Developing the distribution network of a healthcare logistics company considering different types of customers

Developing the distribution network of a healthcare logistics

company considering different types of customers

Master’s Thesis, MSc Technology and Operations Management,

Faculty of Economics and Business, University of Groningen, The Netherlands

Student: Ioannis Kranidis

Student number: S3560414 / Email: i.kranidis@student.rug.nl

First Supervisor: Dr. Ir. S. (Stefano) Fazi Second Supervisor: Dr. O.A. Kilic

Abstract

The present study is focused on cost and benefit analysis of different distribution networks for the delivery of pharmaceuticals, considering the case of a Greek third-party logistic (3PL) company, active in Central Greece. As of now, the system consists of a central warehouse where all dispatches are originated and headed directly to the final customers by means of small trucks and vans. However, a Cross-Dock (CD) facility in the network may help reducing operational cost and CO2 emissions, given the peculiarities of the distribution region. For instance, big trucks cannot deliver to individual clients, to mountainous spots and to heavy traffic city centers. With the existence of a CD facility, big trucks will be able to transfer their deliverables to small trucks or vans, which will be responsible for the final delivery. Furthermore, the selected delivery fleet transports medicines from regional established CD to individuals incurring a dramatic reduction of operational delivery network cost. In this study, firstly the best locations are selected, among eight possible, for the CD facility, by means of a Multi-Criteria Decision Analysis (MCDA). Next, a cost analysis is performed with the help of an integrated in-house model and with data obtained from the case study, comparing the current Scenario against a few settings considering CD facilities. Results indicate that the implementation of a properly located CD facility may reduce the overall yearly costs and carbon dioxide emissions. Due to the efforts of 3PL distribution companies to reduce the costs of delivering products, there is a growing need for an integrated computing tool. Our analysis comprises delivery times, transportation vehicle selection, overall operational cost and CO2 emissions.

Keywords: Distribution Network Design, Pharmaceuticals Network Design, Pharmaceutical Supply

Preface

This Thesis is the final project for the completion of my Master program in Technology and Operations Management. This topic is purposely selected in order to come closer to the Logistics sector and familiarize myself by acquiring critical knowledge, a sector that I would like to get involved with professionally in the future. Apart from that, pharmaceutical logistics is a field that entails logistics peculiarities, really provoking to reveal. I thank warmly Dr. Stefano Fazi for his well-aimed advices, comments, guidelines and assessments, which helped me to overcome the barriers and complete in time my Thesis in a fruitful way. He was always at my disposal for queries and performed a remarkable willingness to be helpful.

I also want to thank Dr. Onur Kilic for the final assessment of my Thesis. Furthermore, I would like to thank very much two executive employees from Unilog. Firstly, Mr. Labros Thireos, the Head of Logistics department, who provided all the data needed for my research and advised me on how to proceed to the structure of the computational integrated built-in model. Secondly, Mr. Panagiotis Ioakimidis and Mr. Stefanos Giakoumatos, both senior environmental consultants and chemical engineers, external partners of UNILOG, who provided me with all the necessary data to estimate the environmental operation footprint of each scenario herein presented.

Contents

2.1 Literature on distribution network in pharmaceutical supply chains ... 11

2.2 Literature on Cross-Docking ... 13

2.3 Contribution ... 15

3. Problem and case description ... 17

3PL case study of our concern ... 17

4. Methodology ... 19

4.1 Selection of Cross Dock facilities using MCDA ... 22

4.2 Tool to calculate operational cost and CO2 emissions ... 27

Figures

Figure 1: The map of southern/western Greece with the 8 main destinations to be served depicted

by red curved arrows. ... 18

Figure 2: Regional segmentation (clusters) and location of the candidate CD spots of the delivery strategy... 20

Figure 3: Loading and unloading practice of a 48-case (cartons) pallet ... 21

Figure 4: Flow chart of orders delivery via a 3PL company ... 21

Figure 5: Cost calculations Model of drugs from DC to customers ... 22

Figure 6: Model structure for the optimal CD location ... 24

Figure 7: Truncated decision tree of optimized vehicle fleet selection, in case of CD establishment . 32 Figure 8: Truncated decision tree of optimized vehicle fleet selection in case of no CD establishment ... 34

Figure 9: Regional segmentation (clusters) and routes on the delivery strategy of Basic Scenario 1. 37 Figure 10: Regional segmentation (clusters) and routes on the delivery strategy (CD in Patras Scenario 2) ... 38

Figure 11: Regional segmentation (clusters) and routes on the delivery strategy (CD in Tripolis Scenario 3) ... 39

Figure 12: Operational Cost in Euros per case delivered in the four predetermined geographical areas of the research for the three Scenarios investigated. ... 43

Figure 13: Weekly cost per Scenario ... 44

Figure 14: CO2 emission (tn/year) per Scenario ... 44

Figure 15: Distance covered (in km) per vehicle type employed on weekly basis for Basic Scenario 1 45 Figure 16: Distance covered (in km) per vehicle type employed on weekly basis for Scenario 2 ... 46

Tables

Table 1: Yearly and daily cost to operate CD ... 28

Table 2: Types of used vehicles and their respective capacity in cases and pallets. ... 28

Table 3: Shipping cost of the used vehicles types expressed in €/ Km. ... 29

Table 4: Distances between DC and clusters [32]. _{... 29}

Table 5: Tolls for possible routes and vehicle type [31] _{... 30}

Table 6: Vehicle types available for use from DC to CD ... 31

Table 7: Vehicle types available for use from CD to clusters ... 33

Table 9: Parameters to calculate CO2 emissions of Scenario 2 ... 41

Table 10: Ranking of the candidate CD spots on the delivery strategy. ... 42

Table 11: Summarized outcome data of the examined scenarios ... 47

Table 12: Orders and cases per customer type and per cluster ... 57

Table 13: Criteria evaluated regarded the real estate parameters of candidate CDs ... 58

Table 14: Criteria evaluated regarded the distances and centrality from the DC to the candidate CDs. ... 59

Table 15: Criteria evaluated regarding the number of cases/pallets per cluster & the number of clients per annum ... 60

Abbreviations

AHP Analytic Hierarchy Process

ANP Analytic Network Process

BOMILP Bi-Objective Mixed Integer Linear Programming

CD Cross Docking

DC Distribution Center

EPA European Environmental Agency

IT Inbound Trucks

MCDA Multi Criteria Decision Analysis MHSA Ministry of Health and Social Affairs

ΝΜΑ National Medicines Agency

OT Outbound Trucks

PSCND Pharmaceutical Supply Chain Network Design

QFD Quality Function Deployment

SAW Simple Additive Weighting

WSM Weighted Sum Model

1. Introduction

The global health economy is developing faster, compared to the overall economy. However, despite the growth in manufacturing, storage, and distribution technologies, many pharmaceutical companies are still unable to satisfy the market demands in a steady way (Mousazadeh et al., 2015). This results in inequity in healthcare provision and complexity in the healthcare systems, which highlights the need for efficient management of healthcare supply chains. This need becomes more important taking into account that the pharmaceutical industry is one of the major drivers of the healthcare sector (Narayana et al., 2014).

The supply chain of pharmaceuticals to final customers is typically carried out by 3PL providers, who take the pharmaceuticals from the producers and take care of their dispatch from central warehouses (Hellenic Association of Pharmaceutical Companies, 2018). A crucial aspect of the transportation network of pharmaceuticals is the customers’ requirements and locations (Kumar et al., 2009). In specific regions, the use of cross dock facility to change the type of vehicles for the last delivery may be necessary. For example, in parts of Greece, the morphology of the terrain makes it a necessity to use vans or small trucks to reach the final recipients e.g. pharmacies to remote villages and difficult to approach down towns. Due to the absence of intermediate cross-dock facilities, these vans start their journey far away from the final destinations.

the allocation of customer demands to distribution centers. Yet, the optimal selection of transportation fleet and the clients’ time windows based on yearly delivery practices were not taken into consideration. The deployment of CD facilities in a supply chain is recognized as high-risk investment due to large costs for starting up and guaranteeing constant drug flow. Hence, there is the need of calculation tools to be implemented that assess the benefits and costs of different settings in order to give proper managerial directions. To our knowledge, literature does not provide a simple and integrated tool for decision makers of supply chain operations, to assess costs and benefits of CD implementation. The above given publications of Izadi and Kimiagari (2014), Martins et al. (2017) and Mousazadeh et al. (2015) do not approach the problem in multi-parametric paths, which means involving CD candidate decision making, optimal selection of the engaged transportation fleet and making right exploitation of real data provided by 3PLs with great clientele share in pharmaceutical logistics. Additionally, retail and wholesalers’ drugs delivery network data are not employed simultaneously, in the aforementioned researches.

The present research assesses different network configurations for a Greek 3PL company, considering the potential use of intermediate CD facilities. This can be done by means of a two-step approach. Firstly,the best location of the CD facility is drawn, by means of a MCDA, which compares eight different candidate sites in pre-decided geographical entities (clusters). Secondly, three different Scenarios are developed. The first one, the basic and current Scenario, corresponds to the common transportation setting of a 3PLs in Greece. This Scenario entails a central warehouse (Distribution Center, abbreviated DC), where all dispatches are originated and directed to the final customers/recipients. The second and third Scenarios consists of a single CD of unlimited storage capacity between DC and the final customers considering the two highest ranked location from the MCDA method.

2. Theoretical Background

2.1 Literature on distribution network in pharmaceutical supply chains

According to Haial et al. (2016), transportation network design is one of the most important decision problems in supply chain management. They introduced a framework which deals with the design of pharmaceutical supply chain and comprises three steps: (i) identification of the configuration of the current distribution network, (ii) design of an optimal distribution network while determining the appropriate location-allocation decisions, and (iii) choosing the most appropriate transportation network strategy through the application of a MCDA method. Transportation strategy for pharmaceutical supply chains entails identification of the configuration of the current distribution network, optimal distribution network design and the choice of the most appropriate transportation network strategy through the application of a MCDA method.

Knop (2019) presented an evaluation mechanism to bring out improvements of drug supply chain network based exclusively on customers’ expectations & services satisfaction. It was considered to be one-sided, since logistics is skewed on the side of the final recipients (customers), disregarding critical working parameters of 3PL operators and overall operational constrains and cost.

regional operational particularities, such as those in southern Greece with a short list of clients to serve.

The design of a supply chain of pharmaceuticals is always a challenge and need to be as much reliable and efficient as it can be. For the design purposes of a Pharmaceutical Supply Chain Network Design (PSCND), Mousazadeh et al. (2015), developed a Bi-Objective Mixed Integer Linear Programming (BOMILP) model and tested the theoretical functionality via an applied case study. This model was solved via an efficient memetic algorithm with adaptive variable to make several decisions as opening of pharmaceutical manufacturing centers and main/local distribution centers.

Settanni et al. (2017) published an integrated system approach to Pharmaceutical Supply Chain modelling. It involves problem conceptualisation with boundary definition; design, formulation and solution of mathematical models, through practical implementation of identified solutions.

Franco & Lizarazo (2017) concluded an exhaustively review research of Quantitative Models of the Pharmaceutical Supply Chain. He classified all models accessed into three categories, which are summarized as network design, inventory models and optimization of a pharmaceutical supply chain. Most of the articles in the literature are focused on the optimization of the pharmaceutical supply chain and inventory models but the field on supply chain network design is still not deeply studied.

2.2 Literature on Cross-Docking

Cross-Docking is as a relative new logistics concept which hybridizes intermediate nodes into a distribution supply chain system. CDs are part of operational strategies to tackle successfully strict delivery schedules, proactive orders handling and optimized delivery routing to the final recipients. Li et al. (2012) defines Cross-Docks as the operational strategy that moves items through consolidation centers without putting them into storage condition. According to Lahmar (2007) in CD facilities products are moved very fast and directly from inbound trucks (ITs) to outbound trucks (OTs), after being resorted or consolidated with limited storage needs, normally not exceeding 24 hours. Furthermore, Boysen et al. (2008) defined CD facility as a consolidation point in a distribution network, where multiple smaller shipments can be merged to full truck loads in order to rearrange scale economies in transportation.

In this context, truck scheduling is getting very crucial and, in many aspects, determines truck succession loading/unloading processing making use of CD unloading platforms to catch up time frame deliveries. Haial et al. (2016) introduced an operational framework which deals with the design of a transportation strategy for pharmaceutical supply chains. In order to select the most appropriate transportation network strategy, a MCDA method was adopted. CD facilities can be approached by categorizing them into two main groups. According to the first approach, facilities could be considered as a node within a larger transportation network (Dobrusky, 2003) and the second approach entails the operational part of the facility (inbound/outbound doors & staging), and all efforts are made to obtain the optimization at each part of the operational stage (Bozer et al., 2008), (Boysen et al., 2008), (Konur et al., 2013).

problem and proposed four criteria in a manner to draw comparison to each one of the algorithms and demonstrate the strengths of them.Similarly, CD time schedules optimization problems were approached by a variety of metaheuristic algorithms presented by Liao et al. (2012), Liao et al. (2013) and others.

An inappropriate selection of cross-docking center location may result in excessive transportation and handling costs. Masaeli et al. (2018) noted the importance of the strategic location of a CD. The minimum total cost is the main parameter affecting the design of CD for a single parcel-delivery company. According to this, there is a fixed cost to dispatch a vehicle, a variable transportation cost per unit distance consisting of fuel, long-term & regular maintenance costs. Given the distances between interconnected nodes, variable transportation costs can be computed for each vehicle.

Partovi et al. (2006), developed an analytic model for the facility location by providing the following three handling tools i.e. Quality Function Deployment (QFD), Analytic Hierarchy Process (AHP) and Analytic Network Process (ANP). Within his study were regarded both external and internal evaluation factors according to sustain competitive advantage. Mousavi et al. (2019) presented a two-stage mixed-integer programming (MIP) model for the location of cross-docking centers and scheduling trucks for the cross-docking distribution networks in the supply chain. In the strategic stage, is necessary to define the location of possible cross-docks establishments and allocate the trucks schedules to open CDs. CDs will supposedly be located within their customers’ radius coverage.

Ertek (2005) identified three types of CDs, acting either as transfer points, where inbound product flow is synchronized with outbound product flow or as Traditional distribution with warehouses, where warehouses serve as intermediate stage inventory points or even as direct shipments, also named “warehouse bypass”. According to the last type, all products are shipping directly from suppliers to the final demand points bypassing the warehouse/cross dock facility.

needs in CDs. This is only a partial approach to an integrated delivery chain network. Dobrusky (2003), categorize CD facilities into two main groups considering them as a huge network with interconnected nodes. He presents optimal CD location and allocation and carriers selection optimization. The final outcome is very close to our final scope through our developed model. Nevertheless, Dobrusky’s effort lacks customers’ particularities in terms of final delivery cost per case. In addition, the selection of a CD over another in terms of CO2 footprint is not examined in his model.

To summarize some gaps, Li et al. (2012) examined the operation of CDs in realistic level and integrating three different layers (planning, scheduling and coordination) but this operation was not tested in a holistic environment including DC and different customers. Haial et al. (2016) examined a transportation strategy considering a bottom down approach, from suppliers to customers. However, they did not take into consideration neither the use of different type of trucks that can reduce the total costs, nor the environmental footprint of the transportation. Most importantly, their developed model was not tested in a realistic setting with real data.

The calculation of the operation cost of CDs is quite simple, as we primarily take into account fixed costs. However, determining the location of the warehouse, as part of the pharmaceutical distribution network, is a complex problem that this Thesis intends to solve.

2.3 Contribution

The thesis intends to provide answers to the main following questions: • Could a CD be a new viable alternative for the proposed case study?

• What are the most important parameters in the pharmaceutical supply chain? • How to assess the best supply chain options by employing a CD establishment?

In literature, in general, there’s a lack of a developed, friendly interfaced, easy to use operational tool to reengineer 3PL pharmaceutical chain network. Many published research efforts about pharmaceuticals supply network are rather theoretical approaches and do not make use of logistic business real data which is the important part for an employee possessing managerial post to do the decision making, for instance a network reengineering when it is needed. They fail to cope with daily confronted barriers that arise many times out of the blue, even for routine dispatches.

3. Problem and case description

Due to the very competitive environment in Greece, 3PL companies in drug logistic services are obliged to continue to provide their customers with reliable, low cost services and to be able to increase their client’s inventory without adversely affecting their clientele.

A company active many years in this sector in Greece (named UNILOG) agreed a collaboration upon the improvement of its distribution network in the south western part of the country, as they are currently aiming to increase their clientele in some cities and need ways and means to consider it. The chosen firm amid others of the kind, possesses Distribution Center (DC) facilities in east part of Attica prefecture (30 km east of Athens), in central Greece, maintains a well expanded clientele network providing drugs’ logistic services to a plethora types of customers (individuals/clinics, hospitals, pharmacy warehouses/ medical centers, pharmacies), at different areas of “Peloponnesus” southern region in Corinth, Tripolis, Nafplion, Kalamata and a part of the western Greece in Pyrgos , Patras, Agrinio and Arta cities. The customers of the company in each of the 8 areas are divided in 4 main aforementioned categories. All categories are part either of wholesalers or retail branch.

The delivery starts from a central warehouse DC to deliver the orders to customers into eight different areas and in distances from 110 Km to 340km in the same day (Figure 1). Consequently, the delivery of the orders to its customers is done by van and small trucks, because it is not possible to reach certain areas with large trucks.

Figure 1: The map of southern/western Greece with the 8 main destinations to be served depicted by red curved arrows.

4. Methodology

The following chapter is divided into two subchapters which refer relatively to the two main steps that we followed for our research. The first step was to select the best two locations (among 8) for the establishment of a CD, with the help of the MCDA method. In the second step, three different Scenarios are formulated and assessed with the help of Excel, in order to determine the best one, concerning costs and CO2 emissions. The first Scenario is the current setting of the company, the second one is formulated with a CD established in the best location and the third one with a CD in the second-best location.

Figure 2: Regional segmentation (clusters) and location of the candidate CD spots of the delivery strategy

Figure 3: Loading and unloading practice of a 48-case (cartons) pallet

The flow chart (business cycle), from the time the order is received by the customers of pharmaceutical companies, to the sequent stage of ordering a 3PL service company to deliver the drug commodity to the final recipient (client), is shown in Figure 4.

Customers from 8 clusters 3PL co Cost screen Pharmaceutical companies CD Delivery to customers Orders Orders 2nd _option

Drugs delivery 1 option

Drugs delivery

Figure 4: Flow chart of orders delivery via a 3PL company

Cost calculations Model of drugs

from DC to customers

Matrix to select the vehicle type based on the Number of “cases”

& driving time from DC

CD operational Cost

Vehicle operational

Cost

Cost per case per

customer

Minimum Overal cost to

deliver “cases” to

customers

Figure 5: Cost calculations Model of drugs from DC to customers

4.1 Selection of Cross Dock facilities using MCDA

evaluations. To achieve improvement of the selected supply network the method of MCDA was used to assess firstly the best available location to operate a CD. Taking the advantage of the close cooperation with 3PL Pharmaceutical Company chief quality employee, numerous real data and valuable information were acquired. The existence of all these data reinforced our decision to develop a MCDA model instead of making use of commercially available optimization operational models. The selection of developing a MCDA model enable the user to try out the results with realistic operational values, to justify model’s outcome so at to proceed to further minor adjustments.

The number of the enlisted customers i.e. approximately 900/ per day, renders the problem analysis highly complex. According to the relevant literature such a number of customers necessitates great computational time to be consumed and is hard to be setup. Therefore, our initial strategy was to aggregate all wholesalers and retail customers into clusters. By creating clusters, (zones quasi operational independent), we reduce the number of variables significantly and thus the complexity of the problem.

Sum of corrected ranking scores of each candidate city

CLUSTER No 1, Nr. customers, Nr. Cases CLUSTER No 2, Nr. customers, Nr. Cases CLUSTER No 3 Nr customers, Nr Cases Cluster No 4, Nr. customers, Nr. Cases Cluster No 5, Nr. customers, Nr. Cases Cluster No 6, Nr. customers, Nr. Cases Cluster No 7, Nr. customers, Nr. Cases Cluster No 8, Nr. customers, Nr. Cases Criterion 1, CD Real estate per cluster Criterion No 2, Nr. of clients per cluster Sub criteria 1. I. Real estate availability, II. Real estate acquisition price/ rent, III. Proximity to highway, IV. Trucks maneuverability, V. Easiness to employ personnel, VI. Permits acquisition easiness, Criterion No 5 Centrality Criterion No 4 kilometric distances to be covered

Selection of CD with Minimum Overall cost to deliver “cases” to customers to 8 clusters

Criterion No3 Nr. overall cases number per cluster

Sub criteria 2.

I. The frequency of the delivery stops II. The number of pallets to be delivered on weekly basis for an individual, III. Cargo’s workload Sub criteria 5. I. The centrality of the candidate CD spot to be selected among the city’s destinations II. The centrality (corrected), taking into account the distances among the candidate cities from the

distribution center in Attica region

Figure 6: Model structure for the optimal CD location

making by means of an overall score that should be directly comparable. We took into account methods such as the Simple Additive Weighting (SAW) method (Zhou & Fan, 2007) and Weighted Sum Model (WSM) method (Triantaphyllou, 2000), where an evaluation scale was set, in order to build our case within the range (0-1). Our five criteria and the weighting factors were selected carefully after thorough interviews with 3PLs employees. Basic criteria include sub-criteria that take certain scores normalized/corrected through a proper scale before use. The approach to 8 CD selection spots and operation Scenarios entails the following considerations:

a) the segmentation of the region of our concern into geographical clusters b) the analytical recording of the clientele in the region

c) the quantities’ records transferred on daily & weekly basis, d) the required number & type of cargo vehicles,

e) the customer service time tables,

f) cargo flows between the nodes amid geographical clusters i.e. neighbouring areas, g) the cost per case to service each customer

h) vehicle operation hours and i) truck’s loading capacity.

More than one CD site selection for a Scenario is excluded a priori, due to the relatively small distances in our regions. The MCDA incorporated criteria and sub-criteria introduced in worksheets i.e.

A) Criteria considered regarding the CD real estate that is: I. real estate availability,

II. real estate acquisition price/rent, III. proximity to highway,

IV. trucks maneuverability,

VI. permits acquisition easiness,

The initial score is resulted by means of a three scaled categorization (high/medium/poor) (see Table 13 in Appendix). Thereafter, the score is undergoing correction (normalized) properly so as to be added to the final score (see Table 16 in Appendix).

B) Furthermore, important criteria in MCDA are the number of clients registered in each geographical entity (cluster), which correspond to data acquired by 3PL company executive interview:

(i) the frequency of the delivery stops

(ii) the number of pallets to be delivered on weekly basis for an individual,

(iii) which is directly related to the cargo’s workload (see Tables 12 & 15 in Appendix). C) The overall cases number in a cluster to be delivered on weekly basis (see Tables 12 & 15 in Appendix).

Clusters that serve many clients get higher weighted overall volume credits, as many commodities by volume should be delivered there. Still, clusters that are located in areas with high nodes score, as a part of the overall routing map, play a more critical role, thereof possess a higher decision weight in the matrix.

D) Criteria are applied with the consideration of the kilometric distances to be covered during the transportation (see Table 14 in Appendix).

E) Last criterion represents the centrality introduced i.e.:

I. The centrality (interconnectivity) of the candidate CD spot to be selected among the city’s destinations (11 grades evaluation scale) and

All the aforementioned criteria are incorporated in the model and indicate the higher score, thereof the most suitable selected city out of 8 overall candidate ones. The city with the highest aggregate score derived from all criteria, receive the CD establishment. Upon that city, the truck selection optimization will be implemented.

4.2 Tool to calculate operational cost and CO2 emissions

Once the optimal locations of the CD installation are selected, the 2nd _{model step entitles the} user to perform, in an automated way, the calculation of delivering cost, the optimal transportation fleet selection and thereafter Co2 emissions estimation. Finally, the flexibility & effectiveness of the above program is tested by purchase prices and operational values given by the compact 3PL company, generating three different potentially viable Scenarios. The first Scenario was the current operational 3PL supply strategy, namely the direct delivery of the drugs to its customers. The second and third Scenarios were the delivery of the drugs by employing a CD in the network, with infrastructure to the two cities that had the highest score in their selection process through the MCDA method that is developed ad hoc.

Goods transportation is being conducted via Greek highways within strict speed limits and driving working hours (max 8 hour per day). They cannot exceed the obligatory maximum hours on driving duty, due to the potential fatigue and, consequently, car accident possibility. The verification of the delivery is done when each order reaches its destination.

The overall annual cost of a CD operation is the sum of individual cost i.e. the values of the 3rd column in Table 1 aregiven on yearly basis each.

Table 1: Yearly and daily cost to operate CD

CROSS DOCKING % OF TOTAL OPERATIONAL _COST €/FACILITY

Wages 45,45 17.000,00 Insurance expenses 5,35 2.000,00 Rent 22,46 8.400,00 Communication expenses 3,21 1.200,00 Equipment 2,14 800,00 Infra-structure expenses 21,39 8.000,00 TOTAL 100,00 37.400,00

working days per annum 250 days/year

CROSS DOCKING daily cost 149,60 €/day

The overall reduced (per km) cost of a certain type of transportation is calculated by means of the summation of the corresponding, to each vehicle type, column’s values in Euros (see Tables 2 & 3).In the above cost, toll fees were not co-estimated.

Table 2: Types of used vehicles and their respective capacity in cases and pallets.

Type of Vehicle _{Capacity (cases)} Transportation 1 Pallet to case _conversion Transportation Capacity (pallets) 34-pallet vehicle (big

truck) 1700

(30-50)* *value of 50 used for our calculations

34 10-pallet vehicle (small

truck) 500 10

Table 3: Shipping cost of the used vehicles types expressed in €/ Km.

SHIPPING COST UP TO 10-PALLET _{CARGO VEHICLE}

34-PALLET CARGO VEHICLE VAN CARGO VEHICLE COST DRIVER _{SHIPPING COST} % OF TOTAL €/km €/km €/km

Wages 63,10 0,404 0,555 0,315 Fuel comsumption 20,40 0,131 0,180 0,102 Vehicle maintenance 9,10 0,058 0,080 0,045 Vehicle tyres 3,30 0,021 0,029 0,016 Compulsory technical service 0,40 0,003 0,004 0,002 Taxes fees 0,40 0,003 0,004 0,002

Vehicle & Cargo

Insurance 3,40 0,022 0,030 0,017

TOTAL 100,00 0,640 0,880 0,499

Since the overall reduced cost per km is well defined and clearly estimated, it is easy to estimate the cost that a certain vehicle needs, in order to cover a distance, e.g. from 3PL establishments to the city of Patras given the distances between destinations (Table 4).

Table 4: Distances between DC and clusters [34].

Distances (km)

From / To Unilog Corinth Tripolis Nafplion Pyrgos Arta Patra Kalamata Agrinio

Unilog 0 114 190 169 326 379 250 270 313 Corinth 114 0 79 58 216 265 136 160 199 Tripolis 190 79 0 57 137 339 202 82 267 Nafplion 169 58 57 0 190 319 182 148 247 Pyrgos 326 216 137 190 0 252 97 117 180 Arta 379 265 339 319 252 0 153 369 85 Patra 250 136 195 178 97 153 0 210 81 Kalamata 270 160 82 148 117 369 210 0 297 Agrinio 313 199 267 247 180 85 81 297 0

Unilog Corinth Tripolis Nafplion Pyrgos Arta Patra Kalamata Agrinio

Table 5: Tolls for possible routes and vehicle type [32]

Tables 6 & 7 present the assessment achieved, in terms of the most preferable vehicle fleet to be selected, for the delivery of a certain number of ‘cases’ in both occasions i.e. destination from DC to CD and from CD to final recipients. Therefore, a random cell value of the column in the middle of the array (i.e. cases to be delivered) correspond to certain transportation fleet that gives the improved operational cost per km.

Our model’s mechanism (truncated decision tree), to do the decision making of the optimal vehicle fleet selection, for each cluster (with or without CD establishment), is given in Figures 7 and 8 respectively.

VAN CARGO VEHICLE 10-PALLET CARGO _VEHICLE 34-PALLET CARGO VEHICLE TOLL FEES € PER ROUTE TOLL FEES € PER _ROUTE TOLL FEES € PER ROUTE

ATHENS → PATRAS 14,6 36,7 52,7 ATHENS → PYRGOS 14,6 36,7 52,7 ATHENS → AGRINIO 37,45 80,5 129 ATHENS → ARTA 40,45 88,1 139,65 ATHENS → KALAMATA 19,5 49,2 71,74 ATHENS → TRIPOLIS 14,4 36,35 53,65 ATHENS → NAFPLIO 12 30,35 45,2 ATHENS → CORINTH 6,7 17 25,1 PATRAS → AGRINIO 6,55 16,5 23,1 PATRAS → ARTA 9,55 24,1 33,75 PATRAS → PYRGOS 0 0 0 PATRAS → KALAMATA 17,9 44,9 63,04 PATRAS → PATRAS 0 0 0 TRIPOLI → KALAMATA 5,1 12,85 18 TRIPOLI → NAFPLIO 2,4 6 8,45 TRIPOLI → CORINTH 4,9 12,35 17,35 TRIPOLI → TRIPOLIS 0 0 0

TRIPOLI → NAFPLIO CORINTH 7,3 18,35 25,8

Table 6: Vehicle types available for use from DC to CD

DESTINATION FROM DC TOWARDS CD OPTIMUM COST

(€/km) CASES PER OPTIMUM SELECTION OPTIMUM TRANSPORT SELECTION

1,101

201

1,602

501

2,202 401 2 VAN

2,402

1701

2,703 701 10-PALLET & VAN

3,203 1001 2 10-PALLET

3,303 601 3 VAN

3,504

1901

3,804 901 10-PALLET & 2 VAN

4,004

2201

4,304 1201 2 10-PALLET & VAN

4,404 801 4 VAN

4,805

3401

4,805 1501 3 10-PALLET

4,905 1101 10-PALLET & 3 VAN

5,105 2401 34-PALLET & 10-PALLET & VAN

5,405 1401 2 10-PALLET & 2 VAN

5,606 2701 34-PALLET & 2 10-PALLET

6,006 1301 10-PALLET & 4 VAN

6,406 2001 4 10-PALLET

6,507 1601 2 10-PALLET & 3 VAN

6,707 2901 34-PALLET & 2 10-PALLET & VAN

Initiation of truck fleet selection

Van operation

cost ≤200 casesLoad

10pallet truck operation cost 34pallet truck operation cost Loading capacity of each vehicle ≤8h/d driving Yes No Load ≤500 cases _Yes No Load ≤1700 cases _Yes No Load ≤1900 cases _Yes No Load ≤2200 cases _Yes No Load ≤3400 cases _Yes No Load >3400 cases _Yes No Yes Driving hours constrains ≤1,101 €/km ≤1,602 €/km ≤2,402 €/km ≤3,504 €/km ≤4,004 €/km ≤4,805 €/km No No No No No Termination out of limits No 1×10pallet optimal truck selection Yes 1×34pallet optimal truck selection Yes

1×34pallet & 1×van optimal truck selection Yes 1×34pallet & 1×10pallet optimal truck selection Yes 2×34pallet optimal truck selection Yes 1×van optimal truck selection Yes No

Table 7: Vehicle types available for use from CD to clusters

DESTINATION FROM CD TO FINAL RECIPIENTS (CLIENTS)

CASES PER OPTIMUM SELECTION OPTIMUM COST (€/km) OPTIMUM MEAN SELECTION

1,101

201

1,602

501

2,202 401 2 VAN

2,703

701

3,203

1001

3,303 601 3 VAN

3,804 901 10-PALLET & 2 VAN

4,304

1201

4,404 801 4 VAN

4,905 1101 10-PALLET & 3 VAN

5,405

1401

6,006 1301 10-PALLET & 4 VAN

6,406

2001

Initiation of truck fleet selection

Van operation

cost ≤200 casesLoad

10pallet truck operation cost 34pallet truck operation cost Loading capacity of each vehicle ≤8h/d driving Yes No Load ≤500 cases _Yes No Load ≤700 cases _Yes No Load ≤1000 cases _Yes No Load ≤1200 cases _Yes No Load ≤1400 cases _Yes No Load ≤2000 cases _Yes No Yes No Driving hours constrains ≤1,101 €/km ≤1,602 €/km ≤2,703 €/km ≤3,203 €/km ≤4,304 €/km ≤5,405 €/km No No No No Termination out of limits No 1×10pallet optimal truck selection Yes Yes Yes Yes 2×10pallet & 2×van optimal truck selection Yes 1×van optimal truck selection Yes Load >2000 cases _Yes No ≤6,406 €/km Yes 4×10pallet optimal truck selection 2×10pallet & 1×van optimal truck selection 2×10pallet optimal truck selection 1×10pallet & 1×van optimal truck selection No No

Figure 8: Truncated decision tree of optimized vehicle fleet selection in case of no CD establishment

Much information was gathered after long lasting interviews with the head of the Logistics and Quality Department employees, at a close distance.

Many design information is given below:

 The covered distances from the distribution center of Athens to delivery targets are fluctuating from 114 to 350 km.

Table 8: Time elapsed (in minutes) to deliver a case per client category

EMPICAL DATA PROVIDED FROM 3PL company INTERVIEWS

Time elapsed (min) /case delivered per client category (min)

Patients (individuals) 3,00

Pharmacy warehouses 0,40

Drug stores 1,50

Hospitals / medical centers 0,85

 The transportation vehicle categories are:

• Vans of 200 cases transportation capacity up to 1,5 t net cargo weight (mean weight 1,2 t),

• 10-pallet truck (mean weight 4,2 t), and • 34-pallet truck (mean weight 9,4 t).

 _{The 34-pallet truck due to its big dimension and thereof poor maneuverability is}

unsuitable to carry out door to door deliveries. It is surely the most economic transportation mean for big freights therefore the most suitable for routings from the DC to a selected CD.

 Many products have to be transported in specific specification of temperature and humidity. Thus, all trucks have to be equipped with adequate cooling systems. (products’ temperature-controlled standards).

 _{The 10-pallet and 34-pallet trucks have the ability to separate their loading volume} capacity with a movable partition wall, separating the temperature-controlled products from the non-temperature-controlled ones.

 _{The selection of a proper truck for a route towards or within a cluster is depended on the} number of orders given by the clients and subsequently of the number of pallets /cases to be delivered.

 _{The Mean Average speed of the used vehicles for highways is 60 km/h and for down town} 30 km/h.

 Deliveries of packages of medicines to residential areas are made only with small trucks  Maximum driving hours are 8 hours

 Daily opening hours of customers are:

o Drug stores (08:00-14:00 and 17:30- 19:30) o hospitals (07:00-15:00)

o pharmacy warehouses (06:00- 14:00) o patients (07:00-13:00 and 17:30- 20:30)

 Since the orders are given to the drivers the company cannot intervene in already scheduled routes until the transported goods reach the final destination (drug stores, warehouses, hospitals etc.), and therefore the delivery time is predetermined.

 _{Short-term consumable products need different handling between the production}

facilities’ gates and the final consignee. Therefore, additional limitations in routing plans should be taken into serious consideration given that we are aware of the percentage of such category to be transferred.

 The Cross Docking operational cost on annual basis according to the 3PL company

experience for south east cities of Greece is given in (Table 1).

 The Loading / Unloading time of large truck into a CD (up to 34 pallets) estimated to be about 30 min.

 The Loading / Unloading time of small trucks into a CD (up to 10 pallets loading capacity) estimated to be about 15 min.

 _{Small vans loading / unloading time into a CD (200 cases loading capacity) is about 15 min.} Considering the output of MCDA ranking we tackled three Scenarios.

a) The 1st _{Scenario (Basic Scenario) makes no use of Cross Docks facility at all. A dense} routing network of transport vehicles cover the clientele demands comprised of the following five routes (Figure 9):

• 1st _{route: Athens→Patras→Pyrgos} • 2nd _{route: Athens→Agrinio→Arta}

• 4th _{route: Athens→Patras} • 5th _{route: Athens→Kalamata}

Figure 9: Regional segmentation (clusters) and routes on the delivery strategy of Basic Scenario 1.

b) Scenario 2, having a CD in Patras and following routes (scored the highest rate of all 7 remaining CD candies) A routing network of transport vehicles that covers the clientele demands is comprised of the following six routes (Figure 10):

• ₅th _{route: CD in Patras→Pyrgos}

• 6th _{route: Athens→Corinth→Nafplio→Tripolis}

Figure 10: Regional segmentation (clusters) and routes on the delivery strategy (CD in Patras Scenario 2)

c) Scenario 3, having a CD in Tripolis and following routes (scored the second highest rate of all 7 remaining CD candies) (Figure 11). A routing network of transport vehicles that covers the clientele demands is comprised of the following six routes:

• 1st _{route: Athens→CD in Tripolis} • 2nd _{route: CD in Tripolis→Kalamata}

• 5th _{route: Athens→Patras→Pyrgos} • 6th _{route: Athens→Agrinio→Arta}

Figure 11: Regional segmentation (clusters) and routes on the delivery strategy (CD in Tripolis Scenario 3)

cost related to cost per 'case' is added to their shipping cost from DC to CD. As regards scenario 2, the estimated Mean reduced Cost (MC), in Euros, of a single ‘case’ was used for further calculations. Thereof:

�(MC)i

8 i=1

Likewise, (MCDOC) denotes Mean reduced CD Operational Cost (€) per cluster i.e.:

�(MCDOC)𝑖𝑖

8 i=1

Finally, (MVOC) denotes the Mean reduced Vehicle Operational Cost (€) per cluster i.e.:

�(MVOC)𝑖𝑖

8 i=1

And thereafter, Mean reduced Cost (€) per cluster equals the summation of Mean reduced CD Operational Cost (€) per cluster and the Mean reduced Vehicle Operational Cost per cluster respectively: �(MC)i 8 i=1 = �(MCDOC)𝑖𝑖+ 8 i=1 �(MVOC)𝑖𝑖 (1) 8 i=1

Today, it is widely accepted that the use of carbon-based fuels is one of the major threats to the environment due to the greenhouse effect and this parameter is highly rated from the 3PL company, due the pressure from their pharmaceutical company and has to be considered for the selection an effective supply chain. In Europe, transport accounts for 30% of greenhouse gas emissions (Green House Gas Emissions) and, in particular, road transport is responsible for 70% of emissions generated by transport, as reported by the latest, annual statistical reports of the European Environmental Agency (EPA).

above information selected and evaluated gives the final CO2 in (t/year) emission per Scenario (Table 9). The calculations of CO2 emissions were based on the mileage per week /or per year and the average consumption of the three types of vehicles. (Euro 5).

Table 8: Parameters to calculate CO2 emissions of Scenario 2

PARAMETER SCENARIO 2

Van (200 cases) 10pallets Truck 34pallets Truck

km/week 2491 1214 1050

km/year 129532 63128 54600

mean consumption (Liters/100 km) g CO2/km 1

overall yearly consumption (Liters) 12953,2 10100,5 19110,0

tn CO2/year per vehicle type/net weight 32,4 11,6 7,1

tn CO2/year (overall) 51,1

5. Results

5.1 MCDA results

The final MCDA ranking scores of Tables 12, 13, 14 & 15 (Appendix) are summarized in Table 10. The individual criterion score is presented for each cluster participated. Cities with bold figured score (last column), denote the sites that achieved the higher score therefore are the eligible ones in the final ranking.

Table 9: Ranking of the candidate CD spots on the delivery strategy.

RANKING OF INDIVIDUAL SELECTION CRITERIA _RANKING FINAL Criterion _Real Estate Number of cases per annum Number of customers per cluster Centrality score excl. 3PL DC Centrality score incl. 3PL DC City Corinth 0,0 0,8 1,2 0,7 2,0 4,7 Tripolis 1,0 1,6 1,2 0,6 1,8 6,2 Nafplion 0,0 1,6 1,2 0,5 1,8 5,1 Pyrgos 0,3 0,8 0,4 0,6 1,4 3,5 Arta 0,7 1,6 0,4 0,0 0,2 2,9 Patras 0,3 4,0 1,6 0,9 2,0 8,8 Kalamata 0,3 2,4 0,4 0,2 1,2 4,5 Agrinio 0,3 1,6 0,8 0,2 1,2 4,1 5.2 Final outcome

The outcome of the examined case study can verify that the resulted optimum CD spot supports the best applied scenario. The research results show that the Scenario 2 has the lowest operations cost. The total calculated cost of Scenarios comprises the total operational cost of a CD and the transportation cost for each dedicated route.

0,00 1,00 2,00 3,00 4,00 5,00 6,00 7,00

PATRA AREA PYRGOS AREA KALAMATA AREA AGRINIO & ARTA AREAS TRIPOLI, NAFLION & CORINTHOS AREAS

Individual operational Cost (euro/case delivered) in different areas

BASIC SCENARIO 1 (no Cross docking, 5 daily routes from Athens) SCENARIO 2 (Cross Docking in Patra, 4 routes from CD, 2 route directly from Athens) SCENARIO 3 (Cross Docking in Tripoli, 2 routes from CD, 4 routes directly from Athens)

Figure 12: Operational Cost in Euros per case delivered in the four predetermined geographical areas of the research for the three Scenarios investigated.

Figure 13: Weekly cost per Scenario

Moreover, the CO2 emissions (tn/year) are presented in Figure 14. As we see, the Scenario 2 has the lowest CO2 emissions.

Basic Scenario 1 employs an optimized selected number of vans and small trucks (small vehicles to meet the delivery demands) and no big trucks whatsoever. Still the fleet cover on weekly basis the greatest kilometric distance to fulfill the delivery targets, (i.e. 7.665Km/week). The emitted CO2 gases are the second highest compared to Scenarios 2 & 3. Scenario 2 achieves lower CO2 emissions in comparison to scenario 1 and 3 by 34% and 40% respectively.

Figure 15depicts the resulted kilometric distances, covered by an optimized selected vehicle’s fleet mixture, which is employed for the Basic Scenario 1 purposes on weekly and by extension on annual basis. 34-pallet trucks are excluded as total unsuitable to carry out the delivery to final recipients.

Figure 15: Distance covered (in km) per vehicle type employed on weekly basis for Basic Scenario 1

Scenario 2 (Figure 16) compared to Scenarios 1 & 3 (Figures 15 & 17) presents the most equal balanced distribution of the three transportation types. Compared to Scenario 1 and 3, all types of vehicles for Scenario 2 travel the fewest (4.755 km /week), consequently, as well as on annual basis and has the lowest CO2 emissions.

5744

1921

0

Basic Scenario 1

van vehicle (200 cases) (1 pallets equals 50 cases) 10-pallet vehicle (small truck)

Figure 16: Distance covered (in km) per vehicle type employed on weekly basis for Scenario 2 Concerning Scenario 3 (Figure 17), compared to Scenario 1 and 2, all types of vehicles travel, overall, the second highest distance per year (7.255 km /year). Nonetheless, emits the highest CO2 quantities (in tonnes) (see Figure 14).

Figure 17: Distance covered (in km) per vehicle type employed on weekly basis for Scenario 3

2491

1214

1050

Scenario 2

van vehicle (200 cases) 10-pallet vehicle (small truck) 34-pallet vehicle (big truck)

5196

1423

636

Scenario 3

In Table 11, the results derived from the examined scenarios in the model are presented. Table 10: Summarized outcome data of the examined scenarios

scenario 1 scenario 2 scenario 3

6. Discussion

The model’ s outcome indicates that the second Scenario i.e., cross dock facility in the city of Patras with four local routes and two more directly from Athens to the eastern Peloponnesus, was the optimal selection, in all aspects, compared to the other two scenarios analyzed. It performs the lowest delivery cost per case, the minimal fuel consumption and thereof the lowest CO2 emissions. Scenario 2, which entails one CD in Patras, undertakes a huge order list to serve, within a very strict time schedule. It is evident to expect this, considering that the city of Patras, including its outskirts, is heavily populated, which means that the local delivery routing net should be very dense. That fact underlines the importance of a CD establishment to support the final delivery. In general, local deliveries favour the use of small transportation vehicles. In a nutshell, the number of orders, corresponding to cases’ number and the routing distances to be covered, regulate and determine the CD operation.

The research was staged by taking into account various types of customers involved in the investigation. The patients at home (as a part of the clientele) are the most time-consuming customers (see Table 8), thus it’s certainly expected to influence the truck routing decision-making and the overall driving hours needed for the delivery to be finished. For example, bigger trucks are unsuitable to deliver parcels to individuals. This is an undesirable situation that automatically predetermines the type of the transportation vehicle that is going to be deployed.

6.1 Limitations

The main limitation of the research is that it cannot be applied for other transportation means (e.g. ships, trains, airplanes etc.), unless significant structural changes of the model take place. All vehicles are subjected to the assumption that during their one-way routes they are fully loaded and, consequently, the delivery schedules are performing on the optimum level. The common practice on daily routine dictates that 100% loading capacity of the trucks can rarely be achieved under normal circumstances.

The basic parameters examined in our model were both the tonnage of the pharmaceutical freights and the covered delivery distances. The given results were finally very close to real facts. Nonetheless, apart from road transportation means, it is not applicable to other e.g. vessels, railways, aircrafts etc. Furthermore, the whole model structure was based upon 8 regional clusters. Any alteration, regarding new cluster addition, incurs model modifications to meet the new scenario demands, which is a decision-making requirement.

The proposed model was set up to evaluate maximum 8 different CD candidate locations and calculate CO2 emissions upon the delivery operation. Assumptions were taken regarding the transferred load. Thus, vans/small/big trucks are considered to be, when optimally selected, fully loaded. The model can handle less than eight (8) clusters without any interventions, but for more than 8 CD candidates, serious alterations are needed.

The model can calculate the CO2 emissions for maximum three different vehicle types which are operating complying with any of the Euro 4, Euro 5 and Euro 6 standards. The calculation of CO2 emissions for more than three types of vehicles can be implicated with minor model changes.

6.2 Future Research

It is a common practice, not only in Greece, that trucks of 3PL companies do not return to their base (DC) unloaded. The companies usually come to synergies and undertake goods transportation from the provincial clusters towards Attica. Thus, the overall distances that the vehicles cover (back and forth) are profitable. The ultimate aim of the model structured is to afford a tool for examining such a potentiality, with prior built-in minor modifications and ultimately further reduction of case delivery cost.

For future research, it will also be interesting to examine the issue of further improvement of the model structure, by taking into account the loading percentage, the real time consuming for each route (dynamic time constraints approach), vehicles GPS data incorporation and automated parameters re-estimation. Moreover, the use of more than three vehicle categories with higher loading variety, the special characteristics of the goods delivered, such as the power supply to sustain the commodity within certain temperature limits, could be the object of further examination as well.

7. Conclusions

The detailed literature review, along with the collaboration with a 3PL company, operating in pharmaceutical sector, gave the proper boost in order to accumulate immediate valuable data and, consequently, to evaluate and exploit them in a very constructive and efficient way. Estimations were made regarding the average transportation cost.

The scope of the Thesis was to investigate the potential ability for a pharmaceutical distribution supply chain network to be reengineered and improved. That was carried out on behalf of an operating 3PL company established in Attica region, in Greece. This 3PL has a certain clients’ list and covers the delivery needs in a specific area in southwest Greece. The target was the reduction of the delivery time, and thus the improvement of customer service and the overall operational cost. An analytical built-in model developed ad hoc, which provides a simple tool to estimate the costs and the Co2 emissions of each proposed operational scenario among three. It also introduces and concurrently provides an additional asset on the decision-making for the best scenario to be adopted. The CD of Patras (scenario 2) pointed out the highest score using our MCDA Model and this assumption was confirmed by comparing this scenario with the other two.

The first one (Basic Scenario 1), is the currently operating scenario of the company. The other two alternative Scenarios were developed by supporting the operation of a CD facility. The optimum CD spot selection was accomplished by evaluation criteria, among other candidates’ spots proposed. CDs were spread in suitable sites of clusters i.e. segmented regions that serve as a sub-network of certain number of clients. Clusters correspond to already existed counties, so as to serve in an optimal way for delivery purposes, in Peloponnesus peninsula in southern Greece.

low overall cost difference. Yet, Scenario 2 achieves lower CO2 emissions in comparison to scenario 1 and 3 by 34% and 40% respectively. The operational cost in Scenario 2, compared to basic scenario 1 and scenario 3 could result in a reduction of total annual expenditure by 30%, and 32% likewise. The final selection turns out to be the optimal one by the perspective of total CO2 emissions, therefore has a greener footprint.

The main conclusions extracted from the above case study are summarized as follows: - Patients, as expected, are by far the most time-consuming client category

- Wages cover almost half of the yearly operational expenses of a CD

- Scenario 2, i.e. one CD in Patras, is the less fuel consuming alternative and CO2 emissions - Scenario 2, turned out to be the most profitable alternative on annual operation cost basis - (Tripolis, Nafplio, Corinth) area and (Arta, Agrinio) area, present insignificant changes in

all Scenarios in terms of the overall operational cost per case delivered.

References

1. Boloori Arabani, A., Zandieh, M., Fatemi Ghomi, S. M. T., (2011) “Multi-objective genetic-based algorithms for a cross-docking scheduling problem”, Applied Soft Computing, Vol. 11, pp 1954–4970

2. Boysen, N., Fliedner, M., Scholl, A., (2008) “Scheduling inbound and outbound trucks at cross docking terminals”, OR-Spektrum, Vol. 321, pp 135-161. doi:10.1007/s00291-008-0139-2.

3. Bozer, Y.A., Carlo, H.J., (2008) “Optimizing inbound and outbound door assignments in less-than-truckload crossdocks”, IIE Transactions, Vol. 4011, pp 1007-1018

4. Dobrusky, F.G., (2003) “Optimal location of cross-docking centers for a distribution network in Argentina”, Thesis (M. Eng. in Logistics), Massachusetts Institute of

Technology, Engineering Systems Division

5. Dulebenets, M.A., et al., (2019) “Development of statistical models for improving efficiency of emergency evacuation in areas with vulnerable population”, Reliability

Engineering & System Safety, Vol. 182, pp 233-249

6. (EPA) Official Eurostat website (data 2017), last retrieval on 20/03/2020:

https://ec.europa.eu/eurostat/statistics- explained/index.php?title=File:Modal_split_of_inland_freight_transport,_EU-28,_2012-2017_(%25_share_in_tonne-kilometres).png

7. (EPA) Official Eurostat website (data 2019), “Road freight transport statistics”, available on website, last retrieval on 20/03/2020:

https://ec.europa.eu/eurostat/statistics-explained/index.php/Road_freight_transport_statistics#EU_road_freight_transport_ continues_to_grow

8. Ertek, G., (2005) “A Tutorial on Crossdocking”, 3rd International Logistics & Supply Chain

Congress, Turkey

10. Grigoratos, T., et al., (2019) “Real world emissions performance of heavy-duty Euro VI diesel vehicle”, Atmospheric Environment, Vol. 201, pp 348-359

11. Haial, A., et al., (2016) “A framework for designing a transportation strategy: The case of a pharmaceuticals supply chain”, 3rd International Conference on Logistics Operations

Management (GOL)

12. Hellenic Association of Pharmaceutical Companies, Data monitoring committee, “Facts & Figures, 2014”, retrievable from the official website: https://www.sfee.gr/i-farmakeftiki-agora-stin-ellada-gegonota-ke-stichia-2014/

13. Hellenic Association of Pharmaceutical Companies, “The pharmaceuticals market in Greece, facts and figures 2018”, retrievable from the official website: https://www.sfee.gr/i-farmakeftiki-agora-stin-ellada-gegonota-stichia-2018/

14. Hellenic Statistical Association (2019), data retrievable from the official website: https://www.statistics.gr/el/statistics?p_p_id=documents_WAR_publicationsportlet_I NSTANCE_qDQ8fBKKo4lN&p_p_lifecycle=2&p_p_state=normal&p_p_mode=view&p _p_cacheability=cacheLevelPage&p_p_col_id=column-2&p_p_col_count=4&p_p_col_pos=1&_documents_WAR_publicationsportlet_INSTA NCE_qDQ8fBKKo4lN_javax.faces.resource=document&_documents_WAR_publicatio nsportlet_INSTANCE_qDQ8fBKKo4lN_ln=downloadResources&_documents_WAR_p ublicationsportlet_INSTANCE_qDQ8fBKKo4lN_documentID=393923&_documents_W AR_publicationsportlet_INSTANCE_qDQ8fBKKo4lN_locale=el

15. Izadi, A., Kimiagari, A.M., (2014), “Distribution network design under demand uncertainty using genetic algorithm and Monte Carlo simulation approach: a case study in pharmaceutical industry”, Journal of Industrial Engineering International, 10, Article number: 50, https://doi.org/10.1007/s40092-014-0050-1

16. Knop, K., (2019) “Evaluation of quality of services provided by transport & logistics operator from pharmaceutical industry for improvement purposes”, Transportation

Research Procedia, Vol. 40, pp 1080-1087

18. Kumar, S., Dieveney, E., Dieveney, A., (2009) “Reverse logistic process control measures for the pharmaceutical industry supply chain”, International Journal of Productivity and Performance Management, Vol. 58, pp 188-204

19. Lahmar M. (2007) ‘Facility Logistics: Approaches and solutions to next generation challenges’, Engineering Management Innovation Series, Auerbach Publications

20. Li, Z., He, W., Sim, C., Chen, C., (2012) “A Solution for Cross-docking Operations Planning, Scheduling and Coordination”, Journal of Service Science and Management, Vol. 5(2)

21. Liao, T.W., Egbelu, P.J., Chang, P.C., (2012) “Two hybrid differential evolution algorithms for ‘optimal inbound and outbound truck sequencing in cross docking operations”, Applied Soft Computing, Vol. 12(11), pp 212-229

22. Liao, T.W., Egbelu, P.J., Chang, P.C., (2013), “Simultaneous dock assignment and sequencing of inbound trucks under a fixed outbound truck schedule in multi-door cross docking operations”, International Journal of Production Economics, Vol. 141, pp 212–229

23. Macharis, C., Bernardini, A., (2015), “Reviewing the use of Multi-Criteria Decision Analysis for the evaluation of transport projects: Time for a multi-actor approach”,

Transport Policy, Vol. 37, pp 177-186

24. Martins, S., Amorim, P., Figueira, G., Almada-Lobo, B., (2017) “An optimization-simulation approach to the network redesign problem of pharmaceutical wholesalers”, Computers & Industrial Engineering, Vol. 106, pp 315-328

25. Masaeli, M., Alumur, S.A., Bookbinder, J.H., (2018) “Shipment scheduling in hub location problems”, Transportation Research Part B: Methodological, Elsevier, Vol. 115(C), pp 126-142

26. Mousavi, S.M., Antuchevičienė, J., Zavadskas, E.K., Vahdani, B., Hashemi, H. (2019) “A new decision model for cross-docking center location in logistics networks under interval-valued intuitionistic fuzzy uncertainty”, Transport, Vol. 34(1), pp 30-40

27. Mousazadeh, M., Torabi, S.A., Zahiri, B., (2015) “A robust possibilistic programming approach for pharmaceutical supply chain network design”, Computers & Chemical

28. Narayana, S. A., Elias, A.A., Pati, R.K., (2014) “Reverse Logistics in a Pharmaceutical Supply Chain: A Systemic Analysis”, The International Journal of Logistics Management, Vol. 25, pp 379-398

29. Partovi, Y.F., (2006), “An analytic model for locating facilities strategically”, Omega, Vol. 34(1), pp 41-55

30. Settanni, E., Harrington, T.S., Srai, J.S., (2017) “Pharmaceutical supply chain models: A synthesis from a systems view of operations research”, Operations Research

Perspectives, Vol. 4, pp 74-95

31. Sousa, R.T., Papageorgiou, L.G., Shah, Ν., (2011) “Global supply chain planning for pharmaceuticals”, Chemical Engineering Research and Design, Vol. 89(11), pp. 2396-2409 32. Tolls data website, last retrieval on 20/03/2020: http://diodia.com.gr.

33. Triantaphyllou E. (2000) ‘Multi-Criteria Decision-Making Methods: A Comparative Study’, Applied Optimization

34. Road distances data, last retrieval on 20/03/2020: https://www.viamichelin.com 35. UNILOG S.A. official website: htpp: www.unilog.gr