A framework for measuring transport efficiency in distribution centers

Transport Policy 45 (2016) 99–106

Contents lists available at ScienceDirect

Transport Policy journal homepage: www.elsevier.com/locate/tranpol

A framework for measuring transport efficiency in distribution centers Milan Andrejić n, Nebojša Bojović, Milorad Kilibarda University of Belgrade, Faculty of Transport and Traffic Engineering, Vojvode Stepe 305, 11000 Belgrade, Serbia

art ic l e i nf o

a b s t r a c t

Article history: Received 30 January 2015 Received in revised form 23 July 2015 Accepted 26 September 2015

Performances of distribution systems are largely affected by the performances of transport systems. This paper is devoted to the analysis of the efficiency of transport subsystems in distribution centers. Transport is a logistics process with the highest energy consumption. In the transport systems two aspects of measuring efficiency are identified. The first aspect is the fleet efficiency which is related to the higher level of decision making. The second aspect of decision making is the vehicle efficiency as operational level of decision making. The main objective of this paper is to propose models for measuring transport efficiency, as well as to identify main factors that affect transport efficiency. The proposed models are based on Principal Component Analysis and Data Envelopment Analysis approaches. According to the results fleet management system, catchment area, vehicle capacity, the age of vehicles and manufacturers are the basic factors that affect transport efficiency. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Efficiency Fleet Vehicle Principal component analysis Data envelopment analysis

1. Introduction Due to a dynamic market and environmental changes, the distribution centers have to realize their activities and processes in an efficient way. During the last three decades logistics has been recognized as one of the most important service industries. Logistics has a crucial role in supply chains, which reflects in connecting members into supply chains. Distribution centers (DCs) are complex logistics systems which connect producers with other participants in the chain, including end-users. Two basic subsystems of DCs are warehouse and transport subsystems. The most important part of transport subsystem refers to fleet and vehicle operating. Efficiency is a very important indicator of companies' operations analysis, and it is one of the basic and the most frequently used performances. Measuring, monitoring and improving efficiency are the main tasks for companies in the 21st century. The importance of efficiency measuring and importance of effectiveness measuring have been recognized in literature (Chow et al., 1994; Adler and Berechman, 2001; Hackman et al., 2001). “Single ratio” indicators have been used for estimating the efficiency of logistics systems for a long time. These indicators do not provide enough information about the system operating. Numerous authors have advocated the use of approaches such as the Data Envelopment Analysis (DEA) method (Min and Joo, 2006). On the n

Corresponding author. Tel.: þ 381 11 30 91 304; fax: þ381 11 30 96 704. E-mail addresses: [email protected] (M. Andrejić), [email protected] (N. Bojović), [email protected] (M. Kilibarda). http://dx.doi.org/10.1016/j.tranpol.2015.09.013 0967-070X/& 2015 Elsevier Ltd. All rights reserved.

other side, Adler and Golany (2001, 2002) have suggested using the Principal Component Analysis (PCA), a methodology that produces uncorrelated linear combinations of original inputs and outputs, to improve discrimination in the DEA with a minimal loss of information. When there are an excessive number of inputs and outputs in relation to the number of decision making units (DMUs), DEA method is inapplicable. The main object of this paper is to propose a model for measuring transport efficiency on different levels, as well as to identify the key factors that affect transport efficiency. To the best of our knowledge there were no papers that simultaneously evaluate the transport efficiency on the described levels and investigate factors that affect transport efficiency. The research is conducted on the DCs that operate in Serbia. Two measurement levels (aspects) are identified in this paper. The first level relates to measuring efficiency of fleets, while the second relates to vehicle efficiency. The information obtained from the company management is used in the process of the model development. The examination of the company management, fleet size, gravity area (catchment area – the area that DC serves), the age of vehicle, vehicle capacity, vehicle manufacturer, influence on the efficiency scores are also objective of this paper. This paper consists of six sections. Section 2 gives a review of the transport efficiency measuring in the literature. Hypotheses are also defined in the next section. Transport efficiency measuring aspects are described in Section 3. The models for evaluating transport and the vehicle efficiency are also proposed in this section. The results of efficiency measurement are given in Section 4. Examination of the factor influence on the efficiency scores and

M. Andrejić et al. / Transport Policy 45 (2016) 99–106

100

hypotheses testing are realized in Section 5. The overall conclusions and future directions are presented in Section 6.

2. Literature review and hypotheses definition Road transport has dominant role in the distribution of goods. Importance of transportation and vehicle fleet management is recognized in the literature (Pedraza-Martinez and Van Wassenhove, 2012). There are different approaches for freight transport performance measurement in the literature (Table 1). Kim (2010) has evaluated technical and scale efficiency of the 62 individual trucks in logistics. The DEA model for efficiency evaluation is specified with three output categories and five costs categories which represent the input. Cruijssen et al. (2010) analyzed efficiency of the freight transportation in Flanders. Simons et al. (2004) defined the Overall Vehicle Effectiveness (OVE). They also stated that transport efficiency is important at an economic, social and environmental level. The authors defined five transport losses or waste: driver breaks, excess loading time, fill loss, speed loss and quality delay. Mckinnon (1999) analyzed KPIs for the food supply chain with special emphasis on the degree of empty running, fuel efficiency, deviations from schedule, time utilization and vehicle utilization. Donselaar et al. (1998) investigated transport and distribution effectiveness. Energy efficiency has become a critical issue for logistics systems. In a situation of increasing global energy demands and rising energy costs, conserving energy is becoming a very important issue. There are many papers that investigate indicators of energy efficiency in transport and logistics in general. Kalenoja et al. (2011) studied indicators of energy efficiency of supply chains, with emphasis on energy consumption, water and electricity consumption, fuel consumption and material use, habitat improvements and damage due to enterprise operations, quantity of non-product output returned to process or market by recycling or reuse. They also analyzed the requirements of ISO 14301 standard (environmental performance evaluation) in the context of energy efficiency. Neto et al. (2009) recognized the problem of balancing

environmental and business concerns. A comprehensive overview of environmental performance metrics for the green supply chain, that range from air emissions to energy recovery and recycling, is given in the paper of Hervani et al. (2005). The main environmental pressures and undesirable outputs in road transport are defined by Kuosmanen and Kortelainen (2005). The authors also emphasized the importance of economic variables such as mileage and fuel consumption. Sarkis and Talluri (2004) analyzed eco-efficiency using qualitative and quantitative inputs and outputs. Chow et al. (1994) examined the definition and measurement of performance in logistics research. The authors defined “distribution effectiveness” as adequacy, consistency, accuracy, timeliness, initiative, responsiveness. Numerous factors affect transport efficiency. These factors are not sufficiently investigated in the literature. The situation is no better in practice. In this paper several factors are categorized according to two efficiency measurement levels. Three factors are identified for each level. Fleet management, fleet size and gravity area are basic factors that affect fleet efficiency. On the operational level vehicle efficiency is influenced by the age of vehicle, vehicle capacity and vehicle manufacturer. In the first hypothesis it is assumed that information system (Fleet Management System-FMS) affects fleet efficiency. In that sense the first hypothesis investigates the difference of the fleet efficiency scores between two analyzed companies (company A and company B) with different fleet management systems. Because of the importance and differences in each fleet management systems we predict differences in the efficiency scores in different companies. The first hypothesis has the following form: H1. : There is a difference in fleet efficiencies of company A and company B caused by the fleet management system. Many past studies analyzed relationship between size of logistics systems and efficiency. (Hackman et al., 2001; Banaszewska et al., 2012). However, the aforementioned papers analyzed warehouses and distribution centers. Because of the previous literature we assume that fleet size affects efficiency scores since of economies of scale. In that manner the next hypothesis is set:

Table 1 Review of transport efficiency measuring. Publication Andrejić et al. (2013)

Indicators

Vehicles, Forklifts, Fuel, Electricity consumption, Invoices (Demands), Warehouse overtime, Time truck utilization, Failures in warehouse, Failures in transport, Write off expired goods Cruijssen et al. (2010) Equipment (e.g. number of trucks, number of trailers, total loading capacity etc.), Labor (e.g. total wages, (drivers’), total hours worked, number of employees, etc.), Added value, Profit, Intangible assets (market information, customer contacts, goodwill etc) Donselaar et al. Costs, Wages, Hours/truck, Hours/driver, Speed, Turnover/trip,(un)loading time/trip, Turnover, (1998) Load factor, Loading capacity, Percentage km driven empty, Turnover/truck, Number of trucks Hervani et al. (2005) Total energy use, fugitive non-point air emissions, total electricity use, total fuel use, other energy use, major environmental, social, and economic impacts associated with the life cycle of products and services Strategic, tactical and operational measurement levels Mckinnon (1999) Vehicle space utilization, Vehicle time utilization, Empty running, Deviations from schedule, Fuel consumption Neto et al. (2009) Masses entering the treatment system, Output masses that are recycled Kalenoja et al. (2011) Quality, Time, Costs, and Flexibility, Environmental indicators such as energy consumption or carbon dioxide emissions, ISO 14031 performances (environmental condition indicators, management performance indicators and operational performance indicators) Kim (2010) Fuel cost, Oil cost, Supplies cost, Labor cost, Tax, Insurance, Transportation distance, Transportation amount Kuosmanen and Kor- Mileage, Fuel consumption, Undesirable outputs (CO2, CH4,N2O, CO, NOx SO2, emissions…) telainen (2005) Sarkis and Talluri Qualitative inputs (managerial plans, Green Purchasing program, ISP 14000…) (2004) Quantitative inputs (raw material intake, energy, materials used, employees…) Qualitative outputs (biodiversity impacts, greenhouse impact, community response…) Quantitative outputs (water emissions, air emissions, solid wastes, products, penalties…) Simons et al. (2004) Labor, Energy consumption, Operating costs, Vehicle emissions, Fuel,, Transport losses or wastes (driver breaks, excess loading time, fill loss, speed loss, quality delay)

Field

Indicator types

Transport and distribution Transport systems and vehicles

Operational, Financial, Quality, Utilization Equipment (Capacity), Operational, Financial, Energy Equipment (Capacity), Operational, Financial Energy, Environmental

Transport and distribution Transport

Transport systems Energy, Utilization and vehicles Logistics network Environmental Transport Financial, Energy, Quality, Environmental Transport systems and vehicles Transport systems and vehicles General organization

Financial, Operational, Energy Operational, Energy, Environmental Environmental, Qualitative, Operational

Transport systems Financial, Operational, and vehicles Energy, Quality

M. Andrejić et al. / Transport Policy 45 (2016) 99–106

H2. : Large fleets are more efficient than small fleets because of economies of scale. Based on the information obtained from the management of company, the third hypothesis is set. Namely, management assumption is that DCs located in large cities have more efficient fleets than DCs located in small cities. Banaszewska et al. (2012) analyzed similar problem in the case of express depots. Due to the mentioned reasons, we predict that the fleets that operates in large cities with wider catchment area are less efficient than the fleets with narrower catchment area in smaller cities. The third hypothesis has the following form:

Table 2 Aspects of measuring transport efficiency.

Decision making level Responsible Indicators

H3. : There is a difference in efficiency scores of fleets that serve large and small cities as a result of different environmental factors.

Factors

Vehicle capacity is one of the basic vehicle characteristics. Capacity of vehicle influences the other vehicle parameters like fuel consumption, but also the shipped tons, etc. Kim (2010) investigated relationship between efficiency score and truck capacity. Because of the previous research in literature and practice we predict the differences in efficiency because the vehicles capacity. In this paper next hypothesis is set:

Numerical example

H4. : There is a difference in efficiency scores between vehicles with different capacities. The age of vehicles largely determines technical characteristics. Based on interviews with fleet managers and dispatchers the assumption about the influence of the age of vehicle on the efficiency score is set. This assumption is generally known, but to the best of authors' knowledge, there are no papers that explicitly examine this dependence. From the mentioned reason in this paper we predict the differences in efficiency because the age of production. The fifth hypothesis has the following form: H5. : There is a difference in efficiency scores between vehicles with different age of production. One of the basic management decisions in the transport systems is the vehicles procurement. There are different opinions about vehicle manufacturers. To the best of authors' knowledge, there are no papers that explicitly examine this dependence. Due to various technical and exploitation characteristics of vehicles from different manufacturers, we predict the differences in the efficiency scores of vehicles of different manufacturers. In that sense next hypothesis is define: H6. : There is a difference in efficiency scores between vehicles of different manufacturers.

3. Transport efficiency measurement aspects – data and model definition In the observed DCs more attention is paid to performance monitoring of the fleets. This is the first level of transport efficiency measuring. On the other side, individual vehicles, as basic elements directly affect fleets efficiency. In that manner, the vehicle efficiency is the second aspect of measuring transport efficiency in this paper (Table 2). Measuring aspects tested on two numerical examples are analyzed in this paper. The first example relates to measuring efficiency of 13 DCs, while the second analyzes efficiency of 170 vehicles. 3.1. Fleet efficiency – tactical level The measuring fleet efficiency is tactical level of decision making (Table 2). This aspect of measuring is very important for

101

Model (approach)

Fleet efficiency

Vehicle efficiency

Tactical

Operational

Transport manager Number of vehicle; Fuel costs (103 m.u.); Total trucks time (h); Distance driven (km); Shipped tons (t); Vehicle utilization (%);

Dispatchers Fuel consumption (l); Number of deliveries (stops);

Transport management; Fleet size; Gravity area; 13 Distribution centers (7 of company A and 6 of company B) PCA–DEA

Distance driven (km); Shipped pallets; Vehicle capacity, The age of vehicles, Manufacturers 170 vehicles (40 standard and 130 refrigerated vehicles) DEA

transport managers. Number of vehicles, fuel costs, the total distance driven and amount of transported tons are some of the basic information that transport managers use in the decision making process. Sometimes it is difficult to make a decision in the presence of numerous independent indicators. In that manner, it is necessary to define new measure that integrates all indicators. Efficiency index is one of the most frequently used measures (Andrejić et al., 2013). Among others, in this paper the efficiency of thirteen fleets of two trading companies which operate in Serbia and have a similar sale network, products range and distribution system are analyzed. Every DC has its own fleet. The trucks in the observed fleets have similar capacity and operating systems. The spatial distribution of distribution centers is shown in Fig. 1. The main problem in the DEA method application is the selection of the input and output variables. This problem is already recognized in the literature (Boussofiane et al., 1991). From the standpoint of the processing approach, transport subsystem in DC is a system which uses a number of inputs (resources) in order to generate certain outputs. In the observed DCs managers monitor six main indicators that describe fleet operations. In this paper they are grouped in three input and three output variables. The input variables include number of vehicles, fuel costs and total vehicle time in operation, while the output variables include total distance driven, tons shipped, and vehicle utilization. The fleet size is described by the number of vehicles (trucks). Fuel costs are one of the most frequently used indicators of energy consumption. Fuel costs are expressed in monetary units (m.u.). Time related indicators are also important in fleet efficiency analysis. Total truck time relates to total time that each vehicle operates. This paper analyzes total vehicle time in operation. On the other side, total distance driven as an output variable, is expressed in kilometers. Tons shipped are a frequently used variable for vehicle efficiency evaluation. The last output variable is vehicle utilization. It is expressed in percentage and represents the ratio of volume of goods and the volume of vehicle cargo space. Input and output variables of the observed fleets are shown in Table 3. Data relates to representative month of 2011. As mentioned before, DEA is one the most frequently used methods in efficiency evaluation. In the DEA literature, there are recommendations about relation between number of DMUs (Decision Making Units) and number of variables (Zhou et al., 2008; Drake and Howcroft, 1994). Considering the fact that it is necessary to determine the efficiency of 13 fleets (DMUs) using six variables, discriminatory power of standard DEA models is questionable. In order to improve discriminatory power of DEA, the

M. Andrejić et al. / Transport Policy 45 (2016) 99–106

102

Fig. 1. Spatial distribution of DCs.

Table 3 Input and output variables for efficiency evaluation. DMU

Number of vehicle

Fuel costs (103m.u.)

Total trucks time (h)

Distance driven (km)

Shipped tons (t)

Vehicle utilization (%)

DMU1 DMU2 DMU3 DMU4 DMU5 DMU6 DMU7 DMU8 DMU9 DMU10 DMU11 DMU12 DMU13 Average St. dev.

29 15 37 21 18 24 12 12 22 16 9 16 13 18.77 7.80

2370.07 590.18 5142.88 3837.26 1098.17 3308.41 1057.71 1951.62 3492.53 2320.57 1001.68 2306.46 1904.92 2337.11 1314.85

6500 2647 10,771 7142 3776 6079 3133 1052 2462 1525 669 1519 939 3708.77 3049.91

114,122 32,795 226,242 159,893 53,641 153,413 61,369 27,248 54,769 38,813 17,989 35,612 33,334 77,633.85 64,955.08

5,994,686 917,034 19,279,019 7,474,618 1,326,588 3,988,841 1,192,797 1369 3423 1721 889 1881 1334 3,091,092.31 5,471,198.10

86.10 87.04 81.17 98.14 98.71 99.82 89.93 94.92 98.14 85.71 90.08 93.48 79.66 90.99 6.80

M. Andrejić et al. / Transport Policy 45 (2016) 99–106

103

Table 4 Vehicles descriptive statistics and efficiency scores. DMU Standard vehicles (40) Average St. dev. Refrigerated vehicles (130) Average St. dev. All vehicles (170) Average St. dev.

Fuel consumption (l)

Number of stops

Distance driven (km)

Shipped pallets

694.62 1004.29

156.33 192.84

2516.78 2953.62

602.13 706.44

961.98 1099.27

169.86 128.97

3839.45 3204.26

418.75 410.69

899.07 1080.77

166.68 145.98

3528.24 3188.57

461.89 500.00

PCA–DEA approach is proposed (Andrejić et al., 2013). The PCA represents data reduction technique of multivariate data. The PCA explains the variance structure of a matrix of data through linear combinations of variables, consequently reducing the data to a few principal components (PCs), which generally describe 80–90% of the variance in the data (Sharma, 1996). If most of the population variance can be attributed to the first few components (dummy variables), then they can replace the original variables with minimum loss of information (Adler and Golany, 2001, 2002). The PCA improves discrimination of the DEA models. According to Hair et al. (1995), a random vector X¼[X1, X2,…, Xp] (the p is the number of original inputs/outputs chosen to be aggregated) has the correlation matrix C with eigenvalues λ1 Z λ2 Z…Z λp Z0 and normalized eigenvectors l1,l2,…,lp. Consider the linear combinations, where the superscript t represents the transpose operator:

XPCi = lit = l1i X1 + l2i X2 + ... + lpi Xp , i = 1, 2, ... , p

(1)

Var (XPCI ) = lit Cli , i = 1, 2, ... , p

(2)

t

Correlation (XPCI , XPCK ) = l i Clk, i = 1, 2, ... , p , k = 1, 2, ... , p , i ≠ k

(3)

The PCs are the uncorrelated linear combinations ranked by their variances in a descending order. Adler and Golany (2002) set additional constraints that require the weight of PC1 to be at least that of PC2, the weight of PC2 to be at least that of PC3 and so on. For the efficiency evaluation we use PCA–DEA software (Adler and Yazhemsky, 2010). The PCA–DEA model for DMUa used in this paper has the following form.:

max Vo, Uo, Upc, Vpc

t a Uot Y oa + U pc Y pc

(4)

Subject to:

V ot X oa

t a + V pc X pc =1

(5)

t t V ot X o + V pc Xpc − Uot X o + U pc Ypc ≥ 0

(6)

Vo ≥ 0

(7)

Uo ≥ 0

(8)

t V pc Lx ≥ 0

(9)

t U pc Ly ≥ 0

(10)

Vpc , Upc , free

(11)

o and pc are indexes of original and principal component variables. Xpc and Ypc represent the input and output matrix, X a t t and Y a are input and output column vectors of DMUa, V pc and U pc

Average age of vehicle (years)

Average capacity (t)

15

11.79

10

10.19

11

10.55

represent vectors of weights assigned to inputs and outputs, while L x and L y relate to the matrix of the PCA linear coefficients of input and output data (Adler and Yazhemsky, 2010). 3.2. Vehicle efficiency – operational level There is insufficient literature regarding the efficiency of a single vehicle. Measuring vehicle efficiency relates to operational level of decision making. Conducted research shows the great importance of the vehicles efficiency for the distribution process. In the observed DCs vehicle operation is monitored by dispatchers. They analyzed four main indicators: fuel consumption, number of deliveries (stops), distance driven, and transport pallets. The age of vehicle, capacity, manufacturer, are identified as key factors that affects efficiency. As in the previous case, it is also difficult to make decision about vehicles in the presence of numerous independent indicators. The integrated measure of the vehicle efficiency is necessary. Data for 170 vehicles and four indicators are collected. All vehicles serve the DCs in Serbia and operate in similar conditions. Each vehicle represents the DMU. Fuel consumption represents input while number of deliveries (stops), distance driven and transport pallets represent outputs (Table 4). These indicators are common indicators for a set of observed vehicles. There were problems in the data availability. For certain vehicles there are additional data, but for successful efficiency measure of all 170 vehicles it is necessary to have the same indicators for all vehicles. In the process of hypotheses testing 130 refrigerated vehicles are observed. The main reason is the homogeneity of the sample, and the lack of certain information relating to the hypotheses testing. It is important to note that some indicators are common for operational and tactical level (Table 2). For example, fuel consumption is important for the fleet, but also for the particular vehicle. An interesting difference is reflected in the unit of measurement. For the fleet operating fuel consumption is expressed in m.u. (cost), while at the operational level, the quantity of fuel that vehicle consumes is more important. Similarly, shipped tons describe the fleet operating, while shipped pallets describe vehicle operating. The pallets as a loading units provide much more information about packaging. This information is very important at the operational level. Since the number of DMUs is significantly higher than the number of variables standard CCR (Charnes, Cooper, Rhodes) DEA model is proposed (Charnes et al., 1978). Similar model is also proposed in Kim (2010). This model has been widely applied in the literature and therefore it is not described in detail in this paper.

4. Case study results The efficiency scores are analyzed in this section. The first part relates to measuring the fleet efficiency according proposed the PCA–DEA model, while the second part relates to vehicle efficiency

M. Andrejić et al. / Transport Policy 45 (2016) 99–106

104

Table 5 Fleet efficiency scores.

Table 6 Vehicle efficiency scores.

DMU (Fleet)

Company Standard CCR DEA model

PCA–DEA modeln

Benchmarks

DMU1

A

0.96

0.85

DMU2 DMU3 DMU4 DMU5

A A A A

1.00 1.00 1.00 0.87

1.00 1.00 1.00 0.85

DMU6

A

1.00

0.88

DMU7 DMU8

A B

1.00 0.92

1.00 0.84

DMU9

B

0.74

0.57

DMU10

B

0.82

0.66

DMU11 DMU12

B B

1.00 0.79

1.00 0.67

DMU13

B

1.00

0.77

DMU7-0.69,DMU30.29 1 2 5 DMU7-0.65,DMU2 -0.46 DMU11-0.71,DMU30.41,DMU7-0.05 3 DMU11-1.01,DMU40.04 DMU11-0.92,DMU40.16 DMU11-0.85,DMU40.09 6 DMU11-0.96,DMU40.08 DMU11-0.80,DMU40.08

0.93 0.09 53.85

0.85 0.15 38.46

Average St. dev. Efficient (%)

n In PCA–DEA model 2 input and 2 output principal components are extracted with 97% and 98,46% of retained information.

measuring with standard DEA approach. 4.1. Fleet efficiency measuring The efficiency scores for thirteen fleets of two trading companies which operate in Serbia are analyzed in this section. Model for fleet’s efficiency evaluation, as higher level of efficiency measuring is input oriented. The main objective of each fleet is to reduce number of vehicle while maintaining the same level of output. Also the objective is to reduce the fuel costs and total truck time. Results of two different approaches are shown in Table 5. For the CCR DEA model we used EMS software, while for PCA–DEA model we use PCA–DEA software (Adler and Yazhemsky, 2010). The average efficiency according to the CCR DEA model is 0.93, standard deviation 0.09 and 53% of the observed fleets are efficient. In comparison to the CCR DEA model, the discriminatory power (more inefficient units, lower average efficiency and higher standard deviation – Kao et al., 2011) of the PCA–DEA model is increased. Namely, average efficiency score is 0.85, standard deviation is 0.15 and only 38% of fleets are efficient. For this reason the model with higher discriminatory power (second model) will be considered. 13 fleets observed in this paper can be divided in two groups. Five DMUs are in the efficient group, while eight fleets are in the inefficient group. The proposed model is input oriented, so the inefficient fleets can improve their efficiency by reducing the input variables. The fleet with the lowest efficiency score is the ninth fleet. This fleet is only 57% efficient. This also means that observed fleet can achieve the same level of output with 43% less resources. Similarly DMU 1 can improve efficiency reducing resources for 15%. Other inefficient fleet can improve efficiency in a similar manner. Proposed model also gives information about the benchmarks (reference units) as indicated in the last column of the Table 5. For example reference set for DMU 9 are DMU 4 and DMU 11 with corresponding coefficients respectively 0.16 and 0.92.

DMU Standard vehicles (40) Vehicle 1 (standard) Vehicle 4 (standard) Average St. dev. Refrigerated vehicles (130) Vehicle 94 (refrigerated) Vehicle 101 (refrigerated) Average St. dev.

Efficiency score

0.53 1 0.60 0.20 0.5 1 0.57 0.22

Number of efficient

5 (40)

20 (130)

% of efficient

12.5

15.38

4.2. Vehicle efficiency measuring Standard CCR DEA output oriented model for 170 vehicle efficiency evaluation is used in this paper. On one side the aim of the fleet is to realize all tasks with the least possible number of vehicles, while on the other side the aim of each vehicle is to realize the highest possible outputs. Unlike to the model for fleet efficiency evaluation, a model for measuring efficiency at the operational level is input-oriented. The aim of each vehicle is to shipped as much as possible pallets to the largest possible number of customers. In this case CCR DEA model has enough discriminatory power (number of DMUs is much higher than the number of variables). According to the vehicle type, all vehicles are separated in two groups. The first group consists of 40 standard vehicles, while the second one consists of 130 refrigerated vehicles. The EMS software is used in for efficiency evaluation (http://www. holger-scheel.de/ems/). Efficiency scores are shown in Table 6. According to Table 6, it is easy to see that refrigerated vehicles on average are five model years earlier than the average model years of standard vehicles. The average capacity of standard vehicles is about 1.6 t larger than of refrigerated vehicles. The average efficiency score for 130 refrigerated vehicles is 0.57. Standard vehicles with average efficiency score 0.60 are more efficient than refrigerated vehicles. Efficiency of four randomly selected vehicles (two standard and two refrigerated) is shown in Table 6. As mentioned before, the model is output oriented. In that manner vehicle 4 may become efficient if with the same energy consumption, realizes 47% more deliveries, shipped pallets and driven km. Similarly refrigerated vehicle 94 with the same energy consumption can realize 43% more deliveries, shipped pallets and driven km. Vehicles 4 and 101, represent efficient vehicles. Mentioned vehicles are also benchmarks for inefficient units. There are numerous factors that may affect efficiency scores. In the next section they are analyzed in more details.

5. Factors that affect transport efficiency As mentioned before, six hypotheses are tested in this paper. The first three relate to the fleet efficiency, while the last three relate to the vehicle efficiency. In this paper non-parametrical tests were chosen, because the tested efficiency scores are not normally distributed. For the hypothesis testing the approach proposed by Brockett and Golany (1996) is used. The approach involves a fourstep procedure. Since we have two different groups of fleets for the first three hypotheses we use Man–Whitney rank test (Demirbaga et al., 2010; Moreno, 2008). For the last three hypotheses we use the same methodology with the difference in statistical tests, because there are more than two groups in the last hypotheses (Brockett and Golany, 1996; Sueyoshi and Aoki, 2001; Bayraktar et al., 2012).

M. Andrejić et al. / Transport Policy 45 (2016) 99–106 Table 7 Descriptive statistics for sub samples in hypotheses H1–H3. H1 Sub sample 1 Average 0.99 St. dev. 0.01 Size 7

H2

105

Table 10 Hypotheses tests results H4–H6.

H3

Sub sample 2

Sub sample 1

Sub sample 2

Sub sample 1

Sub sample 2

0.79 0.15 6

0.94 0.06 8

0.82 0.20 5

0.83 0.10 7

0.99 0.01 6

Table 8 Hypotheses tests results H1–H3. Mann–Whitney (α¼ 0.05)

H1

H2

H3

U (α¼ 0.05) Z Asymp. Sig. (2-tailed)-p

4.000  2.640 0.008

12.000  1.178 0.284

5.000  2.298 0.022

According to the results presented in Table 8 there is a difference in fleet efficiency of the observed companies. It can be observed that the average fleet efficiency of company A is about 99%, while the efficiency of company B is 79% (Table 7). This suggests that the efficiencies of the fleets are largely affected by fleet management system. It can be explained by the fact that company A has better information system. Differences in vehicle routing and scheduling, among other things, affect vehicle utilization. In order to verify the second hypothesis, the observed set is divided according to the number of vehicles into small and large fleets. For the critical point eighteen vehicles are taken. The observed set is divided into five big and eight small fleets. The hypothesis of the difference in the fleet efficiencies between small and large fleets is not proved according to the Mann–Whitney and significance level of 0.05. In this case hypothesis is rejected and it is concluded that there is no difference in the efficiency of large and small fleets. According to the observed example economies of scale do not affect efficiency in the observed example. The results are not in accordance with Hackman et al. (2001), but are consistent with results presented in Banaszewska et al. (2012). To identify whether efficiency scores depend on the fleet depot location, observed set is divided into two subgroups according to the gravity area. The first group consists of DCs located in cities with more than two hundred thousand people, while the second consists of DCs located in small cities. According to the results in Table 8 there are differences in the efficiency scores across small and large catchment areas. Various environmental factors that characterize catchment area are the cause of these differences. The fleets that operates in large cities with wider catchment area are less efficient (in average 0.82) than the fleets with narrower

Kruskal–Wallis (α ¼ 0.05)

H4

H5

H6

Chi-square Asymp. sig. (2-tailed)-p

126.341 0.000

59,751 0.000

4.307 0.038

catchment area in smaller cities (in average 0.94). Traffic congestions, greater distances, more frequent disorders, are some reasons of mentioned differences. The hypothesis is confirmed. The results are not in accordance with results presented in Banaszewska et al. (2012). Previous three hypotheses related to the factors that determine fleet efficiency on the higher level of decision making. The next hypotheses analyzes factors that affect vehicle efficiency. Since there is a more than two subgroups of vehicles according to the last three hypotheses Kruskal–Wallis rank test (Brockett and Golany, 1996; Sueyoshi and Aoki, 2001; Bayraktar et al., 2012). The descriptive statistics for sub samples and testing results are shown in Table 9 and Table 10. The fourth hypothesis analyzes differences in efficiency scores between vehicles with different capacity. From the standpoint of capacity observed set of vehicles can be further divided in eight subgroups (2 t, 4 t, 5 t, 7 t, 10 t, 15 t, 24 t and 25 t). According to the results in Table 10 there is a difference in efficiency scores between vehicles with different capacity. According to the results in Tables 9 and 10 it is easy to conclude that vehicles with less capacity are more efficient. The hypothesis of differences between vehicles with different capacity is proved. Vehicle capacity affects efficiency. Results are in accordance with the conclusions made in Kim (2010). In order to verify fifth hypothesis observed vehicles are grouped in ten subgroups according to the age of vehicles (Table 9). The total range of the observed set is 25 years. It is divided in ten subgroups according age of vehicles (in the first group are vehicles produced before 2000, in the second are vehicles produced in 2001, in the third are vehicles produced in 2002, etc.). On an average the vehicles are ten years old (Table 4). The results of statistical differences between observed groups in Table 10 show that there is a statistical difference between mentioned subgroups. The age of vehicles affects efficiency. The hypotheses H5 is proved. As in the previous two cases, observed vehicles are grouped into nine sub-groups depending on the manufacturer. Both tests' results show that there is a difference in efficiency scores between vehicles of different manufacturers (Table 10). The hypotheses H6 is proved. Names of manufacturers are not specified. However, it is important to note that in the manufacturer selection process numerous factors like vehicle price, maintenance costs, warranty, etc. are also important.

Table 9 Descriptive statistics for sub samples in hypotheses H4–H6. Sub samples

H4

H5

H6

Average St. dev Size Average St. dev Size Average St. dev Size

1

2

3

4

5

6

7

8

9

10

1.00 0.00 21.00 0.46 0.09 19.00 0.60 0.03 9.00

0.73 0.00 17.00 0.60 0.13 20.00 1.00 0.00 9.00

0.61 0.00 22.00 0.53 0.06 16.00 0.93 0.11 28.00

0.50 0.00 15.00 0.38 0.14 9.00 0.50 0.00 13.00

0.44 0.00 18.00 0.55 0.07 5.00 0.87 0.17 31.00

0.41 0.00 9.00 0.38 0.01 10.00 0.71 0.06 6.00

0.33 0.00 17.00 0.30 0.11 12.00 0.67 0.10 12.00

0.34 0.01 11.00 0.52 0.23 15.00 0.44 0.00 7.00

0.48 0.04 20.00 0.93 0.13 15.00

0.27 0.09 4.00

11

12

M. Andrejić et al. / Transport Policy 45 (2016) 99–106

106

H2 H1 Transport management

+

behavior, etc. Mentioned factors are not examined in this paper and in the future models it would be desirable to introduce these indicators. The future models should also include indicators of the greenhouse gas emission, and the other undesirable outputs.

H3

Fleet size

Gravity area

+

H4

Fleet efficiency

Vehicle capacity

+

H5 +

Age of vehicle

Vehicle efficiency +

H6 Manufacturer

Fig. 2. Factor that affects transport efficiency.

Testing the set of six hypotheses, only one of them is not proved. As previously mentioned, two basic levels of the transport efficiency measurement are observed (Fig. 2). Based on the observed DCs it may be concluded that the efficiency of the fleet is significantly influenced by fleet management system and catchment area. Fleet size does not have enough statistically significant influence. On the operational level three main factors are found. It is also found that the vehicles with smaller capacity are more efficient than the larger vehicles. The age of vehicles is also important for the vehicle efficiency. Mangers in some DCs have preferred certain manufacturers. According to the last hypothesis, managers assumptions are accurate to some extent.

6. Conclusions This paper analyzes the transport efficiency in Serbian DCs. Two basic aspects of analyzing transport efficiency are described. On the tactical level the PCA–DEA model for fleet efficiency evaluation is proposed. The PCA–DEA model has more discriminatory power than the standard DEA models. Number of vehicles, fuel costs and total trucks time in operation are used as inputs while total distance driven, tons shipped and trucks utilization as outputs in the proposed model. On the operational level vehicle efficiency is evaluating with the standard DEA approach. Fuel consumption is used as input, while number of deliveries (stops), distance driven and shipped pallets are used as outputs. According to the examples analyzed in this paper it is concluded that it is not possible to define universal model for measuring transport efficiency. The process of measuring the efficiency among others depends on concrete example, and on the measurement level. It is important to note that the proposed models also give the information about corrective actions for the particular fleet and vehicles. Numerous factors influence the transport efficiency. In this paper the main factors are identified. For the fleet efficiency and for the vehicle efficiency three factors are recognized. In that manner six hypotheses are set. Five of them are proved. The fleet management is crucial for the fleet efficiency. For the vehicle efficiency, and indirectly for the fleet efficiency, the age of vehicle, capacity and manufacturer are essential. The fleet efficiency is also influenced by a great number of factors upon which the company management has no influence: weather conditions, market situation, competition, driver

References Adler, N., Berechman, J., 2001. Measuring airport quality from the airlines' viewpoint: an application of data envelopment analysis. Transp. Policy 8, 171–181. Adler, N., Golany, B., 2001. Evaluation of deregulated airline networks using data envelopment analysis combined with principal component analysis with an application to Western Europe. Eur. J. Oper. Res. 132, 260–273. Adler, N., Golany, B., 2002. Including principal component weights to improve discrimination in data envelopment analysis. J. Oper. Res. Soc. 53, 985–991. Adler, N., Yazhemsky, E., 2010. Improving discrimination in data envelopment analysis: PCA–DEA or variable reduction. Eur. J. Oper. Res. 202, 273–284. Andrejić, M., Bojović, N., Kilibarda, M., 2013. Benchmarking distribution centres using principal component analysis and data envelopment analysis: a case study of Serbia. Exp. Syst. Appl. 40, 3926–3933. Banaszewska, А, Cruijssen, F., Dullaert, W., Gerdessen, J.C., 2012. A framework for measuring efficiency levels – the case of express depots. Int. J. Prod. Econ. 139, 484–495. Bayraktar, E., Tatoglu, E., Turkyilmaz, A., Delen, D., Zaim, S., 2012. Measuring the efficiency of customer satisfaction and loyalty for mobile phone brands with DEA. Exp. Syst. Appl. 39, 99–106. Boussofiane, A., Dyson, R.G., Thanassoulis, E., 1991. Applied data envelopment analysis. Eur. J. Oper. Res. 52, 1–15. Brockett, P., Golany, B., 1996. Using rank statistics for determiningprogrammatic efficiency differences in data envelopment analysis. Manag. Sci. 42, 466–472. Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efficiency of decision making units. Eur. J. Oper. Res. 2, 429–444. Chow, G., Heaver, T.D., Henriksson, L.E., 1994. Logistics performance: definition and measurement. Int. J. Phys. Distrib. Logist. Manag. 24, 17–28. Cruijssen, F., Dullaert, W., Joro, T., 2010. Freight transportation efficiency through horizontal cooperation in Flanders. Int. J. Logist.: Res. Appl. 13, 161–178. Demirbaga, M., Tatoglu, E., Glaister, K.W., Zaim, S., 2010. Measuring strategic decision making efficiency in different country contexts: a comparison of British and Turkish firms. Omega 38, 95–104. Donselaar, K., van Kokke, K., Allessie, M., 1998. Performance measurement in the transport and distribution sector. Int. Phys. Distrib. Logist. Manag. 28, 434–450. Drake, L., Howcroft, B., 1994. Relative efficiency in the branch network of a UK bank: an empirical study. Omega 22, 83–90. Hackman, S., Frazelle, E., Griffin, P.M., Griffin, S.O., Vlasta, D.A., 2001. Benchmarking warehousing and distribution operations: an input–output approach. J. Prod. Anal. 16, 79–100. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1995. Multivariate Data Analysis. Prentice Hall Publishers, New York. Hervani, A.A., Helms, M.M., Sarkis, J., 2005. Performance measurement for green supply chain management. Benchmarking: Int. J. 12, 330–353. Kalenoja, H., Kallionpää, E., Jarkko, R.J., 2011. Indicators of energy efficiency of supply chains. Int. J. Logist.: Res. Appl. 14, 77–95. Kao, L.J., Lu, C.J., Chiu, C.C., 2011. Efficiency measurement using independent component analysis and data envelopment analysis. Eur. J. Oper. Res. 210, 310–317. Kim, T., 2010. Efficiency of trucks in logistics: technical and scale efficiency. Asian J. Qual. 11, 89–96. Kuosmanen, T., Kortelainen, M., 2005. Measuring eco-efficiency of production with data envelopment analysis. J. Ind. Ecol. 9, 59–72. Mckinnon, A., 1999. Vehicle Utilisation and Energy Efficiency in the Food Supply Chain. School of Management Heriot–Watt University, Edinburgh. Min, H., Joo, S.J., 2006. Benchmarking the operational efficiency of third party logistics providers using data envelopment analysis. Supply Chain Manag: Int. J. 11, 259–265. Moreno, J.J., 2008. Efficiency and regulation in Spanish hypermarket retail trade: a crosssection approach. Int. J. Retail Distrib. Manag. 36, 71–88. Neto, J.Q.F., Walther, G., Bloemhof, J., van Nunen, J.A.E.E., Spengler, T., 2009. A methodology for assessing eco-efficiency in logistics networks. Eur. J. Oper. Res. 193, 670–682. Pedraza-Martinez, A.J., Van Wassenhove, L.N., 2012. Transportation and vehicle fleet management in humanitarian logistics: challenges for future research. Euro J. Transp. Logist. 1, 185–196. Sarkis, J., Talluri, S., 2004. Ecoefficiency measurement using data envelopment analysis: research and practitioner issues. J. Environ. Assess. Policy Manag. 6, 91–123. Sharma, S., 1996. Applied Multivariate Techniques. John Wiley & Sons, New York. Simons, D., Mason, R., Gardner, B., 2004. Overall vehicle effectiveness. Int. J. Logist.: Res. Appl. 7, 119–135. Sueyoshi, T., Aoki, S., 2001. A use of a nonparametric statistic for DEA frontier shift: the Kruskal and Wallis rank test. Omega 29, 1–18. Zhou, G., Min, H., Xu, C., Cao, Z., 2008. Evaluating the comparative efficiency of Chinese third-party logistics providers using data envelopment analysis. Int.. J. Phys. Distrib. Logist. Manag. 38, 262–279. http://www.holger-scheel.de/ems/ (the EMS Homepage; accessed 15 November 2013).