1
Master thesis in Supply Chain Management
University of Groningen - Faculty of Economics and Business
Role of collaboration in transportation networks
Urban Consolidation Centres - Adding Pickup Services to the Equation
19-07-2019
Joost Hoekstra S2571862 j.hoekstra.24@student.rug.nl Supervisor Dr. I. Bakir Co-assessor Dr. E. Ursavas Word count: 10.0122
ABSTRACT
Purpose: The goal of this thesis is to investigate the improvement of the Urban Consolidation Centre (UCC) effectiveness by offering pickup serivices. While research shows high potential benefits from UCC implementation on all levels of the Triple-bottom-line, many attempts do not survive the start-up phase. A new perspective assuming a receiver focus shows promising reults. An additional offering of pickup services could improve the business model by developing additional cashflow. This thesis investigates what the impact of the additional pickup services will be and how different stakeholder may be affected and make use of the improved logistics system.
Methodology: By means of a simulation experiment this thesis makes use of a deductive research approach. A quantitative model is used to represent the developed scenarios in which the conventional scenarios and proposed scenario are represented. From literature research, some expectational statements are developed to guide answering the research question. In addition, a sensitivity analysis is performed resulting in descriptive and prescriptive results as in how the model will interact in different stakeholder scenarios. Performance of the different scenarios is monitored in terms of total costs of urban last-mile logistics, total distance traveled and vehicle capacity utilization.
Findings: Where the sole use of multimodal transportation lacks to improve logistics systems efficiency, the addition of pickup services to UCC service offering shows high potential returns. Next to monetary improvements, a decrease in total distance traveled of 48.98% is shown. In combination with the use of more sustainable transportation vehicles, this indicates a large benefit for the whole Triple-bottom-line. Managerial insight are derived; (i) the adoption and promotion of pickup services in an early stage of the UCC business development is supported to create a steady cashflow, (ii) the (re-) alignment of fleet composition to the customer demand in a later stage, (iii) the effective offering of subsidies by focussing on operational costs to incentivize the use of E-trikes and, (iv) the potential use of pickup services to make the UCC less sensitive to environmental changes and lower sensitivity. Practical implications: Manny UCC initiatives fail due to lacking the ability to align all stakeholder motivations. By improving insight in UCC performance and possible courses of actions in how to influence performance into a certain direction, this research adds to the holistic view motivating stakeholders to actively participate in the UCC development.
3
TABLE OF CONTENTS
Abstract ... 2
1. Introduction ... 4
2. theoretical background ... 7
2.1 Collaboration in urban last-mile deliveries ... 7
2.2 UCC costs and benefits – multiple stakeholder perceptions ... 8
2.2.1 Receivers ... 8
2.2.2 Carriers ... 9
2.2.3 Local Government & citizens ... 10
2.2.4 UCC operators ... 10 2.3 Short-haul vehicles ... 11 2.4 Customer orders ... 12 3. Problem description ... 13 3.1 Scenario development ... 13 3.2 Expectations ... 14 4. Research methodology ... 15 5. Computational study ... 18 5.1 Scenario comparison ... 18 5.2 Sensitivity analysis ... 19
5.2.1 Setting the parameters ... 19
5.2.2 Analysing the results ... 21
6. Discussion and conclusion ... 28
6.1 Statement reflection ... 28
6.2 Stakeholder interaction with the logistics network performance ... 30
6.3 Limitations and future research ... 32
4
1. INTRODUCTION
The cost of last-mile logistics comprise over 50% of product transportation cost (Dolan, 2018), this has instigated the logistics industry to increase efficiency. Successful execution of last-mile deliveries can assign a competitive edge to a firm, while failure in doing so puts it in a vulnerable position. However, not so long ago it was still recognized as the more expensive, least efficient and most polluting section of the supply chain (Gevaers, Van de Voorde and Vanelslander, 2011). Logistics service providers (LSP), also referred to as carriers, have focused on optimizing the delivery process from their perspective. From a city’s or retailer’s perspective however, the optimal solution is often found in a different direction (Rooijen and Quak, 2010). An effective way of increasing last-mile delivery performance is collaboration within the logistics network (Guajardo and Rönnqvist, 2016). Besides the financial consideration, social and environmental pressure is accumulating to the attractiveness of collaboration in transportation networks (Basso et al., 2019), displaying a potential win-win-win situation to all stages of the triple-bottom-line (Browne, Woodburn and Allen, 2007). The main goals of collaborations are the reduction of costs, uncertainty, CO2 emissions, congestion and noise reduction. While the idea of collaboration in logistics receives growing attention in research (Björklund and Johansson, 2018), the number of successful implementations stays limited (Basso et al., 2019).
6
To what extent do pickup services influence UCCs logistics performance in terms of; total costs of urban last-mile logistics, total travel distance and vehicle capacity utilization and how does this influence the different stakeholders?
By adding pickup services to the logistics network effective use of vehicle capacity can be increased. On the other hand, it results in additional complexity to the routing challenge. As these services are recognized as a valuable addition to the UCC business model (Rooijen and Quak, 2010), more insight into its effects on the logistics network effectiveness is needed. Supporting the pickup services aspect with quantitative measurables adds knowledge to a holistic logistics-management perspective, which has received little attention so far (Browne, Woodburn and Allen, 2007; Verdonck, Caris, Ramaekers and Janssens, 2013). While this new receiver-focused perspective has shown positive results, its ability to fulfilling a city’s logistics needs when operating at full potential while considering the economic, social and ecological needs is still unknown. As a full working system is only possible when all stakeholders’ needs are serviced, deriving insight to the full potential of the system enables and benefits all stakeholders, as indicated by Cruijssen, Cools and Dullaert (2007) inability to quantify the operational savings is a huge barrier to develop horizontal collaboration. Especially practitioners interested in setting-up or running a UCC will benefit by better understanding the impact of the receiver focussed perspective.
The goal of the simulation experiment is to show the impact of adding pickup services to the UCC service offering and making an impact on the business model’s efficiency quantifiable. Optimal vehicle deployment and routing are determined for three different scenarios which are described in chapter 3.1. The problem discussed in this paper can be defined as the vehicle routing problem with simultaneous pickup and delivery with a heterogeneous fleet or HVRPSPD, as described by (Avci and Topaloglu, 2016). Using the insights derived from the simulation study and sensitivity analysis, the relation to practice in terms of stakeholder impact is discussed.
7
2. THEORETICAL BACKGROUND
2.1 Collaboration in urban last-mile deliveries
One of the most cited papers on collaboration in transport (Lambert D. M., Emmelhainz M. A. and Gardner. J. T., 1999) defined a partnership in logistics as: “a tailored business relationship based upon mutual trust, openness, shared risk, and shared rewards that yields a competitive advantage, resulting in business performance greater than would be achieved by the firms individually”. The main goal of collaborative logistics is aimed at improving the logistics chain, by which the total fulfillment costs are lower compared to when the companies proceed on an individual basis. While horizontal collaboration can be used in different stages of the Supply chain, this research will focus on collaborative transportation in last-mile deliveries in urban areas. Horizontal collaboration can be achieved by sharing capacity or orders. When capacity is shared collaborating parties make use of the partners’ transport capacity or fleet, while in the situation of order sharing the participating parties share, combine or exchange customer orders to improve delivery efficiency (Verdonck et al., 2013; Fernández, Roca-Riu and Speranza, 2018). In the case of a UCC we speak of combining customer orders, it enables joint route planning which received much attention in current research (Verdonck et
al., 2013). This is the idea behind Binnstadservice.nl as depicted in figure 2.1.
Figure 2.1, Depicting the logistics efficiency increase by use of a UCC (Source: TNO)
8
2.2 UCC costs and benefits – multiple stakeholder perceptions
There exists a misalignment between the interests of stakeholders in urban last-mile logistics. Most LSPs are still mainly focused on the economic factors related to the logistics system efficiency. However, in urban logistics there exists other stakeholders, e.g. other infrastructure users, local residents, and local government, with more intangible considerations (Muñuzuri, Larrañeta, Onieva and Cortés, 2005). This increases the complexity of developing and implementing a solution to urban logistics challenges satisfying all stakeholders’ needs. As Janjevic, Lebeau, Ndiaye, Macharis, Mierlo and Nsamszinshuti (2016) point out, it is essential to research scenarios which satisfy all stakeholders. In this research we focus on the use of consolidation centres for achieving a higher level of efficiency due to its evident potential from prior research showing cost savings above 25% of total logistics costs (Fernández, Roca-Riu and Speranza, 2018,), let alone the less materialistic savings, and recent developments into the field requesting more academic support as stated on the website of Goederenhubs, a UCC initiative related to Binnenstadservice.nl. We will discuss the perception of different stakeholders towards the implementation of a UCC in the existing logistics network and include their concerns to the feasibility of UCC establishment. This will be structured by focussing on the costs and benefits per stakeholder deriving from the implementation of a UCC.
2.2.1 Receivers
9
2.2.2 Carriers
10
2.2.3 Local Government & citizens
Urban logistics decisions are primarily based upon commercial motives, with little regard for the impact on environmental and social aspects. This results in a wide variety of negative side effects on the environment such as air pollution, greenhouse gas emissions, noise nuisance, congestions and safety issues, visual intrusion, stench and vibration (Browne and Gomez, 2011, Quak and Tavasszy, 2011). As citizens’ interests are represented by the local governments we will focus on their motivations. It is necessary to recognize there exists misalignment between citizens and governments interests and motivations as depicted by the agency theory (Ketchen and Hult, 2007). UCC implementation leads to reduced distance traveled in urban areas. A reduction in logistics distance traveled and a decreased a number of stops decreases congestion levels and lowers safety risks. Combining the decrease in travel distance with the use of low emission vehicles this results in a reduction of greenhouse gas emission and local air pollution as well (Allen et al., 2012). There are reports on urban logistics improvements by the implementation of a UCC by means of vehicle load utilization up to 100% and vehicle trips and travel distances have shown decreases of up to 80%, likewise for the greenhouse gas emissions. Kerbside usage shows decreases by 7%, in Tenjin, up to 42% in Monaco (Allen et al., 2012). Governmental interference in urban logistics operations can be done by means of urban area vehicle restrictions (Allen et al., 2012) or supporting urban logistics initiatives like UCCs by means of subsidies which are both seen as a Key success indicator (Rooijen and Quak, 2010). However, the potential resistance to the urban area vehicle restrictions from carriers, and in some cases explicit as for example in the case of the Italian city Vicenza described by Ville, Gonzalez-Feliu and Dablanc (2012), decreases its attractiveness as it increases the stakeholder misalignment. Furthermore, financial short and long-term viability is within reach as the receiver focused perspective enables this.
2.2.4 UCC operators
11
2.3 Short-haul vehicles
Besides the consolidation of freight, the use of alternative modes of transport is one of the most evident actions to achieve sustainable city logistics. By decoupling the long-haul distance transportation from the urban logistics by means of a UCC, the logistics network is enabled to make use of sustainable modes of transportation within the city centres (Clausen et al., 2016). The BSS fleet entails new electric-assisted cargo bikes, or E-cargo trikes and natural gas-powered vans, E-vans from now on (Rooijen and Quak, 2010). For this research, we will consider the use of E-trikes and E-vans, as the use of E-vans is exponentially growing (CBS, 2019). These modes of transport have low emission of greenhouse gasses. Until shortly the downsides for the use of E-trikes entailed limited capacity and reduced service span. However, new E-trikes, have a service span of 45 km while carrying 250 kg of load and a volume of 40ft3. The E-van capacity has a volume capacity of 140ft3 (Simoni, Bujanovic,
Boyles and Kutanoglu, 2018). By using the vehicles in parallel, also referred to as multimodal transportation, the most efficient mode of transport will be used depending on the customer orders, resorting in effective use of the capacity. A more complete overview of the vehicle attributes can be found in table 2.1.
Table 2.1, Transport vehicle attributes, derived from Simoni et al. (2018) and Lia et al. (2014)
Diesel van Electric van E-Cargo bike
Speed (km/hour) 25 25 25* (16-32) Capacity (ft3) 180 140 40 Service range (km) - 160 45 Purchasing costs (€) 22.000 26.400 6.160 Operating costs (€/km) 1.21 1.21 1.43 Emission costs (€/km) 0.03575 0.00143 0.00022
The table suggests the load capacity of a traditional diesel van overgrows the E-trike with a factor 4.5 which is supported by the research of Lia et al. (2014). However, driving speed is countered by Lia et
al. (2014), who suggest E-trikes actually outperform van travel speed in the narrow urban areas,
12 back the investment. The operational costs lack accounting time for the stops, unloading and loading the cargo. This is compensated by a penalty based on the average hourly wage (De Groot et al., 2017) in the logistics branch, which accounts for € 8,31, or € 6,23 for delivery only, per customer location.
2.4 Customer orders
By including pickup orders to the service offering of BSS the characteristics of customer orders are dispersed. In relation to the routing challenge, customer orders can be categorized into three sections. In this research, only the first two are included. The first being delivery orders, these include cargo delivered to the UCC which has to be delivered to the customer location. Second, there are pickup orders, this includes orders which have to be picked up at the customer location and transported back to the UCC. A third possibility would be when orders have to be picked up and delivered during the same tour, this kind of orders are not included in this research to constrain model complexity.
13
3. PROBLEM DESCRIPTION
The addition of pickup services to the service offering of BSS is of great interest to researchers as well as to practitioners (Rooijen and Quak, 2010). By increasing efficiently usage of the vehicle’s capacity the transportation network can be utilized to a higher level. These logistics improvements are supplemented by the application of multi-modal short-haul vehicles.
In order to retrieve insight on the size of the impact by including pickup services to the UCC offering a simulation experiment will be done. In order to analyze the performance differences of the suggested changes, multiple scenarios will be developed, tested and compared on their performance. The performance of the different scenarios will be measured on traveled distance, capacity utilization, and monetary performance.
3.1 Scenario development
In order to get insight into the potential operational performance of a UCC including pickup services, three scenarios are developed. The first scenario represents a logistics network without the use of a UCC. There are two LSPs in place, one performing delivery services and one performing pickup services. The second scenario describes a situation including last-mile delivery by use of a UCC, however, in this scenario, the pickup services are still performed by a different LSP not using a UCC and its multimodal vehicles. In the third and final scenario Pickup services are added to the logistics network including the UCC combining the delivery and pickup services. A summary of the scenarios can be found in table 3.1. Table 3.1, Representing the three described scenarios
Scenario 1 Scenario 2 Scenario 3
Description Conventional deliveries by carriers
White label last-mile service for delivery, independent conventional pickup service
White label last-mile delivery including pickup and delivery LSPs active in
last-mile logistics
Two, one for deliveries one for pickups
Two, one UCC for deliveries and one conventional LSP for pickups
14
3.2 Expectations
From the literature research, some expectations for the performance of the model are made. These are represented by the development of a number of statements.
Statement 1: Scenario 2 will outperform scenario 1 in terms of average total costs of logistics.
The use of multimodal transportation will decrease the logistics costs of the deliveries. Thereby the total costs of logistics of scenario 2 will be lower compared to scenario 1.
Statement 2: Scenario 2 will not outperform scenario 1 in terms of average total travel distance.
The tour start-up cost is lower for E-trikes while the operational costs per km are higher. This entails that a larger total distance will be accepted to some degree as this is not the primary objective. The total distance traveled will suffer as using multiple E-trikes need to travel further for a lower total cost.
Statement 3: Scenario 2 will outperform scenario 1 in terms of average vehicle capacity utilization.
The short-haul vehicles have a smaller vehicle capacity and the utilization of this capacity will be better aligned to the delivery need. Therefore, the availability of multimodal transportation will make sure the vehicle capacity will be utilized to a higher level.
Statement 4: Scenario 3 will outperform scenario 1 in terms of average total costs of logistics.
The use of consolidation of delivery and pick-up freight will improve the efficiency of the logistics network. The improved effective use of vehicle capacity in addition to the use of multimodal transportation will save on average total logistics costs.
Statement 5: Scenario 3 will outperform scenario 1 in terms of average total distance traveled.
Using a single vehicle for deliveries and pickups largely eliminates empty hauling of vehicles. This decreases the total distance traveled by vehicles to a large extent. However, as mentioned before, the usage of E-trikes may increases the total distance traveled as long as this is beneficial to the total costs of logistics.
Statement 6: Scenario 3 will outperform scenario 1 in terms of average vehicle capacity utilization.
15
4. RESEARCH METHODOLOGY
For the purpose of this study, a mathematical model is developed in order to monitor and compare the performance of the different scenarios. In order to get a holistic perspective the performance is measured by monitoring the number and types of vehicles used, the total distance traveled, number of stops during the tours and the utilization of vehicle capacity. From these results, the total costs of logistics will be derived. A customer location represents one stop which may include multiple requests from multiple nearby receivers.
Some assumptions have been made concerning the model. The first assumption has to do with the location of the UCC. Research suggests that UCC effectiveness is highest when the UCC is located just outside the city border and close to the highway (Quak and Tavasszy, 2011). Thus, when a UCC is optimally placed it can be assumed assume the additional distance traveled by delivering the goods via the UCC instead of going straight to the receivers as marginal. The UCC location decision is left out of the model for simplicity, but this assumption should be taken into account before deriving any conclusions as in practice the UCC might not always be placed in an optimal location and any resulting additional costs should be taken into account when considering the system's effectiveness. This also explains why the potential UCC location will be taken as a start and ending point for the carriers in all scenarios. A second assumption is the UCC’s capacity. While vehicle capacity becomes of increasing interest after adding pickup points to the problem at hand, we do not include UCC capacity as it falls outside the scope of this research. The UCC capacity is assumed to be sufficient at all times or infinite. This assumption with regard to capacity also goes for the fleet size and the number of available vehicles. Third, we leave the supply of the UCC by carriers, necessary to fulfill the receivers’ demand out of scope during this research. We assume the supply of the UCC is always done in a timely fashion, neither do we take receiver time windows into account in the simulation. In practice however, it could be argued that any delay due to reasons out of control of the UCC operators, and in control of the supplying carriers are not supposed to lead to any penalties for the UCC operators. But with the growing focus on supply chain interdependencies adding this in future research would add to the holistic understanding of UCC performance.
16 The problem is formulated as a Mixed Integer Programming (MIP) problem, in which the decision variables are restricted to be integer or binary values. This way the optimal solution can be computed for small instances of data, other models were available (Avci and Topaloglu, 2016) but these are commonly applied for heuristic solution useful for larger data but not guarantee optimal solutions. Optimal solutions are more reliable for comparing different scenarios.
Sets
K; set of all vehicles
N; set of all customer nodes V; set of all nodes with V = {0} u N A; set of all arcs, with A = {(i,j) ∈ N2 : i≠j}
Parameters
cij; cost for traveling from node i to node j using vehicle k
di; delivery demand of customer i
pi; pickup demand of customer i
Qk; vehicle capacity
M; a large positive number
Decision variable
xijk; 1 if node j is visited after node i by vehicle k, 0 otherwise
Additional variables
ui; variable used to prohibit subtours
lsk; load of vehicle k when leaving the depot
17 The above-explained variables, parameters, and sets result in the following mathematical representation of scenario 3. The other scenarios are simplified versions of the mathematical model. Objective function Minimize:
∑
𝑖,𝑗∈𝑉∑
𝑘∈𝐾𝑉𝑐(𝑖𝑗𝑘) ∗ 𝑥(𝑖𝑗𝑘) + ∑
𝑗∈𝑁∑
𝑘∈𝐾𝐹𝑐(𝑘) ∗ 𝑥(0𝑗𝑘)
(1)
Subject to:∑
𝑖∈𝑉∑
𝑘∈𝐾𝑥(𝑖𝑗𝑘)
= 1
∀𝑗 ∈ 𝑁
(2)
∑
𝑖∈𝑉𝑋(𝑖𝑗𝑘) − ∑
𝑖∈𝑉𝑋(𝑗𝑖𝑘)
= 0
∀𝑗 ∈ 𝑉, ∀𝑘 ∈ 𝐾
(3)
∑
𝑗∈𝑁𝑥(0𝑗𝑘) ≤ 1
∀𝑘 ∈ 𝐾
(4)
𝑢(𝑖) − 𝑢(𝑗) + |𝑁| ∗ ∑
𝑘∈𝐾𝑋(𝑖𝑗𝑘)
≤ |𝑁| − 1 ∀𝑖, 𝑗 ∈ 𝑁
(5)
1 ≤ 𝑢(𝑖) ≤ |𝑁|
∀𝑖 ∈ 𝑁
(6)
𝑙𝑠(𝑘) = ∑
𝑖∈𝑉∑
𝑗∈𝑁𝑑(𝑗) ∗ 𝑥(𝑖𝑗𝑘)
∀𝑘 ∈ 𝐾
(7)
𝑙(𝑖) ≥ 𝑙𝑠(𝑘) − 𝑑(𝑖) + 𝑝(𝑖) − 𝑀(1 − 𝑋(0𝑖𝑘))
∀𝑖 ∈ 𝑁, ∀𝑘 ∈ 𝐾
(8)
𝑙(𝑗) ≥ 𝑙(𝑖) − 𝑑(𝑗) + 𝑝(𝑗) − 𝑀(1 − 𝑋(𝑖𝑗𝑘))
∀𝑖, 𝑗 ∈ 𝑁, ∀𝑘 ∈ 𝐾
(9)
0 ≤ 𝑙𝑠(𝑘) ≤ 𝑄(𝑘)
∀𝑘 ∈ 𝐾
(10)
0 ≤ 𝑙(𝑗𝑘) ≤ 𝑄(𝑘)
∀𝑗 ∈ 𝑁
(11)
𝑥(𝑖𝑗𝑘) ∈ {0,1}
∀𝑗 ∈ 𝑁
(12)
18
5. COMPUTATIONAL STUDY
In this chapter, the simulation results based on the mathematical models are discussed. In addition, the sensitivity analysis is performed. The developed model presented in the previous chapter is implemented in Python 3.7 using a Microsoft device using Windows 10. Gurobi Optimizer 8.1.1 software has been used to solve the quantitative model to optimality. Next, to the mathematical model presented in the previous chapter, two simplified models are solved representing scenario 1 & 2. All models are solved simultaneously, using the same device, software and generated data. The HVRPSPD is also known as an NP hard problem, which is only solvable to optimality for small instances (Avci and Topaloglu, 2016). Preliminary test-runs are performed on the benchmark and parameters selected for the sensitivity analysis. Hereby the representativeness of the considered parameters and related benchmark values are checked. Afterward, it is decided to add a cost for stopping at a customer node in order to improve cost balance of the fixed-route costs and operational cost. This improves the representativeness of the model.
5.1 Scenario comparison
To start, the logistics network performance for the different scenarios are computed. Forty instances of data are generated and the averages are compared. With the generated data for the benchmark situation, an average 223,55 ft2 of goods is transported from and to the receivers. In the first scenario,
two LSPs operating diesel-vans are used, one LSP to perform the delivery services and one LSP to perform the pickup services. In the second scenario half of the time, a single E-van is preferable to do the delivery services and the other half of the time three E-trikes are used. For the pickup services, a single diesel sufficed every time. For scenario 3, 62.5% of the time a single E-van is preferred to service all customer orders, with 35% of data instances multiple E-trikes are used and 2.5% of the situations a combination of an E-van and a single E-trike is the preferred deployment of vehicles. The benchmark performance results are presented in Table 5.1.
Table 5.1, Comparing scenario benchmark performances
Scenario 1 Scenario 2 Scenario 3
Average logistics cost per day € 117.34 € 119.94 € 75.37
Average total distance traveled 6558 m 7922 m 4402 m
Average Capacity utilization per day 28.79% 38.15% 80.55%
19 The results in the table show scenario 2 does not outperform scenario 1 on all KPIs. The average total costs of logistics increases slightly with 2.22%. The average total distance traveled increases by 20.80%. The average capacity utilization results in an increase of 9.36%. Next, scenario 1 and scenario 3 are compared. Here we see an improvement for all three KPIs. First the average total costs of logistics, scenario 3 results in a decrease of 35.77%. Second, the average total distance traveled is decreased by 48.98%. Finally, capacity utilization is increased by 51.76%. Figure 5.1 shows an example of scenario 3 where three E-trikes are deployed to perform all delivery and pickup services.
Figure 5.1, Example figure including 3 E-trikes servicing all customer delivery and pickup demand indicated by the three distinctly colored routes.
5.2 Sensitivity analysis
For every parameter being analyzed 40 instances of data are generated per step. Every data point in the plots represents the average results of these 40 data instances. The resulting average costs of logistics, average travel distance, and capacity Utilization are plotted and analyzed. In addition, the amount and type of transportation vehicles are recorded.
5.2.1 Setting the parameters
20 of the order size, the relative vehicle capacities, the fixed cost of transport (expressed by the purchase price of the E-Vehicles), the operational cost of transport and the size of the distribution area. For all parameters, a benchmark scenario value is developed by means of desk-research and the preliminary test runs. The sensitivity of every parameter is tested by varying a single parameter value while fixing the other parameters to their benchmark scenario value. For an overview of all parameter benchmark values settings and steps used in the sensitivity analysis see Table 5.2.
Table 5.2, Parameter values for the sensitivity analysis
Parameter Minimum value Maximum value Step size Benchmark value Number of
customer-nodes
4 8 1 6
Average size order request
15 25 2,5 20
Standard deviation order request size
2 8 1,5 5
Relative vehicle capacities (Etrike -E-van – Diesel van)
1 – 2,5 – 3 1 – 4,5 – 6 0 – 0,5 – 0,75 1 – 3,5 – 4,5 Purchasing price E-Vehicles 0% 100% 25% 100% Operating costs E-Vehicles 50% 100% 12,5% 100%
Pickup request ratio 60% 100% 10% 90%
Distribution area size (city centre)
91 ha (Nijmegen) 171 ha (Groningen) 805 ha (Amsterdam)
130 ha
21 combining of multiple customers in a single customer location supports the use of a low standard deviation for the order request size. The vehicle capacities, fixed-route costs (based upon the purchasing costs) and operational costs are derived from Simoni et al. (2018). The pickup request ratio is tested in the preliminary tests, a level of 90% showed to deliver optimal results. To show the full potential of adding pickup services to the service offering the benchmark value is set to this optimal level of 90%. The sensitivity analysis includes this parameter to depict what happens when this level is not achieved or even when a 100% level is achieved. Finally, distribution area sizes representing different Dutch city centres (CBS, 2017) are considered to increase the generalizability of the results for different city sizes.
5.2.2 Analysing the results
The results are analyzed and discussed in the following section, in addition, some plots are added for a graphical representation of the scenarios sensitivity to parameter changes.
Number of nodes
When a UCC focuses on the receiving end of the services it is interesting to know how the business performs in the start-up phase and what happens when a larger customer base is achieved. As shown in figure 5.2, the sensitivity analysis returns the lowest coefficient regarding total costs of logistics for scenario 3 suggesting this scenario will keep outperforming the other scenarios during the start-up phase as well as after a growing number of customer locations.
0 50 100 150 200 4 5 (B) 6 7 8 Av era ge cos ts o f lo gis tics ( €)
Number of customer locations
Total cost S.1 Total cost S.2 Total cost S.3
22 Figure 5.3 shows total travel distance to reach an optimum for scenario 2 & 3 when seven customer locations are included. Scenario 1 shows steady growth. This optimum is a result from an ideal situation for the use of E-vans as in this situation the usage of E-trikes is relatively very low at 5% compared to 35% and 100% when respectively six or eight nodes are serviced.
Vehicle capacity is most efficiently used when six customer locations are included as can be seen in figure 5.4. At this point a single E-van or three E-Trikes can be used to cover all customer demand, depending on the dispersion of the nodes but mainly the deviation of the order sizes. Beyond this level of customer nodes, a combination of an E-van and an E-trike is often necessary to cover all demand.
0 2000 4000 6000 8000 10000 4 5 (B) 6 7 8 Av era ge d is ta n ce t ra ve le d (m )
Number of customer locations
Total dist S.1 Total dist S.2 Total dist S.3
0 0.2 0.4 0.6 0.8 1 4 5 (B) 6 7 8 Av erage v eh . cap . u tili za tion
Number of customer locations
Utilization S.1 Utilization S.2 Utilization S.3
Figure 3.3, Total travel distance vs. number of customer locations
23 Order-sizes
As the average order-size increases the effectiveness of the UCC decrease in terms of costs. Total logistics costs increase for scenario 2 & 3 while total costs for scenario 1 remains stable. This can be explained by the high capacity of the diesel-van, which is more effective for higher-order sizes. However, the total distance traveled decreases as E-vans become the preferred mode of transport over E-trikes. As shown in figure 5.5, vehicle capacity utilization of the UCC reaches its optimum at an average order size of 20 as here E-vans and E-trikes capacity can be used to optimality. Before this point, customer orders utilize vehicle capacity up to a lower level and afterward combinations of the E-vans and E-trikes become necessary to satisfy the magnitude of all customer demand.
Standard deviations
Next, the sensitivity of the performance is tested against the standard deviation for values ranging from ten up to forty percent of the mean. The total cost of logistics shows minimal sensitivity to the changes. For scenario 2 & 3 the total travel distance is highest at the benchmark value of 5, equal to 25% standard deviation from the mean, With this level of deviation from the mean, the economical preference for E-trikes grows with 15% as compared to the lower deviation level of 3,5 and 20% as compared to a higher standard deviation level of 6,5. It could be expected that a lower deviation would be preferable for the use of E-trikes. With the mean order size at 20 ft2 and E-trike capacity of 40 ft2
fitting two loads in one trike is highly dependable on the negative or compensating deviations for delivery order sizes and pickup orders. The utilization of vehicle capacity reacts negatively to the standard deviation as can be seen in figure 5.6. The usage of E-trikes becomes less attractive as explained above. 0 0.2 0.4 0.6 0.8 1 15 17,5 (B) 20 22,5 25 Av era ge v eh . cap . u tiliza tio n
Average order sizes (ft2)
Utilization S.1 Utilization S.2 Utilization S.3
24 Relative vehicle capacity
The total cost of logistics shows a slight decrease related to the growing relative vehicle capacity. When the relative capacity of 1 – 2.5 – 3 increases up to 1 – 4.5 – 6, for E-trikes, E-vans, and diesel-vans respectively, the total costs of logistics for scenario 1 decreases by 7.62%, for scenario 2 a decrease of 4.95% and for scenario 3 a decrease of 6.42% is shown. As the vans become better capable of satisfying all customer demand the usage of E-trikes and combinations become less attractive and total distance traveled decreases. Striking is the increase of E-trike usage in scenario 2 for the relative vehicle capacity of 1 – 4 – 5.25, as shown in figure 5.7, while this does not seem to be the most attractive mode of transportation for scenario 3. A possible explanation for this could be unfavorable pickup order sizes and locations making the customer nodes difficult to combine for E-trike routes including the pickups. Vehicle capacity utilization decreases with the growing relative vehicle capacity sizes for scenario 1 & 2. Even though at the lowest level additional trucks have to be deployed 22.5 % of the time to satisfy all customer delivery requests.
0 0.2 0.4 0.6 0.8 1 2 3,5 (B) 5 6,5 8 Av era ge v eh . cap . u tiliza tio n Standard deviation
Utilization S.1 Utilization S.2 Utilization S.3
Figure 5.6, Vehicle capacity utilization vs. standard deviation
0 2000 4000 6000 8000 10000 1 - 2,5 - 3 1 - 3 - 3,75 (B) 1 - 3,5 - 4,5 1 - 4 - 5,25 1 - 4,5 - 6 Av era ge d is ta n ce t ra ve le d (m )
Relative vehicle capacity
Total dist S.1 Total dist S.2 Total dist S.3
25 Tour start-up costs
As the tour start-up costs are only reduced for the E-vehicles, scenario 1 is expected not to react to the changes during the sensitivity analysis of this parameter. A 100% decrease in tour start-up costs results in a 21.678% and 43.12% decrease of total costs of logistics for scenario 2 and scenario 3 respectively. The total distance traveled is positively related to the tour start-up costs. This can be explained by the fact that lower tour start-up costs decreases the attractiveness of E-trikes as these maintain a higher operational cost. As the tour start-up costs decrease it becomes more attractive to use additional vehicles, which decreases the utilization of vehicle capacity as shown in figure 5.8.
Tour operational costs
As the tour start-up costs are only reduced for the E-vehicles, scenario 1 is expected not to react to the changes during the sensitivity analysis of this parameter. A 50% decrease in tour operational costs results in a 19.43% and 53.82% decrease of total costs of logistics for scenario 2 and scenario 3 respectively. This is the lowest total costs of logistics measured during this simulation experiment. When the tour operational costs are decreased it would be expected that longer traveling distances using fewer vehicles, in order to prevent high tour start-up costs, would be preferred. However, figure 5.9 indicates a low level of total travel distance at 50 % operational cost. Just at 62.5% a peak in travel distance seem to appear. The vehicle capacity utilization stays stable for all scenarios when the tour operational costs are decreased.
0 0.2 0.4 0.6 0.8 1 0% 25% 50% 75% (B) 100% Av era ge v eh . cap . u tiliza tio n
Tour startup costs
Utilization S.1 Utilization S.2 Utilization S.3
26 Pickup ratio
The average costs for an increased pickup ratio by 10 percent in the range under research are on average 4.72%, 4.87% and 2.79% for scenario 1, scenario 2 and scenario 3 respectively. The total cost increases are shown in figure 5.10. The total travel distance shows stable growth for scenario 1, while total travel distance seems to decrease for scenario 2 and scenario 3. Vehicle capacity usage shows a small increase for scenario 1 and scenario 2 and a larger increase for scenario 3.
0 2000 4000 6000 8000 10000 50% 62.5% 75% 87.5% (B) 100% Av era ge d is ta n ce t ra ve le d (m )
Tour operational cost
Total dist S.1 Total dist S.2 Total dist S.3
Figure 5.9, Total traveled distance vs. Tour operational costs
0 20 40 60 80 100 120 140 60% 70% 80% (B) 90% 100% Av era ge cos ts o f lo gis tics ( €) Pickup ratio
Total cost S.1 Total cost S.2 Total cost S.3
27 Distribution area sizes
As the distribution area grows the total costs of logistics increases, but notable differences can be observed between the different scenarios. However, as the area sizes are chosen to represent different Dutch cities the step sizes are unequal and are difficult to compare. Therefore, figure 5.11 indicates the percentual growth levels related to the different distribution area size increases. This indicates scenario 1 costs of logistics is relatively more sensitive to area size increases in the smaller-section, between 91 and 130 ha, and larger-section, 171 and 805 ha, and is relatively insensitive to area size increases in the medium-section. Scenario 2 shows a strong sensitivity in the medium-section, and scenario 3 shows to be less sensitive to area size increases in the larger-section. The total distance traveled shows similar results for all three scenarios. Utilization of vehicle capacity stays stable for scenario 1. Scenario 2 shows a slight increase when moving to the larger-section of area sizes, while scenario 3 shows a small decrease of 3.62% in vehicle capacity utilization from 171 ha to the large size city of 805 ha. 0 0.02 0.04 0.06 0.08 0.1 91 - 130 130 - 171 171 - 805 Perc en tu al cos t o f lo gis tics growth (% )
Distribution area indifferences
Scenario 1 Scenario 2 Scenario 3
28
6. DISCUSSION AND CONCLUSION
In order to answer the research question and related statements developed at the outset of this thesis, the results of the computational studies will be discussed. In addition, some managerial insights resulting from the research will be presented. Finally, limitations to the performed research and future recommendations will be discussed. The research question yields:
To what extent do pickup services influence UCCs logistics performance in terms of; total costs of urban last-mile logistics, total travel distance and vehicle capacity utilization and how does this influence the different stakeholders?
First, the performance of the different scenarios will be examined using the developed statements. Next, the sensitivity analysis will be examined in relation to the different stakeholders. How these stakeholders can be affected by the logistics network and specifically the implementation of a UCC servicing both delivery and pick-up order requests is of main interest. In a more prescriptive note, also will be discussed how the stakeholders can effectively influence the logistics networks and its performance. Lastly, conclusions will be derived from the simulation experiment concerning the scientific and managerial implications.
6.1 Statement reflection
The analysis shows that multimodal transport can improve transportation network performance on total costs of logistics, total distance traveled and capacity usage. However, these KPI performances might not always be aligned and therefore react in a different way to the scenario. The first three statements were concerned with the performance of scenario 2, including the UCC for delivery services while a different LSP services the pickup requests, as compared to scenario 1, being the conventional scenario without the use of an UCC. The latter three statements concern the performance of scenario 3, which describes the proposed scenario including a UCC servicing all delivery and pickup requests, as compared to scenario 1. The expectations and the resulting insights from the simulation experiment will now be discussed.
Statement 1: Scenario 2 will outperform scenario 1 in terms of average total costs of logistics.
29 of average total costs of logistics. Therefore, statement 1 is not supported by the results of the simulation experiment.
Statement 2: Scenario 2 will not outperform scenario 1 in terms of average total travel distance.
The tour start-up cost is lower for E-trikes while the operational costs per km are higher. This entails that a larger total distance will be accepted to some degree because this is not the primary objective. The total distance traveled will suffer as using multiple E-trikes need to travel further for a lower total cost. In half of the data instances, the usage of three E-trikes is preferred in the second scenario, this explains the average increase in distance of 20.80% and support for the second statement.
Statement 3: Scenario 2 will outperform scenario 1 in terms of average vehicle capacity utilization.
As expected the availability of multimodal short-haul transportation vehicles with diverse vehicle capacities were better aligned to the delivery need. Therefore, the availability of multimodal transportation in combination with smaller vehicle capacity sizes improved the utilization of vehicle capacity on average by 9.36%. Therefore, statement 3 is supported by the research results.
Statement 4: Scenario 3 will outperform scenario 1 in terms of average total costs of logistics.
The use of consolidation of delivery and pick-up freight in combination with multimodal transportation has successfully improved the transportation networks efficiency. The average total costs of logistics have been decreased by 35.77%, which is a significant improvement, statement 4 is supported.
Statement 5: Scenario 3 will outperform scenario 1 in terms of average total distance traveled.
As expected, the usage of a single vehicle for deliveries and pickups eliminates empty backhauling to a large extent. This decreases the total distance traveled by vehicles significantly. Even with the increased distances from using multiple vehicles, in 37,5% of the occasions, the average total distance traveled is decreased by 48.98%. This indicates close to half the original amount of transportation vehicles on the road for servicing the included customer orders, statement 5 is supported as well.
Statement 6: Scenario 3 will outperform scenario 1 in terms of average vehicle capacity utilization.
30
6.2 Stakeholder interaction with the logistics network performance
The sensitivity analysis indicates what can be expected in similar scenarios by use of interpolation or even extrapolation for scenarios surpassing the tested parameter levels. Next-to the descriptive insights of what stakeholders can expect when a UCC will take place in their logistics network, the results of the analysis can also function in a prescriptive manner to guide shaping the UCC’s constitution or modification in an attempt to influence performance outcomes.
The first managerial insight is related to the timing of adopting pickup services to the UCC service offering. Some of the parameters included in the sensitivity analysis can be used to derive some insight into the model's performance while developing from a starting phase towards a more mature state of the UCCs business. In the start-up phase, the number of customer locations will be fewer, the order sizes will be smaller and the standard deviations will be higher compared to a more mature state of the business. The sensitivity analysis showed when accounting for a small number of customer locations it is already effective to include pickup services, even so, when the pickup ratio is low. As in the start-up phase developing cashflow is of primary importance (Van Duin, Quak and Muñuzuri, 2010), including and actively developing paid pickup services in order to generate cashflow from an early phase by the UCC-operators can be advised.
31 This brings us to the third managerial insight retrieved from this research, concerned with the effective use of subsidies. In the sensitivity analysis, the tour start-up costs and operational costs are distinguished. The tour start-up costs are based upon earning back the investment of the used transportation vehicle. The operational costs are based upon wages, related stops at customers, insurances, fuel expenses and a small penalty for emissions. Subsidies obviously decrease the total costs of logistics in a rather stable manner. However, the effect on secondary KPIs shows different results. The analysis results show an increase of E-vans at the expense of E-trikes at lower levels of tour start-up costs, the same happens to a far lesser extent when decreasing operational costs. This indicates when E-trike usage is to be promoted instead of larger vehicles, even when fuelled by ecological more responsible power sources. Decreasing operational costs or making clear agreements about the use of vehicle types should be made to achieve the desired result.
A fourth and final managerial insight derived from the performed sensitivity analysis is the decreased sensitivity of the UCC when pickup services are included. When pickup services are included in the service offering of the UCC the performance is less sensitive to changes in customer locations dispersion, or service area size, in terms of total costs of logistics and total distance traveled. Also, the number of customer nodes has a slightly smaller impact on the total costs of logistics. Adding pickup services to the UCC service offering diversifies the business and spreads the costs over a larger amount of customer transactions. UCC operators and carriers can benefit from this method of cost dispersion and spread the risk resulting from their business.
This research has found academic relevance by the development of a new HVRPSPD model and finding numerical results related to the questions: To what extent do pickup services influence UCCs logistics
performance in terms of; total costs of urban last-mile logistic, total travel distance, and vehicle capacity utilization? by answering the developed statements. The more practical managerial insights
are mainly derived from the second part of the research question: How does this influence the different
stakeholders?, leaving practitioners with useful handles as how to anticipate to and act upon the
32
6.3 Limitations and future research
33
REFERENCES
Allen, J., Browne, M., Woodburn., A. and Leonardi, J. (2012) ‘The Role of Urban Consolidation Centres in Sustainable Freight Transport’, Transport Reviews, 32(4), pp. 473–490.
Avci, M. and Topaloglu, S. (2016) ‘A hybrid metaheuristic algorithm for heterogeneous vehicle routing problem with simultaneous pickup and delivery’, Expert Systems With Applications, 53, pp. 160–171. Basso, F., D'Amours, S., Rönnqvist, M. and Weintraub, A (2019) ‘A survey on obstacles and difficulties of practical implementation of horizontal collaboration in logistics’, International Transactions in
Operational Research, 26(3), pp. 775–793.
Björklund, M. and Johansson, H. (2018) ‘Urban consolidation centres: retail stores’ demands for UCC services’, International Journal of Physical Distribution & Logistics Management, 48(8), pp. 745-764. Browne, M. and Gomez, M. (2011) ‘Research in urban logistics: a systematic literature review’,
International Journal of Physical Distribution & Logistics Management, 41(9), pp. 896–912.
Browne, M., Woodburn, A. and Allen, J. (2007) ‘Evaluating the potential for urban consolidation centres’, European Transport, 35, pp. 46–63.
Browne, P. M. et al. (2005) Urban Freight Consolidation Centres Final Report. London.
CBS, (2017) Kerncijfers wijken en buurten 2017. Centraal Bureau voor de Statistiek, Statline Publicaties. CBS, (2019) Bedrijfsvoertuigen; voertuigkenmerken, regio's 1 januari. Centraal Bureau voor de
Statistiek, Statline Publicaties.
Cherrett, T., Allen, J., McLead, F., Maynard, S., Hickford, A. and Browne, M. (2012) ‘Understanding urban freight activity - key issues for freight planning’, Journal of Transport Geography, 24, pp. 22–32. Choi, T. Y. and Krause, D. R. (2006) ‘The supply base and its complexity: implications for transaction costs, risks, responsiveness and innovation’, Journal of Operations Management, 24(5), pp. 637-652. Clausen, U., Geiger, C. and Pöting, M. (2016) ‘Hands-on testing of last mile concepts’, Transportation
Research Procedia. Elsevier B.V., 14, pp. 1533–1542.
34 Cruijssen, F., Cools, M. and Dullaert, W. (2007) ‘Horizontal cooperation in logistics: Opportunities and impediments’, Transportation Research Part E: Logistics and Transportation Review, 43(2), pp. 129– 142.
Dethloff, J. (2001) Vehicle routing and reverse logistics: the vehicle routing problem with simultaneous
delivery and pick-up. Working paper, pp. 79-96.
Dolan, S. (2018) ‘The challenges of last mile logistics & delivery technoogy solutions’, Business insider. Van Duin, J. H. R., Quak, H. and Muñuzuri, J. (2010) ‘New challenges for urban consolidation centres: A case study in The Hague’, Procedia Social and Behavioral Sciences, 2, pp. 6177–6188.
Fernández, E., Roca-Riu, M. and Speranza, M. G. (2018) ‘The Shared Customer Collaboration Vehicle Routing Problem’, European Journal of Operational Research, 265(3), pp. 1078–1093.
Gentile, G. and Vigo, D. (2013) ‘Movement generation and trip distribution for freight demand modelling applied to city logistics’, European Transport, Issue 54.
Gevaers, R., Van de Voorde, E. and Vanelslander, T. (2011) ‘Characteristics and typologies of last-mile logistics from an innovation perspective in an urban context. ’ City Distribution and Urban Freight
Transport: Multiple Perspectives, Edward Elgar Publishing, pp. 56-71.
Gonzalez-Feliu, J. (2012) ‘Freight Distribution Systems with Cross Docking: A Multidisciplinary Analysis Transportation Research Forum Freight Distribution Systems with Cross-Docking: A Multicisciplinary Analysis’, Journal of the Transportation Research Forum, 51(1), pp. 93–109.
De Groot, C., Verweij, K., Vos, G., Hoen, A., Otten, M., Den Breejen, J., Pool, J., Schoo, R., Van den Engel, A., Kindt, M., Riskt, J., Ligterink, N., Sluijk, N., Spreen, J. and Verbeek, M. (2017) ‘Gebruikers en inzet van bestelauto’s in Nederland’, Topsector Logistiek, p. 85.
Guajardo, M. and Rönnqvist, M. (2016) ‘A review on cost allocation methods in collaborative transportation’, International Transactions in Operational Research, 23(3), pp. 371–392.
Huang, F., He, J. and Lei, Q. (2018) ‘Coordination in a retailer-dominated supply chain with a risk-averse manufacturer under marketing dependency’, International Transactions in Operational Research, 00, pp. 1–23.
Janjevic, M., Lebeau, P., Ndiaye, A. B., Macharis, C., Van Mierlo, J., Nsamszinshuti. A. (2016) ‘Strategic scenarios for sustainable urban distribution in the Brussels-capital region using urban consolidation centres’, Transportation Research Procedia, 12, pp. 598–612.
35 Karaoglan, I., Altiparmak, F., Kara, I. and Denzig, B. (no date) ‘Technical Report Vehicle Routing Problem
with Simultaneous Pickup and Delivery: Mixed Integer Programming Formulations and Comparative Analyses. ’
Ketchen, D. J. and Hult, G. T. M. (2007) ‘Bridging organization theory and supply chain management: The case of best value supply chains’, Journal of Operations Management, 25(2), pp. 573–580. Lambert, M. D., Emmelhainz, M. A. and Gardner, J. T. (1999) ‘Building successful logistics partnerships’,
Journal of Business Logistics, 20(1), pp. 165-181.
Lia, F., Luè, A., Nocerino, R., Bresciani, C. and Colorni, A. (2014) ‘Promotion of E-bikes for delivery of
goods in European urban areas: an Italian case study’, Transport Research Arena.
Liu, J., Xiao, T., Tian, C. and Wang, H. (2018) ‘Ordering and returns handling decisions and coordination in a supply chain with demand uncertainty’, International Transactions in Operational Research, 00, pp. 1–25.
Miller, C. E., Tucker, A. W. and Zemlin, R. A. (1960) ‘Integer Programming Formulation of Traveling Salesman Problems’, Journal of the ACM. Association for Computing Machinery, 7(4), pp. 326–329. Muñuzuri, J., Larrañeta, J., Onieva, L. and Cortés, P. (2005) ‘Solutions applicable by local administrations for urban logistics improvement’, Cities, 22(1), pp. 15–28.
Österle, I., Aditjandra, P. T., Vaghi, C., Grea, G. and Zunder, T. H. (2015) ‘The role of a structured stakeholder consultation process within the establishment of a sustainable urban supply chain’, Supply
Chain Management: An International Journal, 20(3), pp. 284–299.
Quak, H. and Tavasszy, L. (2011) ‘Customized solutions for sustainable city logistics: the viability of urban freight consolidation centres’, In: van Nunen, J., Huijbregts, P., Rietveld, P., (eds) Transition Towards Sustainable Mobility. Berlin: Springer.
Rooijen, T. Van and Quak, H. (2010) ‘Local impacts of a new urban consolidation centre – the case of Binnenstadservice.nl’, in The Sixth International Conference on City Logistics, pp. 5967–5979.
Simoni, M. D., Bujanovic, P., Boyles, S. D. and Kutanoglu, E. (2018) ‘Urban consolidation solutions for parcel delivery considering location, fleet and route choice’, Case Studies on Transport Policy, 6, pp. 112–124.
Verdonck, L., Caris, A. N., Ramaekers, K. and Janssens, G. K. (2013) ‘Collaborative Logistics from the Perspective of Road Transportation Companies’, Transport Reviews, 33(6), pp. 700–719.
