running the simulation, with all the parameters equal to the initial simulation, except for the loading schedule, the time plot showed an average time inside the dock of 7.5 hours. The last case considered has four pickups but with different time gaps between them, where the first one still happens at midnight, but the second occurs at 4:30, the third pickup is at 8:00, and the last batch of loading trucks arrive at 16:00. With this loading schedule, the time plot shows a average time inside the cross-dock of 6.5 hours. This three different configurations for the loading schedule show that, not only, the amount of pickups influence the lead time inside the dock, but also, the moment of the pickup is also a variable with influence. The last graph in the 3D view area is a bar chart that shows the number of packages aggregated per day and the information in this graph updates at the beginning of each day. Each day where, at least, one operation of aggregation took place has a bar in the graph and, each bar, has a label with the date of the day, with this format YYYY-MM-DD, and the number of orders aggregated. The day with most aggregations is the 30th of November, with 863 orders suffering consolidation. Aside from all the graphs mentioned above, at the end of the simulation happens the creation of an excel file with two sheets. The first sheet has information about every package that exited the cross-dock, showing the portal order id associated with the box and the time of arrival of the mother-box at the cross-dock. Apart from this, exists a column with the moment of loading the box into a truck, the number of packages aggregated with the mother-box, and the total time spent inside the dock. This sheet has 41455 records, a number that represents the total amount of packages that left the cross-dock, taking into consideration all the boxes, and not only the ones that suffer aggregation. This sheet allows a deeper knowledge of the time spent inside the cross-dock and the number of items in the boxes when departing from the dock. After observing the data in the file, is possible to see that the package that holds more items left the dock with 20 articles inside, and the biggest time span in the cross-dock was 46.491 hours, fulfilling the limit of 48 hours inside the facility. Lastly, some calculations were made to realize the number of orders that 4.1 Chinese Case 39 spent more than 24 hours inside the dock, the preliminary limit of time span inside the dock, and to determine the number of packages that spend 9.2 hours, the average time inside the dock, or less in the facility. Only 3884 packages, representing 9.37% of the boxes that left the cross-dock, surpassed the time mark of 24 hours, and the data reveals that 36111 packages, 87.11% of the total packages, spend less or equal time inside the dock that the average time. After doing the others simulations, more calculations were added to this sheet to compare the results obtained in the different situations. Each case has a report, and the results of each case are in Table 4.2. Table 4.2: Results of the three simulations for the Chinese Case Case with 3 pickups Case with 4 pickups and 6-hour gap Case with 4 pickups and different time gap between them Number of packages and percentage that spent more than 24 hours in the facility 3884/9.40% 4020/9.70% 4120/9.94% Number of packages and percentage that spent 9.2 hours or less in the facility 36111/87.11% 35995/86.82% 35579/85.83% Number of packages and percentage that spent 7.5 hours or less in the facility 25435/61.36% 35303/85.16% 34510/83.25% Number of packages and percentage that spent 6.5 hours or less in the facility 13853/33.42% 33928/81.81% 33712/81.32% Biggest time span inside the dock (hours) 46.491 41.783 46.111 After analyzing the data, is possible to see that the number of packages that spend more than 24 hours increased slightly when the average time reduced. In the case of an average time of 7.5, the case represented in the second column, the longest time span inside the dock, was decreased in, almost, 5 hours. The case with four pickups and a time gap of 6 hours between pickups showed overall better results, except for the number of packages that spent more than 24 hours in the cross-dock, and the case of four pickups with different time gap performed worse than this one in all the results. In the second case, 81.81% of the packages spent 6.5 hours or less inside the facility, while with the three pickups only 33.42% of the packages have a time span lower or equal to this value. 46 Proposed Solutions then the first six docks have one more destination assigned. The first step begins at midnight of each day. At midnight, a function checks all the packages with EDD equal to the present day and, after this, starts checking the boutique that has a stock point id identical to the one associated with the order. Due to the number of partners that Farfetch has, not all the partners were represented in the map, to avoid crumbling it, and, when the function does not find a stock point id equal to the one associated with the article, then the item is assigned to leave from a random boutique. To assure the arrangement of the packages, that do not have a stock point id correspondence, through all the boutiques, the first package will come from the first boutique in the population of boutiques, the second box comes from the second agent in the population and so on. If this attribution reaches the last boutique in the population, then the counter resets, and the next item originates from the first boutique. After all the articles having a boutique associated with them, is time to set the departure of vehicles. To verify if a boutique has packages to send, a function checks if the variable named ordersToSend is bigger than zero, if it is happens the creation of a vehicle, and its capacity is the same as the number of packages that the boutique needs to send. If the boutique is in Italy and the city is not Palermo or Catania, the two biggest cities of Sicily, an Italian Island, the presentation of the vehicle is the blue truck, and the variable isATruck changes to true, but if the boutique is located outside Italy or in one of the two Sicilian cities, the presentation is the plane. In order to know what packages the vehicle is carrying, the information about the stock point of the boutiques and, if the boutique has articles placed randomly there, the stock point id of these articles, are stored in an array named idStockPoint. Following the creation of the vehicles, they are bound to the Italian cross-dock, and the first step ends when they arrive there. When they reach the facility, the state chart of the vehicle is in state arrivedAtDestination and is important to understand what happens in this state. If the vehicle that arrived at its destination has the boolean variables toAmerica and americanTruck set to false, the default option for these variables, then the function, defined inside the state, creates a truck in the population of Trucks placed inside the Italian Warehouse agent. After the creation of the agent, a function checks which items have the EDD equivalent to the present day, and stock point id equal to the ones contained in the array of stock point id of the vehicle, and the capacity of the truck created is identical to the capacity of the vehicle. If the vehicle has the variable isATruck set to true, then the presentation for the truck is the blue one, but if the variable is false, then the truck inside the cross-dock presents itself as a silver truck. After the creation of the truck, the vehicle, associated with it, is deleted, and the process described in the Chinese case begins, i.e., the truck seizes a unload dock, the process of unloading begins and so on. The operations inside the cross-dock will not be explored in this section, taking into account that they were already explained in detail previously. The only difference that needs to be enhanced is the fact that, in the Italian facility, does not exist any operation of consolidation, because this operation happens in America. In this facility, the boxes do not have the batch size set, and the docks do not have a location assigned, due to the fact that all the packages go to the 4.2 Transatlantic Bridge 47 same place. The attribution of loading docks is sequentially, i.e., the first package will go to the first dock, the second box goes to the second dock and so on. The moment of loading trucks follows a schedule and, in this case, the vehicles arrive to load at 2:00, 12:00, and 20:00. After loading the trucks, an event that marks the beginning of the third step, they take the articles to a plane, a process not visible in the simulation, that takes all the items together to the USA. After being loaded into the trucks, the variable sentToUnitedStates is set to true to each box that leaves the cross-dock. This boolean variable indicates which boxes already went to the cross-dock in New York. Aside from this, happens the creation of a vehicle marked as toAmerica to later identify which agent from the population of vehicles is the one that carries the packages that left the Italian facility, and the capacity of this agent is equal to the number of packages marked as sentToUnitedStates and not marked as onWayToUnitesStates. This last variable is set to true when the plane starts moving and indicates the packages that no longer exist in the Italian building. This variable is important because the boxes are not immediately deleted from the population of boxes in the cross-dock after leaving, and this variable allows not setting the capacity of the next vehicle bound to the USA wrongly. The fourth step begins when the vehicle arrives at the American facility, and the variable toAmerica is used to check what needs to occur upon its arrival. First, is important to set the capacity of the trucks, because when the plane brings a great number of packages, more than one truck needs to carry them to the cross-dock, because one truck only needs one dock door, and, using only one dock door is inefficient. If a large group of trucks comes to the facility, the total time of the unloading operation decreases, when compared with unloading only one truck that brings a high volume of packages. So, the number of packages that a truck can carry was set to 200, because the unloading racks in the cross-dock located in the US hold 448 items, and, the amount of packages set for the truck’s capacity, allows unloading two trucks without having to wait. While the unloading of the second vehicle happens, the operators are already moving the articles to the loading racks, which avoids having a great set of trucks waiting for unloading, due to lack of space in the unloading racks. For each agent created, the number of packages carried by the vehicle decreases by the capacity of the truck, until being less than 200. When this occurs, is conceived the last agent, bringing the remaining packages. To check which boxes are inside the plane, a function checks which articles inside the Italian Warehouse have the sentToUnitedStates and onWayToUnitesStates set to true, and copies all these packages, until meeting the capacity of the truck, to the population of boxes inside the American cross-dock and deletes them from the population of boxes of the Italian facility. Another important function that happens to all the boxes unloaded in this cross-dock is the function responsible for setting the batch size. This function works in the same base as the one described in the previous section, but with the difference that a package can arrive at this facility in the day after being shipped from the boutique, because of the time spent traveling between Europe and the United States, and is important to know if the box is arriving in the day of shipping or in the next day. To determinate the day of arrival of each package, a function compares the date of 48 Proposed Solutions arrival at the Italian cross-dock and the date on the moment of arrival in the other facility. If the days are different and no boxes with the same portal id are in the facility, then the batch size is the sum of the number of packages shipped in the same day as the box that just arrived and the boxes shipped in the day of arrival at the facility in New York. If the days are equal and, again, no boxes with the same portal id are already in the American facility, then the batch size is the amount of boxes shipped in the same day as the mother-box and the ones with EDD set to the following day. If the box is not the mother-box, then the batch size will be equal to the one set in the mother-box. After this process, begins the usual operations inside the cross-dock, again this was explained in Section 4.1. The loading schedule defined in this cross-dock has four pickups, the first at 2:00, the second occurs at 5:00, the third happens at 15:00, and the last batch of loading trucks comes at 20:00. These times were chosen based on the pattern of arrivals of the planes to this cross-dock. Is important to enhance that at each loading moment, more than one truck can come to the same dock door. In this simulation, each state will have a vehicle delivering the packages to it, if, at least, one package marked as readyToLoad inside the cross-dock has that state as its destination. If a loading dock has four states assigned and has packages to deliver in the four, then four trucks come in one moment of loading, and each vehicle takes the packages bound to one of the four states. The last major step begins when the loading trucks leave the American cross-dock. For each truck exiting the facility, is created an agent in the population of vehicles in the Main agent. The newly created agent takes every box loaded into the truck and is bound to the destination of the boxes that it carries, and the variable americanTruck switches to true. If the destination is New York, then the presentation of the agent is the red truck, if not, the agent is the plane. When the vehicle arrives at its destination, the variable americanTruck defines what happens upon its arrival. When the vehicle arrives, the total time in transit of each article inside of the boxes that it carries, if it brings more than one, is recorded and the package has the variable wasDelivered set to true. In this simulation, the function responsible for deleting the agents from the population of boxes in the American cross-dock, happens after each consolidation checking which packages have the number of items inside equal to zero, just like in the Chinese case, but also verifies the agents that have the variable wasDelivered set to true. Another moment of triggering this function is after unloading a package and loading a truck, checking more times if an operation of deletion has to occur. The moment of deleting the agents is crucial since, in this case, the creation of the boxes’ agents happens when the truck enters the logic of the unloading process and not after being unloaded, like in the previous case. The simulation needs to know which was the unloaded package and, to recognize, the boxes the program recurs to the index of the agent in the population. If the index of the agents changes before unloading all the packages, it can cause errors in the program, and this is why is so important to control the moments of deleting agents from the boxes’ population because deleting an agent can change the index of all the other agents. Apart from checking the two variables mentioned above, in the moment of deleting, the function also checks if the variable packagesInbound, a variable equal to the number of packages currently 4.2 Transatlantic Bridge 49 being unloaded, is zero, if all the conditions are satisfied, the elimination of the agents can happen. Results The simulation for this case uses 12 load docks and 12 unload docks in the Italian facility, and 15 loading entries and 14 unloading gates in the American cross-dock. Each cross-dock has 150 operators, and the truck speed and the time to use the mobile application is equal to the value used in the Chinese simulation. Lastly, the plane has a speed of 750 km/h and the truck in the world map moves at 80 km/h. All the statistics used in the Chinese case are used in this simulation too, and all the graphs obtained are in Annex F. In respect to the time plots in the 3D view, each cross-dock has its own. The Italian facility only has two graphics, the two time plots, one showing the average time, in minutes, to unload and to load a truck, and the second shows the average time spent by the packages inside the facility. For this cross-dock, the average time to unload a truck is 44 minutes, while the average time to load a vehicle is 12 minutes, and the average time spent by the packages here is 6 hours. Unlike what happens in the Chinese Case, the average time spent in the crossdock is highly variable during the simulation and, during the spikes, this variable can reach 14 hours. The amount of packages, in the normal days, is between 2000 and 3000, approximately, and, the biggest spike on the number of orders, happens in the 29th of November, represents 9290 packages passing through the facility on only one day. The Black Friday happened in the 23rd of November and this holiday causes this spike, because, usually the average time to delivery to the United States is around six days, so the great number of orders placed during the American holiday will have the EDD set to 29th. The time spent in the cross-docking during the spikes is expected to be higher since the number of orders triplicates, and, taking into account, that the average time is around the double of the normal average time, the value of 14 hours is satisfactory. The American cross-dock, on the other side, has the three graphs talked in the previous section. In this facility, the operations of unloading a truck needs, on average, 180 minutes, while loading a vehicle takes around 18 minutes. The average time to unload is higher than the other cases since this cross-dock receives a great number of packages at the same time. Because of this, the items need to wait an extended time before the unloading actually starts, and due to always carrying an immense amount of packages, usually 200, this also raises the overall time to unload a truck. The day with most aggregations is the 28th of November, with 967 orders suffering consolidation, and the boxes spent, in average, 5 hours inside the dock. This lead time suffers from the same mentioned for the Italian cross-dock, reaching a pike of 35 hours and, again, the orders arriving in the 29th of November are responsible for this spike. The time plot of the average time shows the spike between the 1st and 2nd of December since some packages, placed on the 29th, arrive in the following day, so a lot of boxes spend, just because of this, more than 24 hours in the facility. Adding the time to suffer the operations inside the cross-dock and the wait time for loading, in the worst case, they can wait 10 hours for the loading trucks, which easily increases the time inside the cross-dock. To conclude the analysis of the average lead time, the packages, from orders that 50 Proposed Solutions have the EDD set to the 29th of November and suffer aggregation, are, highly probable, to only leave during the 30th. The graphs visible in the statistics view in the Chinese Case, are also visible in this simulation, and show, for this case, that 24.0% of the orders made during November of 2017, have more than one item, and the articles of these orders represent 46.35% of all the items bought during this month with the USA as destination. The total amount of packages suitable for aggregation is 32939, before considering the 24-hour rule, and, at the end of the simulation, the pie chart that contains information about the number of packages aggregated, shows that 23.5%, 7731 items, did not suffer aggregation because the time gap between the arrival of the boxes of the same order is bigger than one day. The other static pie chart, the one with nine partitions that gives information about the number of items in each order, shows that 63.7% of the orders, with more than one item, are composed by two articles, representing 7677 of the orders. The number of orders with ten or more items, 0.7%, that represents 81 orders, surpasses the number of orders with eight and nine articles, just like in the Chinese Case. Lastly, the stack chart, at the end of the simulation shows, that 6927 packages leave the cross-dock with two articles inside, a reduction of 750 boxes when compared to the value obtained before considering the 24-hour rule, and 33 boxes leave the cross-dock with ten or more items, less 48 than the value initially obtained in the static pie chart. Apart from these results, the excel file created at the end of the simulation now has five sheets, Annex G. The first two sheets are the sheets referent to the Italian cross-dock. The first sheet is the operations sheet and has the information mentioned for this sheet in the previous case, with the exception of the column with the number of aggregations and the percentage of aggregations, because, in this cross-dock, the operation of consolidation does not exist. The day where most packages came to the facility was the 29th of November, the spike previously mentioned, and 9290 boxes pass in this cross-dock. The second sheet for the cross-dock in Rome is the report sheet, showing the portal id, arrival and departure time from the cross-dock of each box, and the total time span of each box. In this sheet was calculated the amount of packages that spend less than 24 hours in the dock and was verified that all the packages meet this criteria, and the largest time span in the facility is 18.998 hours, so the results obtained here are satisfactory and respected the 24-hour rule, that is the aim for cross-docking. This sheet has some more calculations and Table 4.4 shows them. Table 4.4: Results obtained for the Italian cross-dock Number of packages and percentage that spent less than 24 hours in the facility 71064/100% Number of packages and percentage that spent 6 hours or less in the facility 19170/26.98% Number of packages and percentage that spent 14 hours or less in the facility 69192/97.37% Biggest time span inside the dock (hours) 18.998 4.2 Transatlantic Bridge 51 The second and third row of the table presents the number of boxes that have a time span inferior to the average time obtained and to the value of spike. The amount of boxes that spend less than 6 hours, the average time, inside the facility is relatively low and shows that the spikes, in this particular case, are extremely important and need to be taken into account. Looking now at the value of spike, 14 hours, is possible to see that the lead time of, almost, every box is below the value of spike, and this is a great result because, not only the 24-hour rule is fulfilled, but also 97.37% do not need this amount of time to leave the facility. The two sheets mentioned are also created for the cross-dock in New York. The sheet related to the operations shows that in the 28th of November represents the day where more boxes, 10818 packages, came to the facility. The day where most packages suffer aggregation is the 23rd of November, and 2401 boxes suffered consolidation. The column with the percentage of orders aggregated, in this case, shows that a lot of packages that come in day suffer aggregation only in the next one. The case with most highlight happens in the 3rd of December, where only 22 boxes arrive at the cross-dock, but 452 suffer aggregation, giving a percentage, of aggregated packages, of 2054%. Another case is the 2nd of December, where 263 packages arrive at the American facility, but 1778 suffer consolidation, again the percentage surpasses 100% and reaches 676%. These are the only cases where the percentage goes above the 100%. The report sheet for the facility in the USA shows that the biggest time span inside the facility in American soil is 61.435 hours and has 55750 records. This sheet also shows that only 1.35% of the packages spent more than 24 hours inside the dock, which is a phenomenal result and is really close to actually meeting the 24-hour rule. Table 4.5 lists the calculations added in the sheet. Table 4.5: Results obtained for the American cross-dock Number of packages and percentage that spent more than 24 hours in the facility 755/1.35% Number of packages and percentage that spent 5 hours or less in the facility 40422/72.51% Number of packages and percentage that spent 20 hours or less in the facility 54773/98.25% Number of packages and percentage that spent 35 hours or less in the facility 55667/99.85% Biggest time span inside the dock (hours) 61.435 The table with the results of the American cross-dock has the same purpose as the one created for the Italian facility, with the only difference of the average times considered. The lead time in the second row is 5 hours and 72.51% actually spent less than this time in the cross-dock, which is a really good value, because adding only 5 hours to the time in transit of the boxes due to going through the American cross-dock is, almost, meaningless. In the calculations were considered two spikes, one of 20 hours and, the biggest spike, of 35 hours and, practically, all the packages have a 52 Proposed Solutions time span below 20 hours, showing that the spike only affects a small percentage of orders, which is amazing because shows that the cross-dock has the necessary capacity for dealing with spikes of orders, without affecting the usual time inside the facility. Lastly, another excel file shows the lead time of all the articles ordered during November, taking into account the origin and the destination of the box. This file is important to evaluate the impact of cross-docking in the process of shipping, considering that this solution looks at the whole process of delivering the orders to the clients. With the data mentioned, was created a pivot table. This pivot table has the origin and destination of the boxes and shows the average time in transit for each combination. Farfetch provided the usual times in transit for each American state and for each route between Europe and the USA. The pivot table has 25 origins, and 53 destinations, each origin is a European country and the destination is an American state, and always shows the average time in transit for each combination selected. Table 4.6 shows the main cases analyzed. Three of them are related to three American states that represent the main percentage of orders made by the clients. The routes in the table are the three top routes provided by the company. Table 4.6: Average times in transit Average time in transit (days) (1) (2) Given by Farfetch (2) Obtained in the simulation (1) To California 3.4 1.32 0.39 To New York 2.3 0.96 0.42 To Florida 3.3 1.19 0.36 Between Italy-US 3.5 1.17 0.33 Between UK-US 2.8 1.49 0.53 Between Poland-US 3.5 1.09 0.31 After analyzing the six cases listed in the table, is possible to see that the value given by the simulation is always lower than the one given by the company, and it is about 39% of the time given by Farfetch. This value is very good because gives space for enriching the model, with variables like the clearance time of each article, and shows that the whole cross-docking strategy may not have a big impact on the shipping service. The time in transit, in the simulation, includes the time of traveling between the boutiques and the cross-dock in Italy, the time to unload the truck and, if the truck brings items that came to the cross-dock by plane, this time is multiplied by two, to simulate the time that would take to unload the packages from the plane. After the unloading, the time in transit updates when the package leaves the Italian cross-dock, and total time inside this dock is added to the time in transit. The total time necessary to load the trucks is added three times to the total in transit, one for the actual time of loading all vehicles, the second addition is to simulate the time to load the packages into the plane, and the third is to account the time to unload the plane in America. When the packages arrive in America, the time of traveling is added to the 4.2 Transatlantic Bridge 53 time in transit, as well the time to unload the boxes from the trucks, in the American cross-dock, and, again, the time inside the dock is added when the boxes leave the facility. If the packages need to go by plane, then the same logic of adding the time of loading three times happens, and, when the vehicle arrives at its destination, the last addition happens that consists on adding the time of the travel between the facility and the destination to the time in transit. 54 Proposed Solutions Chapter 5 Conclusions and Future Work 5.1 Conclusions One of the major goals of any e-commerce platform is to fully satisfy the client. In Farfetch case, one point affecting the client’s satisfaction is the way of delivering when the client purchases more than one article. The main goal of the present project is to study cross-docking and how could it be implemented in the company. Cross-docking is something totally new in the company, due to this there was nothing to start from, and was important to understand what is this logistic strategy, setting what needs to be study to successfully implement it. The literature review allowed to understand the main operations inside a cross-dock and the steps needed to implement this type of logistic strategy. The operations identified, in the order of occurrence, is the unloading of the trucks, followed by moving the packages from the unload docks to the load docks. If the package is suitable for aggregation, then the next operation is the consolidation of packages, and the last operation is the loading of the packages into the loading trucks. To simulate this strategy in Farfetch was used a software called AnyLogic. Before staring the simulation of cross-docking, the current situation in Farfetch needed to be studied. First, were identify the different parts of Farfetch’s order processing system that will be affected if cross-docking is implemented, and, for each, part identified, was done an As-Is analyze. Cross-docking was analyzed in two different situations. The first case study, the Chinese Case, focus on understanding what needs to change in Farfetch to allow cross-docking and in the logic inside the cross-dock with the aim to gain knowledge about the variables that impact crossdocking and the operations that happen inside a cross-dock. To understand what needs to change in Farfetch, was done an To-Be analyze for all the part of the system in the As-Is analyze. With these two analysis is easy to know how Farfetch is before cross-docking and what will it be after implementing cross-docking, allowing the evaluation of the necessary work to make this strategy happen, in terms of system’s configurations, and what are Farfetch’s current limitations. After the theoretical analyze, the programming of the simulation begun. The first step in the simulation was decide the shape of the cross-dock, followed by building the facility and setting 55 62 REFERENCES [32] Kevin R. Gue and Keebom Kang. “Staging queues in material handling and transportation systems”. In: Proceedings of the 33nd conference on Winter simulation, WSC 2001, Arlington, VA, USA, December 9-12, 2001 (2001), pp. 1104–1108. DOI:10.1109/WSC.2001. 977421.URL:https://doi.org/10.1109/WSC.2001.977421. [33] Target Distribution Center Network | MWPVL.http : / / www . mwpvl . com / html / target.html. Last time visited: February 27th of 2018. [34] John J Bartholdi and Kevin R Gue. “The Best Shape for a Crossdock”. In: Transportation Science 38.2 (2004), pp. 235–244. ISSN: 1526-5447. DOI:10.1287/trsc.1030.0077. URL:https://www2.isye.gatech.edu/people/faculty/John_Bartholdi/ papers/crossdock-shape.pdf. [35] ˙ Ilker Küçüko˘ glu. “The Effects of Crossdock Shapes on Material Handling Costs”. In: ISSN || International Journal of Computational Engineering Research 10 (2016), pp. 2250–3005. URL:www.ijceronline.com. [36] M W P Savelsbergh and M Sol. “The General Pickup and Delivery Problem.” In: Transportation Science 29.1 (1995), pp. 17–29. DOI:10.1287/trsc.29.1.17.URL:http: //pubsonline.informs.org.%20https://. [37] Mu-Chen Chen et al. “The Self-Learning Particle Swarm Optimization approach for routing pickup and delivery of multiple products with material handling in multiple cross-docks”. In: Transportation Research Part E: Logistics and Transportation Review 91 (July 2016), pp. 208–226. ISSN: 1366-5545. DOI:10.1016/J.TRE.2016.04.003.URL:https: //www.sciencedirect.com/science/article/pii/S1366554515300600. [38] Warisa Wisittipanich and Piya Hengmeechai. “Truck scheduling in multi-door cross docking terminal by modified particle swarm optimization”. In: Computers and Industrial Engineering (2017). ISSN: 03608352. DOI:10.1016/j.cie.2017.01.004. [39] What is LTL (Less Than Truckload) Shipping? | Freightquote.https://www.freightquote. com/blog/what-is-ltl-shipping. Last time visited: February 20th of 2018. Annex A: Gantt Chart 63 64 Annex B: Queries Query Used to Obtain the Data for the Simulation of the Chinese Case select X.Country as StockPointCountry, X.CountryId as StockPointCountryId, GLB.Country as UserCountry, GLB.CountryID as UserCountryId, GLB.City as UserCity, GLB.OrderCode as PortalOrderId, GLB.OrderId as MerchantOrderId, GLB.DataCriado as CreatedDate, EstimatedDeliveryDate as EDD from BI_SYNC..glborders GLB join [FFLIVE\BI Analysts].OPS_Del_KPI_Time EDD on GLB.OrderID =EDD.OrderID cross apply (select AD.* from BI_SYNC..StockPoint SP inner join BI_SYNC..AddressInfo AI on SP.IdShippingAddress =AI.IdAddressInfo inner join BI_SYNC..Address AD on AI.IdAddress =AD.IdAddress where SP.stockpointkey = GLB.siteid )X where GLB.DataCriado >= ’2017-11-01’ and GLB.DataCriado <’2017-12-01’ and ( GLB.Country =’Hong Kong’ or GLB.Country =’China’) and EstimatedDeliveryDate is not null and EDD.siteid =GLB.SiteID order by GLB.DataCriado ASC Query Used to Obtain the Information about the Boutiques Placed on the Map in the Transatlantic Bridge Case Simulation select top 401 IdStockPoint,Address1,Country,City,ZipCode from BI_SYNC..StockPoint SP 65 inner join BI_SYNC..AddressInfo AI on SP.IdShippingAddress =AI.IdAddressInfo inner join BI_SYNC..Address AD on AI.IdAddress =AD.IdAddress where Country <> ’Australia’ and Country <> ’Brazil’ and Country <> ’Canada’ and Country <> ’United States’ and Country <> ’China’ and Country <> ’Hong Kong’ and Country <> ’India’ and Country <> ’Japan’ and Country <> ’Korea, Republic of’ and Country <> ’Kuwait’ and Country <> ’Lebanon’ and Country <> ’Macau’ and Country <> ’Malaysia’ and Country <> ’Morocco’ and Country <> ’New Zealand’ and Country <> ’Georgia’ and Country <> ’Saudi Arabia’ and Country <> ’Singapore’ and Country <> ’South Africa’ and Country <> ’Taiwan’ and Country <> ’UAE’ and City <> ’not set’ order by IdStockPoint Query Used to Obtain the Data for the Simulation of the Transatlantic Bridge Case select X.IdStockPoint as IdSTockPoint, GLB.Country as OrderCountry, GLB.State as OrderState, GLB.OrderCode as PortalOrderId, GLB.OrderId as MerchantOrderId, GLB.DataCriado as CreatedDate, EstimatedDeliveryDate as EDD from BI_SYNC..glborders GLB join [FFLIVE\BI Analysts].OPS_Del_KPI_Time EDD on GLB.OrderID =EDD.OrderID cross apply (Select SP.* from BI_SYNC..StockPoint SP inner join( select top 401 IdStockPoint,Address1,Country,City,ZipCode from BI_SYNC..StockPoint SP inner join BI_SYNC..AddressInfo AI on SP.IdShippingAddress =AI.IdAddressInfo inner join BI_SYNC..Address AD on AI.IdAddress =AD.IdAddress where Country <> ’Australia’ and Country <> ’Brazil’ and Country <> ’Canada’ and Country <> ’United States’ and Country <> ’China’ and Country <> ’Hong Kong’ and Country <> ’India’ and Country <> ’Japan’ and Country <> ’Korea, Republic of’ and Country <> ’Kuwait’ 66 and Country <> ’Lebanon’ and Country <> ’Macau’ and Country <> ’Malaysia’ and Country <> ’Morocco’ and Country <> ’New Zealand’ and Country <> ’Georgia’ and Country <> ’Saudi Arabia’ and Country <> ’Singapore’ and Country <> ’South Africa’ and Country <> ’Taiwan’ and Country <> ’UAE’ and City <> ’not set’ order by IdStockPoint)SPId on SP.IdStockPoint =SPId.IdStockPoint where SP.stockpointkey =GLB.siteid)X where GLB.DataCriado >= ’2018-01-01’ and GLB.DataCriado <’2018-04-01’ and GLB.Country =’United States’ and EstimatedDeliveryDate is not null and EDD.siteid =GLB.SiteID order by GLB.DataCriado ASC 67 68 Annex C: Chinese Case’s Statistics Static Pie Chart for the Amount of Orders that can Suffer Aggregation 69 Static Pie Chart Divides the Orders that can Suffer Aggregation Based on the Amount of Articles 70 Non Static Pie Chart that Shows the Amount of Packages that Suffer Consolidation 71 Excerpt of Sheet "operations_report" from the Report Generated for the Case with 3 Pick-ups day packages received packages aggregated orders aggregated incoming trucks outgoing trucks % of packages aggregated 2017-11-02 5 0 0 1 4 0 2017-11-03 49 8 4 24 17 16,32653061 2017-11-04 0 0 0 0 0 0 2017-11-05 1 0 0 1 1 0 2017-11-06 651 222 83 24 19 34,10138249 2017-11-07 162 116 37 27 19 71,60493827 2017-11-08 918 314 125 24 19 34,20479303 2017-11-09 694 397 136 24 20 57,20461095 2017-11-10 718 432 163 24 18 60,16713092 2017-11-11 0 0 0 0 0 0 2017-11-12 1 0 0 1 1 0 2017-11-13 1625 562 213 24 19 34,58461538 2017-11-14 864 436 143 24 18 50,46296296 2017-11-15 4592 1837 712 24 27 40,0043554 2017-11-16 3534 2551 853 24 26 72,18449349 2017-11-17 3806 2106 809 24 29 55,33368366 2017-11-18 1480 891 346 24 19 60,2027027 2017-11-19 0 0 0 0 0 0 2017-11-20 3971 1099 458 24 24 27,67564845 2017-11-21 1753 1250 441 24 22 71,306332 2017-11-22 3167 1149 440 24 28 36,28039154 2017-11-23 2124 1677 563 24 27 78,95480226 2017-11-24 2695 1586 601 24 30 58,84972171 2017-11-25 6 4 2 1 5 66,66666667 78 Excerpt of Sheet "operations_report" from the Report Generated for the Case with 4 Pick-ups with 6-Hour Time Gap day packages received packages aggregated orders aggregated incoming trucks outgoing trucks % of packages aggregated 2017-11-02 5 0 0 1 4 0 2017-11-03 49 8 4 24 17 16,32653061 2017-11-04 0 0 0 0 0 0 2017-11-05 1 0 0 1 1 0 2017-11-06 651 222 83 24 19 34,10138249 2017-11-07 162 116 37 27 19 71,60493827 2017-11-08 918 314 125 24 19 34,20479303 2017-11-09 694 397 136 24 20 57,20461095 2017-11-10 718 432 163 24 18 60,16713092 2017-11-11 0 0 0 0 0 0 2017-11-12 1 0 0 1 1 0 2017-11-13 1625 562 213 24 19 34,58461538 2017-11-14 864 436 143 24 18 50,46296296 2017-11-15 4592 1837 712 24 32 40,0043554 2017-11-16 3534 2551 853 24 36 72,18449349 2017-11-17 3806 2106 809 24 39 55,33368366 2017-11-18 1480 891 346 24 24 60,2027027 2017-11-19 0 0 0 0 0 0 2017-11-20 3971 1099 458 24 30 27,67564845 2017-11-21 1753 1250 441 24 27 71,306332 2017-11-22 3167 1149 440 24 31 36,28039154 2017-11-23 2124 1677 563 24 28 78,95480226 2017-11-24 2695 1586 601 24 35 58,84972171 2017-11-25 6 4 2 1 5 66,66666667 79 Excerpt of Sheet "operations_report" from the Report Generated for the Case with 4 Pick-ups with Different Time Gaps Between Them day packages received packages aggregated orders aggregated incoming trucks outgoing trucks % of packages aggregated 2017-11-02 5 0 0 1 4 0 2017-11-03 49 8 4 24 17 16,32653061 2017-11-04 0 0 0 0 0 0 2017-11-05 1 0 0 1 1 0 2017-11-06 651 222 83 24 19 34,10138249 2017-11-07 162 116 37 27 19 71,60493827 2017-11-08 918 314 125 24 19 34,20479303 2017-11-09 694 397 136 24 20 57,20461095 2017-11-10 718 432 163 24 18 60,16713092 2017-11-11 0 0 0 0 0 0 2017-11-12 1 0 0 1 1 0 2017-11-13 1625 562 213 24 21 34,58461538 2017-11-14 864 436 143 24 18 50,46296296 2017-11-15 4592 1837 712 24 37 40,0043554 2017-11-16 3534 2551 853 24 36 72,18449349 2017-11-17 3806 2106 809 24 39 55,33368366 2017-11-18 1480 891 346 24 27 60,2027027 2017-11-19 0 0 0 0 0 0 2017-11-20 3971 1099 458 24 33 27,67564845 2017-11-21 1753 1250 441 24 31 71,306332 2017-11-22 3167 1149 440 24 39 36,28039154 2017-11-23 2124 1677 563 24 37 78,95480226 2017-11-24 2695 1586 601 24 40 58,84972171 2017-11-25 6 4 2 1 5 66,66666667 80 Annex E: Chinese Case’s View Areas 3D’ View Area 81 2D’ View Area Block Logic’s View Area 82 Program & Stuff’ View Area Statistics’ View Area 83 84 Annex F: Transatlantic Bridge Case’s Statistics Static Pie Chart for the Amount of Orders that can Suffer Aggregation 85 Static Pie Chart Divides the Orders that can Suffer Aggregation Based on the Amount of Articles 86 Non Static Pie Chart that Shows the Amount of Packages that Suffer Consolidation 87 Excerpt of Sheet "operations_report" from the Report Generated for the Italian Cross-dock day packages received incoming trucks outgoing trucks 2017-11-02 24 12 18 2017-11-03 297 126 22 2017-11-04 0 0 0 2017-11-05 0 0 0 2017-11-06 1514 247 22 2017-11-07 724 203 22 2017-11-08 1794 291 22 2017-11-09 1357 241 22 2017-11-10 1202 253 22 2017-11-11 0 0 0 2017-11-12 4 4 4 2017-11-13 1737 276 22 2017-11-14 1132 219 22 2017-11-15 3051 294 22 2017-11-16 2384 269 22 2017-11-17 1989 252 22 2017-11-18 0 0 0 2017-11-19 4 4 4 2017-11-20 2675 306 22 2017-11-21 1853 234 22 2017-11-22 4919 303 33 2017-11-23 116 75 22 2017-11-24 6592 302 33 2017-11-25 1 1 1 2017-11-26 9 8 9 2017-11-27 8031 331 33 2017-11-28 3953 287 22 2017-11-29 9290 341 33 94 Excerpt of Sheet "report" from the Report Generated for the American Cross-dock portal order id arrival time departure time total hours in the dock num articles inside Z6BJUQ 01/12/2017 01:40:11 03/12/2017 15:06:19 61,435 7 VMMWUB 30/11/2017 08:52:38 02/12/2017 15:33:56 54,688 7 FND5DM 30/11/2017 09:15:51 02/12/2017 15:39:00 54,385 14 7AAPRZ 30/11/2017 09:20:25 02/12/2017 15:27:45 54,122 7 YWWVVJ 30/11/2017 09:33:59 02/12/2017 15:14:06 53,668 9 597WGA 30/11/2017 10:44:03 02/12/2017 15:24:48 52,679 13 8R8WKN 21/11/2017 23:46:21 24/11/2017 02:06:39 50,338 9 59UM5A 01/12/2017 03:07:31 03/12/2017 05:06:43 49,986 6 CRF4ZT 30/11/2017 13:40:23 02/12/2017 15:03:39 49,387 8 3553ZS 30/11/2017 15:11:03 02/12/2017 15:25:26 48,239 7 59W6LA 01/12/2017 02:16:46 03/12/2017 02:11:41 47,915 9 8VR6QN 29/11/2017 02:43:26 01/12/2017 02:17:58 47,575 6 N3DXW5 21/11/2017 21:59:01 23/11/2017 15:47:30 41,808 14 2N4RX2 21/11/2017 22:32:03 23/11/2017 16:12:25 41,672 11 59ZFNA 30/11/2017 02:59:19 01/12/2017 20:20:05 41,346 13 RUF4LK 15/11/2017 22:53:24 17/11/2017 15:30:16 40,614 6 BE9GBX 28/11/2017 23:20:09 30/11/2017 15:41:50 40,361 10 SBT4SF 15/11/2017 23:18:35 17/11/2017 15:36:50 40,304 15 CQ8RAT 22/11/2017 00:01:53 23/11/2017 16:08:54 40,116 6 9BB5L9 30/11/2017 10:30:18 02/12/2017 02:27:27 39,952 7 PCZ4E3 30/11/2017 10:36:12 02/12/2017 02:25:03 39,814 8 Z66CFQ 01/12/2017 23:30:28 03/12/2017 15:04:43 39,57 3 59CB6A 01/12/2017 00:09:26 02/12/2017 15:41:08 39,528 6 W3297U 01/12/2017 23:52:03 03/12/2017 15:13:28 39,356 8 SB89QF 16/11/2017 00:46:43 17/11/2017 15:27:58 38,687 11 GJUGR6 30/11/2017 12:13:23 02/12/2017 02:49:09 38,596 7 W33AZU 01/12/2017 00:41:13 02/12/2017 15:15:19 38,568 7 ZPMM8Q 16/11/2017 00:53:31 17/11/2017 15:27:27 38,565 8 68Z3HD 01/12/2017 00:55:42 02/12/2017 15:19:27 38,395 6 KL6EE7 28/11/2017 02:00:33 29/11/2017 15:52:09 37,86 18 95 Excerpt of Sheet "operations_report" from the Report Generated for the American Cross-dock day packages received packages aggregated orders aggregated incoming trucks outgoing trucks % of packages aggregated 2017-11-02 24 4 2 2 4 16,66666667 2017-11-03 297 82 32 3 27 27,60942761 2017-11-04 0 0 0 0 22 0 2017-11-05 0 0 0 0 0 0 2017-11-06 1514 229 104 9 41 15,12549538 2017-11-07 724 346 131 4 68 47,79005525 2017-11-08 1794 363 148 10 74 20,23411371 2017-11-09 1357 564 219 8 83 41,56226971 2017-11-10 1202 460 184 7 79 38,26955075 2017-11-11 0 131 51 0 43 0 2017-11-12 4 0 0 1 3 0 2017-11-13 1737 188 81 10 39 10,82325849 2017-11-14 1132 463 188 7 77 40,90106007 2017-11-15 3051 382 155 17 82 12,52048509 2017-11-16 2384 1356 500 13 111 56,87919463 2017-11-17 1989 1499 504 11 112 75,36450478 2017-11-18 0 489 172 0 60 0 2017-11-19 4 2 1 1 2 50 2017-11-20 2675 198 88 15 45 7,401869159 2017-11-21 1853 868 338 10 100 46,84295737 2017-11-22 966 550 202 5 93 56,93581781 2017-11-23 4069 2401 942 23 141 59,00712706 2017-11-24 1081 79 35 6 80 7,308048104 2017-11-25 5512 2321 933 30 159 42,10812772 2017-11-26 9 0 0 1 16 0 2017-11-27 1102 19 9 6 55 1,724137931 2017-11-28 10818 2288 969 57 165 21,14993529 2017-11-29 1141 1822 680 6 140 159,6844873 96 Excerpt of Sheet "time_in_transit" from the Report Generated for the Transatlantic Bridge Case origin destination time in transit days time in transit hours num of packages United Kingdom Connecticut 0,628496215 15,08390917 1 Switzerland NorthCarolina 0,625301921 15,00724611 1 Netherlands NorthCarolina 0,661255648 15,87013556 1 United Kingdom Florida 0,709749815 17,03399556 1 United Kingdom Colorado 0,770483727 18,49160944 1 Portugal NorthCarolina 0,724749398 17,39398556 2 Italy NorthCarolina 0,717612199 17,22269278 2 Italy NorthCarolina 0,719166655 17,25999972 1 Portugal NorthCarolina 0,722495787 17,33989889 2 Portugal NorthCarolina 0,728874525 17,49298861 2 Portugal NorthCarolina 0,726800509 17,44321222 1 Portugal NorthCarolina 0,725720972 17,41730333 1 Portugal NorthCarolina 0,723634306 17,36722333 1 Portugal NewMexico 0,846636065 20,31926556 1 Portugal NewMexico 0,845854213 20,30050111 1 Portugal NewMexico 0,844426991 20,26624778 1 Portugal NewMexico 0,840401343 20,16963222 1 Italy Connecticut 0,561284074 13,47081778 1 Italy Connecticut 0,569714618 13,67315083 1 Spain Connecticut 0,562739988 13,50575972 1 France Connecticut 0,552729132 13,26549917 1 Belgium Connecticut 0,522342037 12,53620889 1 Germany Maryland 0,561289421 13,47094611 1 France Maryland 0,557562951 13,38151083 1 Germany Maryland 0,550086215 13,20206917 1 Switzerland Massachusetts 0,591786782 14,20288278 1 Italy Massachusetts 0,578163194 13,87591667 1 97 Pivot Table Showing the Average Time Between All the European Countries and the USA 98 Pivot Table Showing the Average Time to All the American States 99 100 Annex H: Transatlantic Bridge Case’s View Areas World Map’s View Area 101 Italian Cross-Dock’s View Areas 102 103