Identifying the perfect pairs: a simulation study into the impact of partner characteristics on the delivery efficiency of last-mile collaborations

(1)

Identifying the perfect pairs: a simulation study into the impact of partner

characteristics on the delivery efficiency of last-mile collaborations

Master Thesis

MSc. Technology and Operations Management

MSc. Supply Chain Management

University of Groningen

Faculty of Economics and Business

(2)

2

Abstract

With the growth in ecommerce, cities have become exposed to a significant increase in freight transportation. To deal with this, urban stakeholders have launched various initiatives that aim to establish the balance between allowing companies some leeway to profit from this growth, while not letting the negative externalities spiral out of control. However, their outcomes have shown that the initiation of last-mile collaborations is the only way to structurally address the problems caused by the overwhelming amount of commercial vehicles in cities.

Though, the extent to which such collaborations will actually improve the situation is dependent on the specific partners that join them. Therefore, it is surprising that only a limited amount of research has focused on understanding how the characteristics of one company allow for it to form a successful alliance with one type of partner, but not with another. In addition to that, none of those studies evaluated these effects in last-mile settings. Hence, the goal of this thesis is to illustrate how different operational characteristics, that specifically describe last-mile transporters, influence the performance of an alliance. Additionally, it will also conclude which companies can therefore best collaborate together in urban areas.

This will all be done through a combination of a simulation- and case study. The latter will ensure that the separate and joint routes constructed for the different sets of partners are simulated in a context that closely represents a real-life setting. In turn, the difference in the total distance of these routes will show how high the collaborative benefits are for each set of partners.

(3)

3

1. Introduction

The last-mile of a logistics network involves the activities that are needed to supply the residents of a city (Lindawati, van Schagen, Goh & de Souza, 2014). But with the rise of ecommerce, paired with a substantial increase in home deliveries, transporting freight to, within and from those urban areas has become a major challenge (Cleophas, Cottril, Ehmke & Tierney, 2019). It has created a situation in which a city’s logistics infrastructure is exposed to a large number of commercial vehicles, which in turn, negatively impacts the level of emission, traffic congestion, accidents and road deterioration in those urban areas (Ranieri, Digiesi, Silvestri & Roccotelli, 2018). Scholars have further found that these externalities are amplified because logistics service providers each use their own networks of hubs and vehicle fleets to address their last-mile (Muñoz-Villamizar, Montoya-Torres & Faulin, 2017). This means that while many providers deliver within the same areas, their networks only address the spatial distribution of their own customers, and the associated small deliveries per drop (Slabinac, 2015). As a result, overlapping routes prevail, causing streets to be visited by multiple commercial vehicles a day. To counter these inefficiencies, scholars have called upon logistics service providers to collaborate by means of consolidating their freight (Gonzalez-Feliu & Salanova, 2012; Cleophas et al., 2019).

Freight consolidation is defined as “the consolidation of many small shipments into a larger one, to be dispatched on the same vehicle” (Qiu & Huang, 2013). The goal of consolidation is to increase the loading rates of vehicles, by delivering shipments of different providers on the same vehicle routes (Verdonck, Caris, Ramaekers & Janssens, 2013). This will in turn lead to a reduction in the total number of vehicles deployed, and thus, result in less freight movement and empty travel (Nadarajah & Bookbinder, 2013; Savelsbergh & Van Woensel, 2016). But, multiple studies found that highly utilized vehicles in itself do not automatically lead to a reduction in the transportation cost per drop (Qiu & Huang, 2013; Gonzalez-Feliu & Salanova, 2012). They discovered that it is the drop density that arises from bundling freight that determines the economies of scale that can be achieved (Triantafyllou, Cherret & Browne, 2014). However, reality suggests that there are even more factors at play than these studies were able to capture. This is proven by the fact that many consolidation initiatives fail, despite an increase in the total number of stops (Verdonck, Ramaekers, Depaire, Caris & Janssen, 2018). Altogether, this highlights that consolidation should not just be initiated between any two companies (Triantafyllou, Cherret & Browne, 2014). Therefore, there is a growing interest in understanding how the various facets of a company’s last-mile determine whether or not a partnership can actually lead to better results for all stakeholders involved.

(5)

5 distinctive last-mile characteristics are missing, while those that are correctly included are tested with values that do not reflect any actual last-mile operations. Consequently, these papers not only lack the foundation necessary to establish what the benefits are that logistics service providers can obtain in urban collaborations, but more importantly, the ability to recommend which partners can best join an alliance, given the characteristics of their last miles.

Owning to the limitations of prior research, the goal of this thesis is to explore which partnership structures, consisting of companies with different characteristics, will lead to more efficient last-mile distributions, and in turn, to a reduction in the negative externalities faced by urban areas. To allow for this, this thesis will conduct a simulation study, in which the use of consolidation is compared to the practice of individual routes for various alliances. More specifically, this thesis intends to capture the real-life complexities of urban transportation by simulating a logistics backdrop that closely resembles an urban setting in which various companies currently fulfil their last-miles. Consequently, simulating this empirical setting will guarantee that the study will cover exactly those characteristics that have the most influence on the performance of last-mile collaborations, namely the partners’ number of stops, order volumes and service times. In turn, it will allow for accurate company profiles to be build, and thus for sensible last-mile partnerships to be tested.

Together, the results of the simulation will show what the savings will be from consolidating the freight of different partners, and hence, it will help companies tackle the process of partner selection by indicating which companies are each other’s best match. To attain this, the following research question will be answered:

“How do partner characteristics impact the effect that freight consolidation has on the joint distribution of last-mile deliveries in urban areas?”

(6)

6

2. Theoretical Background

2.1 City logistics

City logistics covers all the freight movement in urban areas (Slabinac, 2015). On a strategic level, it is defined by the logistics and transportation systems that are applied within a city (Bektas, Crainic & van Woensel, 2015). Operationally, it comprises the terminal- and transportation activities that deal with the distribution of urban goods and their last-mile delivery to customers (Slabinac, 2015; Cardenas et al., 2017). Most of the literature on city logistics focuses on initiatives that reduce the costs of these operational activities. Taniguchi, Thompson, Yamada & van Duin (2001) defined their goal as “totally optimizing the logistics and transportation activities by private companies with the support of advanced information systems in urban areas, while considering the traffic environment, congestion, safety and energy savings within the framework of a market economy”. This definition shows that the main aim of these initiatives is to reduce the nuisances caused by freight transportation, while not penalizing companies for addressing the growth in deliveries (Morana, Gonzalez-Feliu & Semet, 2014). However, many of the more recent initiatives introduced in cities, such as the usage of eco-friendly vehicles or the implementation of access regulations, are only partial solutions to the problems faced in urban areas (Estrada Romeu, Campos Cacheda & Robusté Antón, 2018). The deeper issue that needs to be addressed is that the characteristics of the supply chain are changing, through e.g. just-in-time and the pressure to deliver within small time windows (Allen et al., 2018). These complexities have led to a substantial increase in the number of commercial vehicles driving within the inner-cities (Gonzalez-Feliu & Salanova, 2012). As a result, Estrada Romeu et al. (2018) found that it will only be possible to radically change these distribution effects when initiatives are launched that will allow for structural reductions in the current routes. They suggest that this can be primarily achieved by promoting collaborations among logistics stakeholders, as it will lead to integrated city logistics systems in which the transportation activities are streamlined (Crainic, Ricciardi & Storchi, 2009). Such collaborative strategies are however less explored in urban contexts, especially in areas in which the city’s infrastructure is unequipped to deal with the increase in transport flows (Muñoz-Villamizar, Montoya-Torres & Vega-Mejía, 2015). Following this observation, this thesis intends to make its contributions to the literature stream that focuses on the collaborative last-mile.

2.2 Collaboration in city logistics

(7)

7 subcontractor and the shipper the client (Wang, Kopfer & Gendreau, 2014). Lastly, lateral collaboration involves integrating multiple shippers and carriers into the same logistics network for maximum flexibility, to achieve the most improved logistics performance (Simatupang & Sridharan, 2002; Mason, Lalwani & Boughton, 2007). But Danloup et al. (2015) added to this that the level at which the collaboration occurs, together with its depth (from operational to strategic) and width (from supporting to core activities) determine the entanglement required to achieve joint economic, social and environmental benefits. Pomponi, Fratocchi, Tafuri & Palumbo (2013) continue with that by stating that strategic partnerships can only be established after companies have already engaged in collaborations at the operational level. They found that these prior interactions are required to establish the trust needed to share broader and more strategic activities. Therefore, Mason et al. (2007) concluded that last-mile collaborations should start out being solely horizontal or vertical. The main motivation for companies to integrate their previously separate networks is to eliminate redundancies in their last-mile activities (Rabe, Kleuter, Clausen & Poeting, 2016). Collaborations namely enable for multiple logistics service providers to operate as one enterprise, which will allow them to achieve economies of scale and better utilize their assets (Adenso-Díaz, Lozano, Garcia-Carabjal & Smith-Miles, 2014; Serrano-Hernández, Juan, Faulin & Perez-Bernabeu, 2017). Additional outcomes are increased sales, improved productivity, service level and market position (Cruijssen et al., 2007; Pomponi, et al., 2013). Bektas, Crainic & van Woensel (2015) found that these benefits arise because collaborations provides partners with the opportunity to co-design a logistics network that can overcome the efficiency constraints of their individual ones. The companies will each only integrate those elements of their logistics infrastructures that will optimize their joint performance. Thus, sharing should enable companies to plan their transport activities more efficiently and environmentally friendly (Verdonck et al., 2018). The drawbacks of this practice are that it heightens the partners’ dependency on one another, increases the operational complexity of their last-miles and leads to a loss of control (Moutaokil, Derrouiche & Neubert, 2012; Lindawati et al., 2014). For instance, a company might be afraid to lose its social identity when it considers its own vehicle fleet as a tool for brand recognition (Holguín-Veras & Sánchez-Díaz, 2016). Nonetheless, Savelsbergh & Van Woensel (2016) found that most companies still choose to share their last-mile networks because the space to individually expand in urban area is either too limited or cost prohibitive.

2.3 Consolidation

(8)

8 Carrier-led consolidation occurs on the basis that the partners share or reallocate orders. One of the most prevalent techniques for sharing orders is joint route planning (Verdonck, Caris, Ramaekers & Janssens, 2013). It allows for the delivery requests of the collaborating partners to be collected and combined in a central pool, after which they generate routes that are the most profitable for the system as a whole (Verdonck et al., 2013; Vanovermeire, Cuervo & Sörensen, 2013). But to allow for centralized planning, and thus to reduce scheduling to a single optimization problem, full information disclosure is required (Gansterer & Hartl, 2018). Without this, the benefits obtained from collaborating will otherwise be suboptimal (Simatupang & Sridharan, 2002). This was confirmed by multiple studies, who found that with centralized route planning, the total distance travelled can be reduced by up to 30% (Cuervo, Vanovermeire & Sörensen, 2016; Gansterer & Hartl, 2018). Besides, Estrada & Roca-Riu (2017) discovered that additional benefits of carrier-led consolidation include an increase in load factors, and a reduction in the number of commercial vehicles within urban areas, which in turn leads to a reduction in emission levels, while still allowing for the service levels to be maintained (Muñoz-Villamizar, Montoya-Torres & Faulin, 2017).

But most of the above studies treated collaboration as an agreement between partners that is solely based on the sharing of orders. This kind of cooperation assumes that, after the orders are exchanged, each partner will still use its own depot and vehicle fleet to complete the deliveries (Verdonck et al., 2013). Although this is a viable way of collaborating, it will keep many of the last-mile redundancies alive. Consequently, scholars have found that collaborations in which the partners’ networks are more intertwined outperform simple order sharing initiatives, especially when a consolidation center is used (Zhou, Van Hui & Lang, 2011; Vornhusen, Wang & Kopfer, 2014). Therefore, this thesis will consider consolidation both at the strategic level of terminal consolidation and operationally through freight pooling for less-than-truckload operations (Ballot & Fontane, 2010).

2.4 Influence of partner characteristics

All these consolidation studies hint that the underlying determinant of a collaboration’s effectiveness is the commonality between the partners’ operations. For example, the research of Triantafyllou, Cherret & Browne (2014) showed that consolidating the freight of 92 carriers led to smaller reductions in the total number of trips than for a different set of 13 carriers. The underlying, but unspoken, explanation for this phenomenon is that the deliveries of the latter were more geographically focuses, which allowed for each departing vehicle to be more highly utilized (Janjevic et al., 2016). In contrast, the stream of literature that explicitly addresses the issue of partner compatibility is scarce. Moreover, most of them have approached it from a managerial perspective. Those papers found that partners should mutually strengthen each other on a strategic, operational and cultural level (Breedman, Krols & Verstrepen, 2015), which others specified meant that their strategic visions, managerial practices and collaborative intentions have to be similar or complementary (Ryu, So & Koo, 2009; Verdonck et al., 2018).

(9)

(10)

10

3. Methodology

The purpose of this study is to discover how the operational characteristics of collaborating partners impact the performance of their alliance. To do so, this research will utilize a case study in combination with a simulation approach. The case study will provide insights into the real-life complexities of collaborative urban transportation (Karlsson, 2009). In turn, the simulation will be used to compare a multitude of partnership structures, as it is easy to change or control the experimental factors (i.e. characteristics) of a simulation model (Robinson, 2015). As a result, this research will use the inputs of a real case to simulate multiple collaborative logistics scenarios, which will therefore all be close approximations of an empirical last-mile setting (Law, 2008).

3.1 Case study

The case study involved two companies, that each fulfil last-mile deliveries in Groningen. The first company is a supermarket chain, that delivers its groceries to the homes of its customers. The second one is a parcel delivery company, where this thesis will focus on its traditional parcel- and food division. These settings were deemed appropriate for the study of last-mile collaborations because of two reasons. First, the city of Groningen is pursuing various avenues to efficiently organize different last-mile flows, and in this process, the two companies have taken on a leading role by piloting a form of vertical urban collaboration. However, their collaborative agreement has now reached the point at which the companies need to decide whether consolidation is something they want to pursue together. Second, based on the assessment of the companies’ last-miles, it was found that their operational characteristics are representative of two very prevalent urban groups; transporters with many stops, but small volumes, and logistics service providers with fewer stops, but with more services at each stop. So, the collected data from these cases will ensure that the simulation’s variables reflect the different issues that companies face in city logistics when they initiate an actual collaboration.

3.2 Experimental design

Based on the in-depth analysis of the case companies’ last-miles, three characteristics were identified as being the operational factors that exercise the most influence on the benefits that a partnerships can achieve in urban areas (Table 1).

Table 1: Simulation inputs

(11)

11 these partnerships will be limited to two in this thesis, to prevent the analysis of the variables from exploding.

Concerning the individual instances, carriers can first be characterized based on the number of stops they have in an urban area; i.e. the amount of locations at which they deliver freight. In Table 1, small-sized carriers are represented by 30 to 60 stops, while 120-150 and 210-270 stops identify medium to large-sized companies respectively. The second distinction is based on the volume that carriers drop off at each location. It represent the number of units delivered per stop. The last defining characteristic is customer service time. It represents the time it takes to deliver freight at each location; i.e. the difference between the arrival and departure time at those location. In this simulation, each additional action, such as signing for receipt, will increase the time spent per customer. With regards to the latter two, this study will assume that the delivery requests of a single carrier all have the same order size, and are delivered with the same service time. No variation will be applied to the characteristics of a single partner, to prevent that (unexplained) randomness within the partners’ individual last miles will skew the joint impact of the partners’ characteristics.

3.3 Vehicle routing problem

To define the benefits of last-mile collaborations, the efficiency that companies can achieve with the use of consolidation needs to be compared to the practice of separate routes. The method used to construct these sets of routes in this thesis is a Vehicle Routing Problem with Time Windows (VRPTW). This particular VRP is fitting because it accounts for ‘time’ in ways that are highly relevant in last-mile contexts; e.g. the delivery timeslots promised to customers or the service times at each location. In general, previous research has used VRPs to generate optimal sets of routes. In this thesis, the focus will be on distance minimization, paired with a reduction in the number of commercial vehicles. The basic properties of a VRP are that each customer should be visited exactly once, by precisely one vehicle, and that all vehicles should start and finish their routes at the depot (Laporte, 1992). In addition to this, this thesis will also consider capacity- and total time related constraints; the total volume delivered by each vehicle cannot surpass its capacity, and their total operating time is not allowed to violate a general time window, in-between which all deliveries have to be completed. However, this thesis will not assume that customers have the option to choose a more specific delivery window within the general one. This will still provide companies with the flexibility to schedule their routes, without being restricted by any constraints external to the collaboration. This assumption is appropriate because it reflects the manner in which the case study companies establish their routes; their operational plans only account for the fact that all of their freight has to be delivered within a 3 hour timespan. The mathematical model that represent this VRPTW is presented in the papers of Cordeau et al. (2002) and El-Sherbeny (2010).

(12)

12 Furthermore, the simulation will assume Euclidean distances between each node pair, that can be traversed at the constant speed of 20 km per hour. Lastly, it will be assumed that each carrier has a adequately large fleet of homogeneous vehicles to deliver its volume.

3.4 Performance evaluation

The VRPTW will simulate the shortest routes, given the vehicles’ capacities and their total operating times. Its objective is to first produce the independent set of routes, after which it will be executed again for the partners’ joined operations. The latter step will generate those routes that are the most profitable for the system as a whole. Therefore, the optimum of the collaboration will differ from the sum of the companies’ separate optima. This difference is a reflection of the savings, or additional costs that the collaboration will generate.

This thesis will compute this difference based on the relative distance (%) (1) and absolute distance savings calculations in km (2) introduced by Cruijssen et al. (2007) and Cuervo et al. (2016), where 𝐷() represents the total distance of the joint routes of Partner A and B: 𝐷(𝐴𝐵), or their separate ones: 𝐷(𝐴) and 𝐷(𝐵).

𝐷(𝐴𝐵)−(𝐷(𝐴)+(𝐷(𝐵))

𝐷(𝐴)+𝐷(𝐵) × 100% (1)

𝐷(𝐴𝐵) − 𝐷(𝐴) − 𝐷(𝐵) (2)

It was chosen to include both measures of distance because it will allow for this thesis to present the most comprehensive set of results. For example, the relative measure does not capture that in absolute distance, the 20% savings of one collaboration can still be higher than the 30% achieved by another; in the case where the former had longer routes to begin with. Oppositely, absolute savings do not convey that a 150km saving in distance can still lead to a higher vehicle reduction than one with 200km; when the former previously dealt with very dispersed customers. Incorporating both measures in different parts of the analyses will therefore allow for the bigger picture to emerge on the most suitable partnership combinations.

(13)

13 Table 2: Pseudo-code Tabu Search

(14)

14

4. Results

The following chapter presents the results of the experiments conducted to test the various partnership combinations. The general findings are introduced first, after which the underlying interaction effects between the characteristics are discussed.

The goal of the simulation was to determine how the use of consolidation compares to the practice of split routes for different alliances. In the end, all the partnership structures tested, resulted in positive savings in regards to the routes’ total distance, ranging from a reduction of 5% up to 30%. This shows that it makes sense for companies to collaborate. But also, that the difference between alliances belonging to the bottom or top half of the savings spectrum is substantial. This can be attributed to the level of synergies that arise in the different partnerships, or the lack thereof. Based on this, the following two suggestions summarize the findings with regards to partner selection:

1. Similarity between the partners’ characteristics is simply a requirement when companies are already constructing (semi-) optimal routes by themselves. When their routes are already sufficiently efficient, the best partner to seek out is the one who deranges the other company’s activities the least; meaning that the partner’s own operational requirements should be as non-imposing on the scheduling of routes as possible (i.e. very low order volume / service time).

2. Complementarity is key when companies faced operational inefficiencies prior to initiating the consolidation. Inefficiencies refer to those factors that constrain the construction of routes, such as the interplay between (high) service times and (small) time windows. According to this suggestion, the most preferred partner is the one whose characteristics are less constraining, or at least not in the same way as those of the other company. This will namely allow for the partners to take advantage of each other’s unused vehicle capacity or operating time.

Noticeably, the mechanism that drives both suggestions is the same; the performance of an alliance is the direct result of the freedom provided by the partners’ operational characteristics to schedule the joint routes efficiently. Ultimately, the savings all depend on whether the companies’ characteristics allow for them to capture the benefits of the increase in drop density.

4.1. Equal number of stops (StopA = StopB)

(15)

15 or 2 units. Additionally, the simulation also confirmed that complementarity leads to the vehicles’ being filled by 40% more freight, and that this generated one of the highest reductions in commercial vehicles out of all possible collaborations between equal partners (Appendix A1). Additionally, Figure 3b shows that similar statements are true for equal partners with varying service times, except that these partnerships structures will be most constrained by the time windows, instead of vehicle capacity.

Figure 1-2: VolA x VolB

Absolute Distance savings for StopA=StopB=30(L) or 120(R)

Figure 3: Absolute distance savings for StopA=StopB=270

a: VolA x VolB b: ServA x ServB

Next to the absolute savings, it was found that synergy wise, alliances formed between companies with small order sizes and short service times generate to the greatest savings (example shown in Figure 4). These partnerships namely have the highest probability that all customers, that are located in close proximity of each other, can be serviced by the same vehicle, and thus that the distance in-between stops is minimized. This is because at that point, the partners’ characteristics impose the least amount of restrictions on the routes. It provides the partners with the flexibility to schedule their joint routes efficiently, which in turn, will allow them to take full advantage of the increase in economies of scale. Oppositely, the smallest savings are achieved when the partners’ operations are very constrained. This is especially true when the partners have high service times, as the operating time is the most restrictive factor in this last-mile setting. To illustrate, Figure 5 confirms that the savings are the lowest every time a partnership is formed between companies where at least one of the partners’ operational plans is dominated by its service time (cf. its volume).

0 20 40 60 80 100 120 1 2 5 10 30 A b so lu te D is ta n ce S av in gs (k m ) VolB

VolA = 1 VolA = 2 VolA = 5 VolA = 10 VolA = 30

0 20 40 60 80 100 120 1 2 5 10 30 A bs ol ut e D is ta nc e Sa vi ng s (k m ) VolB

0 20 40 60 80 100 120 1 2 5 10 30 A b so lu te D is ta n ce S av in gs (k m ) VolB

0 20 40 60 80 100 120 1 2 3 5 10 15 A bs ol ut e D is ta nc e Sa vi ng s (k m ) ServB

(16)

16 Figure 4: Relative distance savings for StopA=StopB=120

a: VolA x VolB b: ServA x ServB

Figure 5: VolA x VolB and ServA x ServB together Relative distance savings for StopA=StopB=120

4.2 Unequal number of stops (StopA ≠ StopB)

Figure 6a depicts the absolute distance savings for each alliance structure of StopA x StopB. It highlights that the greatest distance reduction is achieved when both partners service a large amount of customers in an urban area. The larger the pool of customers that need to be visited, the greater the potential to construct optimal routes. It namely results in a very high combined drop density, meaning that the average distance between customers will be small; which will allow for the greatest reduction in kilometres to be achieved. For example, the results of the simulation showed that when Partner A has 30, 120 or 270 stops, and the stops of Partner B range from (30, 60, 120, 150, 210, 270), the distance between the customers diminishes [from 1090 to 490m], [690-430m], and [490 to 370m] for each StopB respectively, the lowest thus being recorded for 270-270. The results also show that the difference between the in-between distance is the smallest when the partnership include at least one partner with an already high drop density (A-270).

In turn, Figure 6b shows that the relative savings are the highest when partners with an equal number of stops collaborate together, where the results are the absolute highest when an alliance is formed between two small companies. These companies will experience the greatest improvements in their performance, as an alliance will substantially counter the issue of detours that individually plague them in their ability to create efficient routes. But noticeably, the success of such collaborations quickly diminishes once one of the potential partners has more stops. Separately, a small company would still benefit from a substantial increase in joint stops, but, this increase would not be significant from the larger company’s point of view. Therefore, the relative savings are the lowest for small-large partnership combinations, as these generates no substantially new ways in which routes can be

0% 5% 10% 15% 20% 25% 30% 35% 1 2 5 10 30 R el at iv e D is ta n ce S av in gs VolB

0% 5% 10% 15% 20% 25% 30% 35% 1 2 3 5 10 15 R el at iv e D is ta nc e Sa vi ng s ServB

ServA=1 ServA=2 ServA=3 ServA=5 ServA=10 ServA=15

0% 5% 10% 15% 20% 25% 30% 35% R el at iv e D is ta nc e Sa vi ng s

VolA=1, ServA=1 VolA=1, ServA=5 VolA=5,ServA=1 VolA=5,ServA=5

(17)

17 generated compared to the partners’ individual operations. Figure 6b e.g. confirms that small companies (30/60) are the least attractive partners for large ones (210/270). More specifically, the simulation showed that a collaboration between companies with 270-270 stops leads to additional savings of 6.10% in distance, 2.7% in time and 14.5% in vehicles compared to alliances with 30-270 stops.

Figure 6: StopA x StopB

a: Absolute distance savings b: Relative distance savings

4.3 Interaction effects between unequal partners

In the next two sections, the interaction effects between combinations of the three characteristics are presented. This will be done by means of the relative distance savings, as initial evaluations showed that this method is able to best capture the all-round effect, both in distance, and in time and vehicle utilization, when neither the drop density (StopA/StopB) nor at least one of the other characteristic (Serv and/or Vol) is assumed fixed. In addition, it should be noted that any characteristic that is not specifically defined takes on the value of one.

4.3.1 Interaction effect: Stop x Vol

The interaction effects between StopB x VolB will be tested for partnerships in which at least one of the partners (StopA) is small (30), medium (120) or large (270). Their results illustrate that there are tipping points before and beyond which the order sizes of companies are similar enough so that the factor volume will become less of a deciding factor in the selection of partners. Once that point is reached, the decision of with whom to collaborate should be based on the level of compatibility between the partners’ other characteristics. Second, the simulation showed that companies can apply some discretion, as to not collaborate with the partner with the absolute smallest volume, when both deal with small order sizes per stop. The difference in relative savings is namely negligible when the partners’ order sizes stay below a certain volume level, as each of their drop-off volumes is still small enough to, together, utilize most of the vehicles’ load spaces. The same phenomenon is observed when both partners have high order sizes. However, at that point, it is because the partners’ order sizes have become too large regardless of the freight’s precise size, to still allow for consolidation to substantially impact their performance.

But those statements do not ring true for partnership structures in which only one of the partners’ order sizes is substantial. In that case, the partners’ volumes will have a more distinct effect on the savings. For instance, Figures 7 and 8 illustrate that a previously preferred partner (i.e. a large partner) becomes one of the least desirable ones, once its volume reaches 10 or more. In that case, the

0 20 40 60 80 100 120 30 60 120 150 210 270 A bs ol ut e D is ta nc e Sa vi ng s (k m ) StopB

StopA = 30 StopA = 60 StopA = 120 StopA = 150 StopA = 210 StopA = 270

0% 5% 10% 15% 20% 25% 30% 35% 30 60 120 150 210 270 R e la ti ve D is ta n ce S a vi n g s StopB

(18)

18 inefficiencies related to that partner’s vehicle utilization (difficult to fill) substantially nullify the benefits obtained from the increase in drop density. To confirm, the simulation results showed that while the difference in relative savings between Partner B having 1 or 5 units of volume, for StopB: 30, 120 or 270, is only 1.2%, 1% and 1.35% when the volume of Partner A is 1, it changes to 3%, 8.6 and 8.7% respectively when partner A has to deliver 10 units of volume per stop.

Figure 7-8: StopB x VolB

Relative distance savings for StopA=120(L) or 270(R)

In addition, StopB x VolA summarizes the effects when the characteristics are applied to both companies. Figures 9-11 show the percentage savings for each StopB x VolA. They highlight that, the higher the order volume of Partner A, and thus the more difficult it is to combine its freight, the more imminent the need is to collaborate with a large partner. A greater pool of orders namely implies that there are more possibilities to optimize the routes. So, the constraining factor of one company is efficiently countered when the other partner’s drop density is high. This insight also shows that it only up until a certain order size that equally-sized companies achieve the highest performance (Section 4.1.2). Once that volume threshold is reached , a large partner is the only one that can compensate for the fact that high order sizes require for a large fleet of vehicles to be deployed; larger partners have a high number of stops, and thus their many small orders can fill up the vehicles’ unused spaces. For example, Figure 10 shows that companies with 210 and 270 stops only become more desirable than more equally-sized partners for StopA=120 once its volume reaches 10 or more. Furthermore, the simulation results also indicated that once one of the partners’ volume reached that point, the average number of stops is 30% higher per vehicle in a partnership between 120-270 than 120-120 stops (Appendix A2). 0% 5% 10% 15% 20% 25% 30% 35% 30 60 120 150 210 270 R el at iv e D is ta nc e Sa vi ng s StopB

VolB = 1 VolB = 2 VolB = 5 VolB = 10 VolB = 30

0% 5% 10% 15% 20% 25% 30% 35% 30 60 120 150 210 270 R el at iv e D is ta nc e Sa vi ng s StopB

(19)

19 Figure 9-11: StopB x VolA

Relative distance savings for StopA=30(L),120(R) or 270(B)

4.3.2 Interaction effect: Stop x Serv

Figures 12-14 illustrate the effects of StopB x ServB on the relative savings of an alliance. Similar to the findings in StopB x VolB, there is often little difference between various collaborations when both partners require little time at each stop. But, differently to volume, the service time threshold is already reached once it crosses 3 minutes (cf. 5 volume units). This is because the service time, in relation to the general time window, places the most dominant restriction on the load that can be assigned to each vehicle. This effect is amplified when the partners have to visit many stops, as it increases the total service time excessively compared to the reduction in driving time. In total, the simulation results showed that the filled capacity of a vehicle increases up to 15% more when a small to medium-sized company, instead of a larger one, has the higher service time in an alliance (Appendix A3). 0% 5% 10% 15% 20% 25% 30% 35% 30 60 120 150 210 270 R el at iv e D is ta nc e Sa vi ng s StopB

0% 5% 10% 15% 20% 25% 30% 35% 30 60 120 150 210 270 R el at iv e D is ta n ce S av in gs StopB

(20)

20 Figure 12-14: StopB x ServB

Relative distance savings for StopA=30(L),120(R) or 270(B)

Figure 15 extends these findings by summarizing the interactions between StopB x ServB x ServA. It shows that the relative savings are highest when each partner has a low service time. But beyond that, the outcome of a collaboration is extremely sensitive towards to the partners’ service times. Therefore, if one partner ‘needs’ to have a high service time, it should be the smallest one of the two. These partnership combinations are the most profitable because they allow for the smallest portion of all stops to be constrained. Therefore, an additional observation is that it cannot be assumed that a similar increase in the partners’ service times will be equally as damaging to their joint performance. Instead, the service time of a small partner must substantially increase more before it will have the same constraining effect on the partnership as the service time of the larger company imposes. Observe for instance that a collaboration between Partner A (StopA=120) and one with 30 to 60 stops achieves higher savings when ServB is both 10 or 15 minutes, than when ServA is 10 minutes. Oppositely, notice that ServA being 10 or 15 minutes leads to smaller reductions in savings than ServB=10 when Partner B now has 210 to 270 stops.

Figure 15: StopB x ServA x ServB Relative Distance Savings for StopA=120

(If ServA>1, ServB=1 v.v.)

ServB = 1 ServB = 2 ServB = 3 ServB = 5 ServB = 10 ServB = 15

(21)

21

4.4 Best last-mile partners

In section 4.3, it was uncovered that partner selection becomes more important when at least one of the partners has characteristics that overshadow its transportation activities. As it is very common for companies operating in the last-mile to be ruled by specific aspects of their operations, this thesis extends the previous analyses by uncovering who the most suitable partners are for three common types of last-mile transporters. Based on those insights, this result chapter will be concluded by detailing what the implications of those observations are on initiating consolidation between the case study companies.

#1: Companies with large order sizes

The first type represents companies that visit a small number of customers in the urban area. But the order volume they deliver per location is large. As a result, it requires them some time to drop each freight off. This description fits companies who deliver their own products to customers’ homes. More specifically, it is most representative of companies who offer a type of freight that customers buy many of, such as groceries. The following levels describe these companies: StopA:30 stops, VolA:30 units, and ServA:1 minute per stop, plus 20 seconds for each unit of volume.

The effects of the characteristics on the savings show that, although these companies prefer to collaborate with partners that have low service times, they need to be most attentive to the partners’ volumes, as it has the greatest impact on the distance saved. Figures 16a-b confirm that it is inappropriate to let these companies collaborate with partners whose orders or service times are large. Figures 17a-b add to this by illustrating that while large partners (i.e. many stops) are preferred regardless of the service time, they diminish in their attractiveness cf. smaller players when their volume increases. This is because a higher number of stops already equates to a higher total volume (i.e. 30 or 270 with Vol = 1). When this volume is then increased, the positive impact of the drop density (i.e. large pool of customers) diminishes as the vehicle savings become almost zero. In other words, these companies should find partners that are able to fill their left-over capacities with multiple small orders, cf. a few orders of medium or large size. The reason being that the small size of the partner’s freight will translate itself into vehicles that are packed with volume for many customers. In turn, this implies that the opportunities to rearrange the routes more efficiently are high, leading to greater savings. So, the best partner for this type of company is one that has a low service time and volume per stop, but a high total number of stops.

Figure 16: Relative distance savings

a: StopB x VolB b: StopB x ServB

0% 5% 10% 15% 20% 25% 30% 35% 1 2 5 10 30 R el at iv e D is ta n ce S av in gs VolB

StopB = 30 StopB = 60 StopB = 120 StopB = 150 StopB = 210 StopB = 270

0% 5% 10% 15% 20% 25% 30% 35% 1 2 3 5 10 15 R el at iv e D is ta n ce S av in gs ServB

(22)

22 Figure 17: Absolute distance savings

#2: Companies with high service times

The second set of companies is a reflection of those that deliver a small amount of volume per stop, but which require for the company to spend a considerable time at each location. This situation is representative of companies that deliver a type of freight that requires for some additional handling to be performed, such as unpacking or installation. It can also be a characterization of those companies whose business entails fulfilling a service, for which they need a small amount of material per stop. Examples of the latter include a repair-, service or handyman. The characteristics of these companies are as follows: StopA:60, VolA:1 and ServA:15 minutes to finish each job.

The results of pairing up these companies with different partners show that each can best initiate a collaboration with a partner whose operations fully complement its own. These companies have a small number of stops, which, paired with their small volumes, leave them with much unutilized vehicle space. But this can only be filled by partners whose services time are low, as their own operations already take up most of the vehicles’ total operating times. Figures 18ab-19ab confirm that when either VolB or ServB is assumed to be one, the highest savings are achieved when VolB is 30 and ServB is 1. However, the interaction effects do highlight that once these companies choose a partner whose volume is that large, the alliance’s performance becomes very sensitive towards changes in service time, especially with regards to total distance savings. For example, it was observed in Figures 20a-b that after the service time threshold is reached (i.e. 3 min; section 4.2.2), a unit increase in service time results in a much larger decrease in savings when VolB is 30 instead of 5. The reason is that at that point, the order pool of the partners will not be big enough anymore to counter the restrictions put on the routes by the interaction between their service times and volumes. Therefore, this group of companies should look for partners with medium to large-sized orders, where the percentage distance savings are the highest with the former, and absolute savings with the latter.

0 10 20 30 40 50 60 1 2 5 10 30 A bs ol ut e D is ta nc e Sa vi ng s (k m ) VolB

0 10 20 30 40 50 60 1 2 3 5 10 15 A bs ol ut e D is ta nc e Sa vi ng s (k m ) ServB

(23)

23 Figure 18: Relative distance savings

Figure 19: Absolute distance savings

Figure 20: Absolute distance savings StopB x ServB

a: VolB=5 b: VolB=30

#3: Logistics service providers with many stops

The last group represents traditional logistics service providers. The core activity of these companies is to deliver the freight that people ordered online from others vendors to their homes. They are characterized by a large number of stops, at which they are able to drop off small amounts relatively quickly. In an alliance, this type of provider is reflected by: StopA:120, VolA:1 and ServA:2 minutes. But this thesis will also consider what these companies’ most preferred partners will be when they grow to have 210 stops. The analysis for this group of companies will focus solely on the relative distance savings. This choice was made to still be able to present a clear overview of the results for both drop densities, and is appropriate because the results of the absolute distance savings did lead to the same conclusions. 0% 5% 10% 15% 20% 25% 30% 35% 1 2 5 10 30 R e la ti ve D is ta n ce S a vi n g s VolB

0% 5% 10% 15% 20% 25% 30% 35% 1 2 3 5 10 15 R el at iv e D is ta nc e Sa vi ng s ServB

0 20 40 60 80 100 1 2 5 10 30 A b so lu te D is ta n ce S a vi n g s (k m ) VolB

0 20 40 60 80 100 1 2 3 5 10 15 A b so lu te D is ta n ce S a vi n g s (k m ) ServB

0 20 40 60 80 100 1 2 3 5 10 15 A bs ol ut e D is ta nc e Sa vi ng s (k m ) ServB

0 20 40 60 80 100 1 2 3 5 10 15 A b so lu te D is ta n ce S a vi n g s (k m ) ServB

(24)

24 The interaction effects indicate that if a logistics service provider wants to enter a long-term strategic alliance, it can best join a collaboration with a partner whose operations are similar to what its own will be in the future; they should opt for a partner with a high number of stops, paired with the lowest possible service time. This will allow them to enjoy the greatest possible savings at each ‘StopA’ along the way; as their delivery efficiency becomes increasingly more similar to that of their ‘efficient’ partner. This is shown in Figures 21a-b, which illustrate that while the combined similarity-complementarity effect between StopA=120 and StopB=120 (with Serv=5) will allow for one of the highest relative savings to be achieved, partnering with a partner with 120 stops will neither be similar nor complementary enough when StopA=210. Oppositely, when StopB is 210 or 270 with ServB=1, the relative savings remain high as the logistics service provider grows. Oppositely, the results show that the preferred number of stops is a decreasing function of volume. Figures 22a-b show that it is only beneficial to choose a partner with a high volume when it has a limited number of stops. Therefore, it can be concluded that a medium (StopA = 120) or large-sized (StopA = 210) logistics service provider should either collaborate with a small company, with a high volume and low service time, or a large company with a small volume and service time (with the latter option results in the largest absolute savings in distance).

Figure 21: Relative distance savings StopB x ServB for VolB=1

a: StopA=120 b: StopA=210

Figure 22: Relative distance savings StopB x ServB for VolB=30

a: StopA=120 b: StopA=210 0% 5% 10% 15% 20% 25% 30% 35% 1 2 3 5 10 15 R e la ti ve D is ta n ce S a vi n g s ServB

0% 5% 10% 15% 20% 25% 30% 35% 1 2 3 5 10 15 R e la ti ve D is ta n ce S a vi n g s ServB

(25)

25

4.5 Implications for case study

The previous insights allude to the fact that some companies are a suitable partner for a multitude of collaborations, while others only fit with a very specific type of partner. This thesis has shown that this is all down to the flexibility that the company’s operations provide it. When this is applied to the case study, the logistics service provider’s parcel division, which is a prime example of type #3, is thus the least constrained in finding a partner compared to its food division and the supermarket, who both belong to type #1 and are thus not preferred by many. The specific scores of the case’s different collaborative combinations are listed in Tables 4a-c. They show that although, none are a total match based on Section 4.4, the collaborations that include the parcel division unsurprisingly come closest to it. Together with this division, the supermarket or food division will namely be able to spread its freight across vehicles already filled with small parcel packages, instead of having them all gathered within the same vehicles. The benefit of this is that their characteristics will therefore be less of an operational constraint on each individual vehicle, as the additional number of locations that can now be visited will compensate for it. The biggest roadblock for these collaborations is however their service times, as the supermarket and food division’s length will disturb the delivery flow that the parcel division can achieve by itself. But in turn, the complementarity effect that exists with the supermarket’s volume, which will generate a large reduction in vehicles, will partly negate for it by reducing the total distance and thus the vehicles’ travel times.

Oppositely, the profitability of an alliance between the food division and supermarket is damped by their limited number of combined stops, meaning that much of the vehicles operating times are spent driving. In turn, the operational requirements of each of the companies are too comprehensive to compensate for those of the other one. Together, their home deliveries result in much inflexibility. First, both of their freights are delivered in large standardized boxes, which makes them inflexible in filling up each other’s empty vehicle spaces. In addition, their service times and associated time windows, also limit them in their ability to plan routes that include many stops, while the other two collaborative combinations do achieve this. However, the one thing that might promote this collaboration more is when the companies are able to align their supporting processes (i.e. such as the time until which an order can be placed online), as this will allow for a more seamless integration of the operations of, in many ways, very similar companies.

(26)

(27)

27

5. Discussion

Various papers have proposed that the logistics issues companies face in cities can be addressed through last-mile collaborations. They suggest that it will result in economies of scale and better utilized vehicles, and in turn, will minimize the negative externalities faced by cities (Cruijssen et al., 2007; Ballot & Fontane, 2010). But whether a collaboration can actually achieve these benefits, is highly dependent on the operational fit between the companies joining the collaboration. So far, only a limited number of studies have examined the impact of the partners’ characteristics on their collaborative performance. Therefore, the results of this thesis will be discussed in light of their previous findings.

A common thread in this research is that the influence that a characteristic can exercise is determined by the context of the collaboration. All last-mile environments include elements that constrain the transportation activities, where some are more self-imposed (e.g. offering small delivery windows) than others (e.g. regulations that limit the size of vehicles). For example, companies that fulfil their last-miles by means of bicycles face the issue of optimizing their routes in light of the bikes’ limited capacity. Thus far, the paper of Cruijssen et al. (2007) was the only one that alluded to the effects of these restrictions on alliances. The authors stated that the partners’ operational flexibility determine whether the companies can render some form of consolidation. This thesis adds to this, that, on the partner’s individual level, companies with the most constrained routes will obtain the greatest benefits from collaborating, except when both partners are similarly constrained in the subsequent collaboration. This implies that in the general sense, selecting a partner is about understanding how each of the companies is constrained by the collaborative context. For example, in this study’s setting, ‘time’ was the most restrictive, which caused the partners’ service times to be the most deciding factor in whether or not companies should collaborate together.

A second common theme is that the characteristics of the partners have to match to obtain the highest possible benefits from collaborating. A subsequent insight of this thesis is that partners can match in different ways. On the one hand, it can reflect partners whose characteristics are the same. Contrary, companies are also a match when their characteristics complement each other, as none of their characteristics will then strengthen the other’s negative impact on their performance. In this regard, this thesis identified three important ways in which partners can be a (mis)match.

5.1.1 Volume

(28)

28 needed to deliver it all. Oppositely, this thesis highlights that while, individually, companies with small order sizes find it difficult to fill up their vehicles without having to make many detours, together, they can easily fill up their small vehicle fleets with orders belonging to many different customers.

Furthermore, Cuervo et al. (2016) already hinted to the benefits of initiating collaborations between partners with high and low order volumes. But those authors defined a ‘large’ order volume to be so large that it filled up more than half a vehicle. Contrary, this thesis tested it with more realistic values for the last-mile, which still resulted in the complementarity gains to arise. Therefore, this thesis supports the findings of Cuervo et al. (2016), although a last-mile partnership with only high order volumes did not lead to zero savings being achieved anymore. However, similar to the research of Verdonck et al. (2018), this thesis still maintains that even then, the savings are the lowest out of all possible combinations when two companies with large order sizes pair up.

In practical terms, these two insights can be translated into the recommendation that companies should either match up with a partner whose order volume is little, or complementary to that of the company. In addition to this, this thesis is the first one to emphasize that complementarity is most beneficial when partners with small numbers of stops become partners. In turn, similarity in terms of both partners having small volumes, leads to the most fruitful outcome when partners with many stops are linked together.

5.1.2 Number of stops

(29)

29

5.1.3 Volume x Number of stops

The third implication integrates the previous two. This thesis showed that while the selection of a partner in relation to its volume is negatively impacted by the order size of the other (i.e. complementarity effect), it is positively influenced by the other partner’s number of stops (Vanovermeire et al., 2013). It was found that a large number of stops can compensate for the other partner’s inefficiencies, unless that partner faces a lot of operational problems itself. Similar to Vanovermeire et al. (2013), this thesis established that small to medium-sized companies, with high order volumes, can best pair up with a large partner. This is the most appropriate match because it will generate the largest possible pool of combined orders for them. Consequently, such a partner will increase the multitude of ways in which the orders can be spread over the vehicles; it increases the probability that different parts of the urban area include both high and low volume stops. But oppositely to the findings of Vanovermeire et al. (2013), this thesis concludes that this no longer holds when both partners are large; it cancels out the positive relationship between volumes-stops. Any constraining factor, such as a partner’s high volume, is amplified when it applies to a large company. At that point, the stops (and related volume) of a large, ‘efficient’ partner are not ‘large’ enough anymore to fill up all the vehicle space left unused in the large fleet that is needed to deliver the freight of the large, ‘inefficient’ company.

In practical terms, these insights lead to the recommendation that companies can best opt to collaborate in areas in which their stops are isolated (i.e. with a low amount of stops). But while a large partner is preferred in those kinds of collaborations, a large one becomes more of a requirement when the collaboration is extended to dense areas (i.e. many stops). This is because the large(r) company’s delivery efficiency will then at least match that of the other. Except, when that partner is sufficiently constrained itself, because at that point, any complementarity effect will only arise with small to medium-sized partners.

5.2 Limitations and Future Research

This thesis is the first to study the impact of the characteristics that specifically identify last-mile transporters, while also testing previously found interaction effects in the urban area. But two limitations have to be mentioned in relation to this study’s recommendations. First, the experiments were solved using a Tabu Search metaheuristic. Using this method was appropriate because it allowed for decent solutions to be constructed in a short amount of time, which was needed to solve the extensive set of instances (32400x3). But this has two implications with regards to the results. On the one hand, the distance savings achieved based on the Tabu Search’s routes are an overestimation of those that an exact algorithm can produce, because the former will not always return the absolute shortest routes. Oppositely, compared to techniques used by transport planners, a Tabu Search does already generate substantially better optimized routes, meaning that its savings are an underestimation of what actual last-mile collaborations can achieve.

(30)

30 match these circumstances will experience slightly lower savings. As a result, this thesis’s first suggestion for future research is to test the impact of the characteristics in a more realistic city backdrop. They should be evaluated in a setting in which the urban context is a reflection of the different areas that make up a city, meaning that different drop densities, delivery times and vehicle speeds should be considered per area.

(31)

31

6. Conclusion

(32)

32

7. References

Adenso-Díaz, B., Lozano, S., Garcia-Carbajal, S., & Smith-Miles, K. (2014). Assessing partnership savings in horizontal cooperation by planning linked deliveries. Transportation Research Part A: Policy and Practice, 66, 268-279.

Adenso-Díaz, B., Lozano, S., & Moreno, P. (2014). Analysis of the synergies of merging multi-company transportation needs. Transportmetrica A: Transport Science, 10(6), 533-547.

Audy, J. F., D’Amours, S., Lehoux, N., & Rönnqvist, M. (2010). Coordination in collaborative logistics. In International workshop on supply chain models for shared resource management, Brussels.

Ballot, E., & Fontane, F. (2010). Reducing transportation CO2 emissions through pooling of supply networks: perspectives from a case study in French retail chains. Production Planning & Control, 21(6), 640-650.

Barratt, M. (2004). Understanding the meaning of collaboration in the supply chain. Supply Chain Management: an international journal, 9(1), 30-42.

Bektaş, T., Crainic, T. G., & Van Woensel, T. (2017). From Managing Urban Freight to Smart City Logistics Networks. In Network Design and Optimization for Smart Cities. 143-188.

Çetinkaya, S., & Lee, C. Y. (2000). Stock replenishment and shipment scheduling for vendor-managed inventory systems. Management Science, 46(2), 217-232.

Cleophas, C., Cottrill, C., Ehmke, J. F., & Tierney, K. (2019). Collaborative Urban Transportation: Recent Advances in Theory and Practice. European Journal of Operational Research, 273(3), 801-816.

Cordeau, J. F., Desaulniers, G., Desrosiers, J., Solomon, M. M., & Soumis, F. (2002). VRP with Time Windows. In The Vehicle Routing Problem. Society for Industrial and Applied Mathematics, 157-193. Crainic, T. G., Ricciardi, N., & Storchi, G. (2009). Models for evaluating and planning city logistics systems. Transportation science, 43(4), 432-454.

Cruijssen, F., Bräysy, O., Dullaert, W., Fleuren, H., & Salomon, M. (2007). Joint route planning under varying market conditions. International Journal of Physical Distribution & Logistics Management, 37(4), 287-304.

Cruijssen, F., Salomon, M. (2004). Empirical study: order sharing between transportation companies may result in cost reductions between5 to 15 percent. CentER Discussion Paper 2004-80.

(33)

33 Danloup, N., Mirzabeiki, V., Allaoui, H., Goncalves, G., Julien, D., & Mena, C. (2015). Reducing transportation greenhouse gas emissions with collaborative distribution: a case study. Management Research Review, 38(10), 1049-1067.

El-Sherbeny, N. A. (2010). Vehicle routing with time windows: An overview of exact, heuristic and metaheuristic methods. Journal of King Saud University-Science, 22(3), 123-131.

Estrada Romeu, M. Á., Campos Cacheda, J. M., & Robusté Antón, F. (2018). Night deliveries and carrier-led consolidation strategies to improve urban goods distribution. Transport, 33(4), 930-947.

Estrada, M., & Roca-Riu, M. (2017). Stakeholder’s profitability of carrier-led consolidation strategies in urban goods distribution. Transportation Research Part E: Logistics and Transportation Review, 104, 165-188.

Gansterer, M., & Hartl, R. F. (2018). Collaborative vehicle routing: a survey. European Journal of Operational Research, 268(1), 1-12.

Glover, F., Laguna, M. (1997). Tabu Search. Kluwer Academic, Boston.

Gonzalez-Feliu, J., & Salanova, J. M. (2012). Defining and evaluating collaborative urban freight transportation systems. Procedia-Social and Behavioral Sciences, 39, 172-183.

Google Developers. (2019). Vehicle Routing Problem with Time Windows | Google Or Tools. Google Developers. Available at: https://developers.google.com/optimization/routing/vrptw

Guo, F., Liu, Q., Liu, D., Guo, Z. (2017). On Production and Green Transportation Coordination in a Sustainable Global Supply Chain. Sustainability, 9, 2071.

Holguín-Veras, J., & Sánchez-Díaz, I. (2016). Freight demand management and the potential of receiver-led consolidation programs. Transportation Research Part A: Policy and Practice, 84, 109-130. Janjevic, M., Lebeau, P., Ndiaye, A. B., Macharis, C., Van Mierlo, J., & Nsamzinshuti, A. (2016). Strategic scenarios for sustainable urban distribution in the Brussels-capital region using urban consolidation centres. Transportation Research Procedia, 12, 598-612.

Karlsson, 2009. Researching Operations Management. 1 ed. New York: Routledge.

Laporte, G. (1992). The vehicle routing problem: An overview of exact and approximate algorithms. European journal of operational research, 59(3), 345-358.

Law, A. M. (2008). How to build valid and credible simulation models. Proceedings of the 2008 Winter Simulation Conference, 39–47.

(34)

34 Mason, R., Lalwani, C., & Boughton, R. (2007). Combining vertical and horizontal collaboration for transport optimisation. Supply Chain Management: An International Journal, 12(3), 187-199.

Morana, J., Gonzalez-Feliu, J., & Semet, F. (2014). Urban consolidation and logistics pooling. Sustainable urban logistics: Concepts, methods and information systems. 187-210.

Moutaoukil, A., Derrouiche, R., & Neubert, G. (2012, October). Pooling Supply Chain: literature review of collaborative strategies. In Working Conference on Virtual Enterprises, Springer, Berlin. 513-525). Muñoz-Villamizar, A., Montoya-Torres, J. R., & Faulin, J. (2017). Impact of the use of electric vehicles in collaborative urban transport networks: A case study. Transportation Research Part D: Transport and Environment, 50, 40-54.

Muñoz-Villamizar, A., Montoya-Torres, J. R., & Vega-Mejía, C. A. (2015). Non-collaborative versus collaborative last-mile delivery in urban systems with si demands. Procedia CIRP, 30, 263-268.

Nadarajah, S., & Bookbinder, J. H. (2013). Less-than-truckload carrier collaboration problem: modeling framework and solution approach. Journal of heuristics, 19(6), 917-942.

Pomponi, F., L. Fratocchi, S. R. Tafuri, and M. Palumbo. (2013). Horizontal Collaboration in Logistics: A Comprehensive Framework. Logistics & Production, 3 (4): 243–254.

Rabe, M., Klueter, A., Clausen, U., & Poeting, M. (2016, December). An approach for modeling collaborative route planning in supply chain simulation. In Proceedings of the 2016 Winter Simulation Conference, IEEE Press. 2228-2238.

Ranieri, L., Digiesi, S., Silvestri, B., & Roccotelli, M. (2018). A review of last mile logistics innovations in an externalities cost reduction vision. Sustainability, 10(3), 782.

Robinson, S. (2015). A tutorial on conceptual modeling for simulation. In Proceedings of the 2015 Winter Simulation Conference, 1820-1834.

Ryu, I., So, S.H., Koo, C., (2009). The role of partnership in supply chain performance. Ind. Manage. Data Syst. 109 (4), 496–514.

Savelsbergh, M., & Van Woensel, T. (2016). 50th anniversary invited article—city logistics: Challenges and opportunities. Transportation Science, 50(2), 579-590.

Serrano-Hernández, A., Juan, A. A., Faulin, J., & Perez-Bernabeu, E. (2017). Horizontal collaboration in freight transport: concepts, benefits and environmental challenges. SORT-Statistics and Operations Research Transactions, 1(2), 393-414.

Identifying the perfect pairs: a simulation study into the impact of partner characteristics on the delivery efficiency of last-mile collaborations