Choosing logistics service suppliers: customers` perspective in benchmarking container terminals

Another perspective from which the problem of port choice is considered is how well the port is integrated in overall logistics network. Tran (2011) developed the model which focuses on total costs and aims at minimizing it. In total costs the author includes ship costs, port tariffs, inland transport costs and inventory costs. He applied brute-force algorithm for solving this optimization problem on real data: the author considered cargo flows between USA and Northern Europe. He concluded that the port choice should be assessed not only from the perspectives of sea port activities and sea routing, but also include inland component.

The abovementioned studies concern mainly the determination of the significant factors of a port by conduction a survey or factor analysis from shipping line`s perspective. However, besides the goal of discovery these factors, some studies aim to provide the solution to the problem of a port choice by implementing the mathematical models. The employed for shipping lines` point of view models are the same as were implemented for the shippers.

The AHP model was used for predicting the best choice of a port and identifying the most important criteria in several studies. Lirn et al. (2004) viewed an extensive set of factors (47) that are attributable to the ports. He distinguished between cost and facility/service factors, the sub-factors of which are shown in the table below. The author added for consideration the factor of port management and administration, which was not considered in previous studies. The author found that the factor of port transportation location is the most important.

Some of the studies suggest high practical importance. For example, Lam and Dain (2012) suggested a decision support system for port selection, using AHP methodology from the shipping lines perspective. The authors state that this methodology is advantageous for its users: “AHP addresses the issue of how to structure complex decision-making problem, identify its criteria, measures the interaction among them and finally synthetize all the information to arrive at priorities which depict preferences.” Thus, the line managers can use this system for solving complex problem of choosing port by utilizing multi-criteria analysis. The suggested model is web-based and can be used by different decision makers. The system is flexible in the way that users can choose criteria by themselves and set preferences as well, according to the pursued purposes and particular cases. The authors claim that such technological advancements would be beneficial for shipping managers to use because it enhances the service quality of shipping companies.

Table 3. The use of AHP in studies: shipping companies` perspective

Authors	Factors
	Service	Costs
Park and Min (2011)	· Port's proximity to import/export businesses · Port service quality · Port security · Port management efficiency · Basic infrastructure · Information technology infrastructure · Intermodal link · Feeder service access · Access to major shipping routes · Ship turnaround time · Port security · Carrier bargaining opportunity	· Container handling costs · Terminal contract costs
Guy and Urli (2006)	1. Port infrastructure · Water depth · Quay length · Crane · Inter-modal interface 2. Service _ turn-around time 3. Geographical location · Immediate hinterland · Extended hinterland · Possibility to serve other port within the same service loop	· Total transit costs
Lirn et al. (2004)	1. Port physical and technical infrastructure · Basic infrastructure · condition · Technical infrastructure · Inter-modal links 2. Port geographical location · Proximity to import/ · export areas · Proximity to feeder ports · Proximity to main · navigation routes 3. Port management and administration · Management and · Administration efficiency · Vessel turnaround time · Port security/safety	· Handling cost of containers · Storage cost of containers · Terminal ownership/Exclusive contract policy

Besides AHP model, the MNL model was also used in several studies for predicting a choice of port from shipping lines` perspective.

In the studies from shippers` perspective the generalized discrete choice model was not developed with the port or transportation industry in mind. The missing element of the direct application of the MNL framework to model port choices in these studies is the indispensable element of network. The service network structure, which is a key element of the container port industry, arises from the main business of a port. The main business of a port, defined in Wang and Cullinane (2005) as the facilitation of cargo transportation from point of supply to point of demand, also bestows a critical role upon the port network-connectivity that determines the competitiveness of a port to a large extent Cullinane, K., Song, D. W., & Wang, T. (2005). The application of mathematical programming approaches to estimating container port production efficiency. Journal of Productivity Analysis, 24(1), 73-92.

Tang et al. (2008) covered this gap by implementing one specific model, called Network-based Integrated Choice Evaluation (NICE) model. This model unites the network characteristics of the port industry into the multinomial logit preference model (MNL) by the connectivity index. The considered model uses published schedules from liner shipping companies to establish the service network of ports and obtain the associated port connectivity indices. The NICE model also applies factor analysis on observational port attributes (such as port charges, turnaround time, annual operating hours, water depth and so forth) to derive port operating dimensions that are mutually and preferentially independent. The NICE model is empirically determined by expressing the port connectivity index as a conditional MNL function of these mutually and preferentially-independent port operating dimensions that allows for an assessment of the marginal contributions of each dimension separately. Tang, L. C., Low, J. M., & Lam, S. W. (2011). Understanding port choice behavior--a network perspective. Networks and Spatial Economics, 11(1), 65-82. Empirical results obtained by applying this model for the process of choice of Asian ports reveal that port efficiency is most important factor in increasing the attractiveness of ports; other attributes which are indispensable for a port to perform in order to be competitive are scale economies and convenience.

Overall, numerous studies dedicate to the perspective of maritime lines in the problem of a port choice. The results vary from one study to another in terms of the order of important factors. However, the main ones remain the same: costs, location, connectivity, efficiency and some other factors. Inside these factors, the sub-factors can be defined, which differ in various studies. Different methods yield different results.

Port operators` perspective. Besides external customers` perspective there are a few studies, which examine the port service providers' interests in choosing a port. Sanchez et al. (2011) compared the important attributes of users and providers for Latin America and Asia region. They pursued the goal of evaluating, whether the perceptions of port operators and shipping lines contradict to each other. They observed that there is a noticeable difference between the perception of port operators about shipping lines` preferences and needs, and the factual shipping lines` needs. The differences are significant in monetary cost, geographical location, and, perhaps most importantly, the speed of ports in responding to the new demands of shipping lines. Sanchez, R. J., Ng, A. K., & Garcia-Alonso, L. (2011). Port Selection Factors and Attractiveness: The Service Providers' Perspective. Transportation journal, 50(2), 141-161

The same topic of comparisons of different parties` interests was discussed in the study of Park and Min who in 2011 proposed a hybrid data envelopment analysis and analytical hierarchy process (AHP) model that allows to identify factors specifically influencing transshipment port selection, evaluates the degree of impact of those factors on a transshipment port selection decision, and then determines the most important ones among various factors. The analysis concerns both carriers of cargo and port service providers. They first identified port selection factors through survey, after they indicated the relative weights (importance) of those factors by implementing AHP model, and after, by using DEA they showed the “contribution” of each factor to port selection. The main distinction from the previous researches of these authors is that this model allows identifying the impact of each factor to the port “attractiveness”. They also showed that in this research cost factor remains important, but at the same time non-monetary qualitative attributes such as intermodal links and feeder service access are also significant when identifying port “attractiveness”. At the same time, such characteristic as technological infrastructure was not revealed as differentiated one. Moreover, the authors proved again that the importance of different characteristics often depends on the perspective. They demonstrated this fact by considering the shipping line and port operators` points of view, which are different in terms of importance of such attributes as port's proximity to import/export businesses, port service quality, port security, and port management efficiency. The study suggests closer and more proper analysis of maritime carriers` needs and preferences for delivering high-quality service.

1.3 Research gap

Extensive literature review has demonstrated that the problem of port choice was examined in numerous studies, beginning from the 1980th. As port customers differ, the choice process, criteria and outcomes vary according to the perspective from which this process is considered. We considered shippers, freight forwarders, maritime carriers and port operators` perspective. However, inside this problem there are different issues that were not fully covered in the existing scientific studies.

In our study we concentrate on freight forwarders and on the choice problem for this specific group of customers. This group nowadays quite often represents a decision-maker while choosing a port, moderately replacing shippers. However, just a few studies reveal the factors that influence decision-making from freight-forwarders` perspective. Furthermore, in comparison to the other considered perspective, for freight forwarders just one study found that would deal with the problem of port choice and result in the choice outcome. Moreover, the review showed the lack of studies that would take into account the relational part of a decision-making process, namely, what role do the shipping companies play in decision-making process, whether they influence the choice? While such groups of factors as costs, infrastructure, hinterland connection and services are explored in various studies, the relational aspects are paid attention only in studies, considering specific sectors of maritime shipping industry, but not for container transportation.

Also, shippers` perspective demonstrated the dispersion in decision-makers preferences when choosing a port. However, there was no attempt to try to classify the preferences of different customers. As soon as shippers` perspective is less relevant, we would like to concentrate on identification of different types of freight forwarders regarding the important factors for choosing a port.

Moreover, the prediction of choice outcome for a decision-maker, according to his system of preferences, exists for shipping companies and shippers, but not for freight forwarders. The scope of models that are implemented for assigning a port for a decision-maker, also lacks objectivity - those methods need to have numerical values for weights of attributes, which, in real life is a hard condition to satisfy. That is why we are going to implement in this study one of the MCDM models, which doesn`t require numeric information about weights of attributes, and is based on aggregated indices randomization method.

Therefore the research questions are the following:

1. What kind of typology can be implemented to differentiate different port customers represented by freight forwarders?

2. How does the process of choosing a port differ from one type of customer to another?

3. How do relational aspects influence the decision-making process?

4. How to rank the alternatives of ports for each type of customer?

The research questions are designed to find answers that would cover both theoretical and practical issues. In general, there is a lack of information concerning a specific type of customers - freight forwarders, who actually make a decision about port choice. That is why for ports and shipping companies this study is also appealing as a source of customer needs` insights.

2. Methodology

2.1 Theoretical framework

As the literature review showed, the process of choosing a port varies from one type of customer to another. In this study we concentrate on the customers, who very often make a choice - freight forwarders. They make the final choice of the port, which can be completely their own decision, or, it can be an advice from the managers of a shipping company. It is also possible that customers can “push” the shipping company to sign contracts with new ports that appeal the end customers (shippers and freight forwarders). That is why the end customers` perspective is very important.

Figure 2.1. Theoretical model of a choice process

In order to understand the logic of the methodology used, let us present the general setup of the considered problem, which is depicted on the figure 2.1., and explain the main parts. A customer considers port not as a separate logistic entity, but rather as a part of the whole logistic network. That is why it is necessary to consider not only physical attributes of ports`, but also take into account the services that operators of port provide, the hinterland connection, and relational aspects. We will elaborate with more detailed description of attributes for each port`s factors in the 3 paragraph.

Customers have certain preferences about the port`s attributes. For some customers only the costs can matter, while for other ones the hinterland connection is of great importance. The best alternatives that correspond to the customers` preferences will differ for each type of such customers` preferences. That is why it is necessary to identify the system of preferences for each type of customers.

We also include in our model relational part. We will try to understand, whether the customers make the decision about port choice rationally, or they are highly influenced by relational (personal) factors. We will include factors that were not previously considered in such type of studies.

Moreover, customer`s choice is influenced not only by the preferences about port attributes, but also by the choosing the shipping company. In most cases a shipper comes to a shipping company, and then, chooses the port among suggested. So, the question arises, how a customer chooses the shipping company? Namely, what attributes are important in this choosing process?

The answers to all these questions we will obtain at the end of the research. In the following paragraphs we will identify the main approaches to data collections that are going to be used in this study, and then present the mathematical model and tool, with the use of which we will obtain the ranking of the alternatives for different customers types.

Figure 2.1 Hypothetical scheme of customers` typology

2.2 Hypothetical typology

Literature review and interviews with experts in maritime container shipping allows us to suggest initial framework of customers` typology. The main distinction criteria are the following:

· Type of cargo transportation: export/import/export-import/inland transportation;

· Frequency of contacts: inconstant/constant basis;

· The way of decision-making: independent/outsourcing of a decision to a company/individually, but taken into account the opinion of a company;

We suggest that based on these characteristics several types of customers will be identified, who share the same preferences about the importance of port`s attributes. The scheme is presented in the picture below. It has 3 levels: the first distinguishes between occasional and regular customer, the second - between export/import/export-import operations, and the third - between the independent decision/individual, but taken into account managers` opinion decision/ outsourcing of decision. We expect that these criteria influence the preferences that freight forwarders have, and thus will help to identify different types of customers.

2.3 Identifying the type of research

There is a set of methods that could be applied to the business research. In order to explain, which of them we will be used in this paper, let us consider the following framework of Timm Paul and Farr Rick.

Table 2.1. Types of research

According to this table, the research we will conduct suits the best for the applied primary research. First of all, gathering data from customers implies the existence of primary research, so does the gathering data from the industry expert. As far as in this paper we are going to solve the specific problem, we use the applied research type.

2.4 Data collection approaches

In this study we are going to use the following empirical study methods:

1) Questionnaire for customers

2) Questionnaire for expert

As was mentioned above, for carrying out the future research it is necessary to obtain some data sets from both expert and customers. In order to do it, it is important to decide which type of gaining data is suitable for each perspective.

Figure 2.2. - Empirical study methods for the obtaining the information

2.4.1 Expert`s perspective

First of all, let us consider expert`s perspective and empirical study methods that are going to be used.

As it was stated in a paragraph above, expert`s estimations are the source of information about the values of the attributes for the container terminals. Some of this information is going to be obtained via interviews. In such interview sessions we pursue two goals - to make sure that in our research we use the same characteristics that are important for the customers and to understand those characteristics that still are not covered in our research, but they exist for the customer when making a decision with which container terminal collaborate to. We initially came up with a set of attributes to apply based on the results of literature review. After, this set was corrected, taking into account the expert`s opinion.

Interviews were conducted in the very beginning with the purpose of understanding, how, in general, the procedure of communication with customers is carried out, how customers, in general, choose port. Basically, interviews were conducted to facilitate further process of research and for forming of general understanding of a problem.

After attributes are identified, we need to obtain its values. The main factors of port choice, attributes of which we explore, are presented below. The experts are asked to evaluate the performance of each port based on different attributes, using the scale from 1 to 5. The questionnaire for expert is based on the same attributes as customers` questionnaire, excluding only those attributes that need to be identified by customers itself, due to individual values of these attributes for every customer. Experts` results are presented in the third chapter of the thesis.

Table 2. 2. Container terminal factors to measure

	Factors of port`s choosing	The scale
1	Physical infrastructure	1…....2…….3…….4…….5
2	Services provided	1…....2…….3…….4…….5
3	Hinterland connection	1…....2…….3…….4…….5
4	Costs	1…....2…….3…….4…….5
5	Relational part	1…....2…….3…….4…….5

2.4.2 Customers` perspective

After completing gathering information about characteristics and its values, we proceed to customers` preferences itself. The questionnaire for customers, in our case for freight forwarders, contains the following structure:

· General information about the client. This information is going to be used as a base for identification of customer typology;

· Port choice criteria. In this part client will evaluate, to which extent each attribute is important for him when choosing a port.

· The ranking of the important attributes;

· The ranking of the ports/terminals. Customer can already have certain preferences about the ports/terminals.

· Criteria for the choice of shipping company. Here a customer evaluates different attributes associating with shipping company choice in terms of its importance.

Fig. 2. The system of data gathering

The questionnaire for customers is attached as an appendix.

In general, the system of obtaining data can be presented the following way:

After completing these steps we can proceed further, namely to describe the model, which is going to be used, and the method, which underlies the model.

2.5 Quantitative model

The main questions that we want to answer in this chapter are: what we want to measure and which models and methods can be used for that. After the implementing the methodology described below we want to get the aggregated quantitative indices that would reflect the “attractiveness” of alternatives (in our case -ports). These indices will take into account the system of preferences of a customer concerning the importance of the characteristics of the terminals.

In order to formulate the considering problem in the right terms and then build proper model let us describe the theoretical model of the problem and its mathematic representation.

2.5.1 Multiple-criteria decision analysis

The model that is used to solve the considered in this thesis problem belongs to the discipline called multiple-criteria decision analysis (MCDA). We were partly considering it in the first chapter. This discipline is a part of operations research which deals with the situation when the decision needs to be made concerning different issues (both for our daily of professional lives) explicitly taking into account multiple criteria. The problem is most likely about making choice between different alternatives. Due to existence of multiple criteria there is no unique optimal solution for such problems; in order to solve it we need to consider the information about preferences of the decision-maker.

There are different classifications concerning MCDA problems and methods. One of the basic ones is to distinguish multi-criteria evaluation problems from that of multi-criteria design (or multiple objective mathematical programming). The main difference is that the former one has explicit set of alternatives, whereas in the latter ones alternatives are implicitly known. Now let us process towards the multi-criteria evaluation problems, as we relate the problem described above to this class of problems.

2.5.2 Multiple-criteria evaluation problems

For such problems each alternative has a set of performance measures for each attribute. The problem may be defined as finding the best alternative for a decision maker (DM), or finding a set of good alternatives. One may also be interested in "ranking" or "classifying" alternatives.

The general procedure within this multi-criteria evaluation problem for solving can be represented the following way:

Fig.3 The scheme of model building

The two main problems arise from the framework stated above: what kind of function should be used when identifying the overall value based on scores for criteria and weights? What weights should be used for the analysis? The answers for these questions distinguish the sets of methods which are used for solving such kind of problem. The last question is of great importance because in real business or daily life we can hardly indicate the exact values for the weights of the attributes. Instead, we are likely to use the equalities and inequalities for identifying which attribute is more important. We proceed to the next paragraph, concerning the description of model that we are going to use in this study. The method is called aggregated indices randomization method (AIRM).

2.5.3 Aggregated indices randomization method

This methods belongs to a set of aggregated indices methods, but implies different approach towards the choosing the weights for the criteria. Let us now briefly explain the mathematical model of the problem of estimating any complex object, and further to apply this model for our particular problem of container terminal estimation for the customer.

In the considered model a DM estimates the alternatives, using a set of attributes. Thus, those alternatives can be called multi-attributes alternatives.

A numerical value of an attribute for a given alternative determines an estimation of the alternative's preference, this estimation being a numerical function of the attribute's value. Such functions of the attributes' values are named individual preference indices. Any individual preference index can be called an individual criterion of preference. That is why, a set of all individual criteria's values for a particular alternative determines a multi-criteria estimation of the alternative's preference.

The initial assumption of the method used is that each of the constructed individual preference criterion is necessary, and the whole set of them is sufficient for a numerical estimation of any alternative's preference. It means that a numerical estimation of an entire alternative's preference is a numerical function of the set of all single preference criteria. Such numerical function of all single criteria of preference is named aggregated preference index, and is treated as an aggregated criterion of the alternatives' preference. Value of an aggregated preference index for a given alternative is its preference estimation which takes into account the whole set of single estimations of the alternative's preference. Hovanov, N., Kornikov, V., & Seregin, I. (1997, June). Qualitative information processing in DSS ASPID-3W for complex objects estimation under uncertainty. In Proceedings of the International Conference" Informatics and Control". St. Petersburg (Russia) (pp. 808-816)

Additionally it is assumed that an aggregative function (i.e. function which determines a corresponding aggregated index) makes allowance for significance (synonyms: importance, influence, weight, etc.) of different single performance indices for the aggregated preference index. Namely, the aggregative function is supposed to be determined by appropriate non-negative parameters which are named weight-coefficients (“weights”), and which play role of single indices' significance estimations.

Briefly, the mathematical model includes implementation of the following steps:

1) Forming the vector of the initial characteristics, any of which is necessary, and all together they sufficiently represent the quality of the evaluated objects.

2) Forming the vector of individual indices which are the functions of the corresponding initial characteristics which evaluate the observable object with the use of m different characteristics

3) Choosing the form of aggregated function , which characterizes the quality of the project in integrally. It is assumed that function depends on the system of weights which are non-negative parameters representing the degree of the importance of each individual index for the aggregated index.

4) The defining the system of weights which shows the “weight” of every individual indexin. The additional clause of normalization allows to define as an estimate of the relative weight of the individual index ,.

Each alternative Ai can be evaluated according to the described above methodology.

Further description of the method has to be connected to the special software that allows to solve two main important issues concerning the forming of the single preference vector and choice of weights based on non-numerical, non-exact, non-complete information.

Let us present further this methodology adopted for our problem of choosing.

2.5.4 Problem formulation in methodological terms

The first step of the methodology presented above is to defineof initial characteristics. For our problem, stands for the attributes of container terminals which were stated in the third paragraph, such as road access, the cost of cargo transshipment and so on.

After, the forming of individual indices needs to be provided. In order to use the software DSS APIS we need the information obtained through the questionnaires from the company about the values of the attributes of container terminals. This information is downloaded into the software with the appropriate normalizing function choice.

The type of the normalizing function is defined according to the following rule: if after increase of the value of the attribute the attractiveness of an alternative increases, then the type of the function is increasing. In the opposite case, the type is decreasing. For example, the higher the costs of transportation are, the less an alternative is attractive. In our case only one attribute has a decreasing function.

After that all information from clients, concerning their preferences about the importance of different terminals` attributes, needs to be numerically estimated. For this purpose in the software it is possible to get certain weights based on this non-numeric, non-exact information for each customer.

The last step is to get the aggregated indicator of overall attractiveness of the object (container terminal) for the customer according to his preferences (which are already transformed into numerical weights). After obtaining all aggregated indicators those alternatives of container terminal can be ranked in descending order for each client.

Moreover, customers can be grouped in different segments according to the type of shipment, regularity of relationships with shipping company, and preferences. For each type of the customers the best alternative can be identified. In this case the information about the values of the attributes should be taken from the experts` interviews.

2.6 Decision support system APIS

For the realization of the method described above, the computer software is used. It is called DSS (decision support system) APIS (Aggregated Preference Indices System) which is used for estimation of complex multi-attribute objects of different nature under uncertainty.

Let us list the main advantages of this software for solving our problem:

1) It allows avoiding the problem associated with different scales of the initial attributes by forming normalizing function of the individual indices.

More precisely, 2 cases of relationships between the initial attributes and individual indices of preference are possible:

· The degree of preference qj is increasing when value of attribute aj is increasing on interval [MINj,MAXj]

· The degree of preference qj is decreasing when value of attribute aj is increasing on interval [MINj,MAXj]

After formation of monotone normalization functions qj=qj(aj), j=1,…,m, values of all individual preference indices for all alternatives can be calculated.

2) It takes into account the absence of exact weights and transform ordinal information into exact weights with the use of formulas.

DSS APIS differs from the other decision-support systems by allowing to use the following types of information:

Non-numeric information - non-numeric information (ordinal information) on weight-coefficients values. It can be represented by a system OI(w)={wr=ws;wu>wv;…} of equalities and inequalities for weight-coefficients.

Non-exact information - non-exact information (interval information) is defined by a system II(w)={aj<=wj<=bj;…} of inequalities and equalities (when aj=bj) for weight-coefficients.

NNN-information on weights - ordinal (non-numeric), interval (non-exact), and non-complete (incomplete) information on weights is a combination I(w) of non-exact information (interval information) II(w) on weights and non-numeric information (ordinal information) OI(w) on weights.

At the end, values of an aggregated preference index Q(q;w) for alternative A(i) calculated These values are elements of set of all possible values of aggregated preference index for alternative A(i), (i = 1,…,k).

Alternatives can be ranked according to its aggregated preference index, which is average aggregated preference estimations. Moreover, APIS allows to get the standard deviations of the estimations. As a result, APIS provides the graphical representation of the result, which shows the ranking of all the alternatives for the decision-maker.

Summary

In this chapter we suggested the methodology to be used in our research. First of all, the type of research is identified - in this study we conduct primary and applied research. The main source of the data needed is questionnaire - for expert and chosen group of customers - freight forwarders.

In the very beginning of this chapter we presented the hypothetical model of customers` typology. It is based on 3 main criteria: the frequency of calling to shipping company, the nature of decision-making in terms of independency, the type of cargo transporting. It is supposed that each type will have some distinctive types of preferences and thus, different solutions of port selection need to be provided.

For solving the problem of port selection we use the method called AIRM and the computer-based program APIS. This program takes into account customers` preferences about the importance of different port`s attributes, the values of those attributes and the preferences about alternatives itself (if there are any). The values of the port`s attributes are obtained from expert. The program allows to aggregate the data into one index that characterizes the overall attractiveness of the port for a customer. After that it provides the benchmarking of ports for the considered customer.

The advantage of this model is that it doesn`t require exact numeric information about the weights of attributes. It generates it itself after obtaining interval and ordinal information about attributes. Thus, at the end we have the benchmarking of alternatives (ports in our case) according to the aggregated indices, and also the weights of the attributes, reflecting, to which extent different attributes are important for the customer.

3. Empirical part

3.1 Data collection

In this study we investigate the customers of ports and shipping companies in Saint-Petersburg, and the way, how they choose a port for cargo transportation. For this purpose we addressed to Maersk Line`s clients, which are mainly represented by freight forwarders and represent the huge part of all shipping companies and ports` clients. Let us now briefly describe terminals and ports, which we consider to be alternatives for selection in Saint-Petersburg:

1) First Container Terminal (FCT). This one is the leader in the field of handling with containers in Europe. It cooperates with a lot of shipping lines and appears to be one of the most important transportation point on the route of cargo.

2) Petrolesport. Petrolesport is a modern technically advanced port complex, which includes container terminal and ferry terminal. Petrolesport performs all necessary activities for the container handling: the loading/unloading, storage, forwarding, and handling of various cargo types, as well as all customs related activities, and other operations. It provides the port services for all kinds of cargo: refrigerated, ferry, timber, general cargoes, and is one of the leading ports in the Northwestern Federal District of Russian Federation.

3) Moby Dick. This port is situated near St. Petersburg, in Krondshtat. It is comparatively less than above-mentioned ones, but also important because there is always demand for this port from the customer side.

4) Sea fish port. This port exists since 1956 and mainly deals with cargo handling and storage. The port consists of 3 terminals, one of which is constructed for container handling. The port is situated on the south of Saint-Petersburg not far from high-way

5) ULCT. (Ust-Lugsky Container Terminal) This container port in one of the largest in Russia and Eastern Europe with highly technological equipment. It is situated 100 km far from St. Petersburg, in the Gulf of Finland.

3.2 Expert`s estimations

Maersk Line collaborated with these terminals and ports for a long time. Let us now present the expert opinion about the values of port attributes that were identified in our questionnaire.

Initially the questionnaire was compiled in such a way so as to completely reflect the main port`s attributes, which were first identified through literature review, and then verified by expert. The questionnaire accounted for 19 attributes that need to be evaluated. However, some of them were excluded due to the equal values for all 5 container terminals. The employed model, APIS, does not allow to proceed with the equal values for attributes because it means that the alternatives are indistinguishable for a customer regarding this particular attribute. Due to this reason we excluded the following attributes: cargo safety, port safety, the absence of drug cases in a port. Moreover, we could not include those characteristics that have different values for different customers, for example, existence of personal relations, the proximity to business (which includes two attributes). These attributes are important for the customers, even if the alternatives are indistinguishable according to these factors. At the end we obtained 13 attributes for which we got numerical values from the experts. You can see these values in the Table 3.1.

Table 3.1. Experts` evaluations of ports` attributes (on to scale from 1 to 5)

№	Attributes	Petrolesport	First Container Terminal	Ust-Lugsky Container Terminal	Moby Dick	Sea Fish port
1	Technical infrastructure	4	3	2	3	2
2	Special equipment	5	4	2	4	2
3	Availability of storage services	3	3	4	4	4
4	Timeliness of loading/unloading of containers	4	4	4	3	4
5	Custom clearing procedure	4	4	4	3	3
6	Qualified personel	3	4	4	3	2
7	Delivery of containers in and out	5	4	4	4	3
8	Developed transportation network around a port	5	4	2	4	4
9	Opportunity to choose the time of cargo taking out	4	4	5	4	3
10	The conditions of cargo storage in port	3	3	5	5	4
11	Port reputation	4	5	4	4	3
12	Recommendation to the clients	5	5	5	5	4
13	Total costs	3	3	5	4	3

In general, each of the port has certain advantages and disadvantages. That is why the choice of a port depends to a large extent to the customer preferences and specific needs. Moreover, the difference in choice could be a result of initial customer differences, such as regularity of applying to shipping line`s services or, type of cargo transportation (export/import), or some other specification. This would result in identifying customer typology.

As for the data from customers, we got 5 different types of answers on questionnaire designed for customers. The results of these answers we will discuss further in the section “customer typology”.

3.3 Findings and the results of empirical study

As the results of empirical study we obtained the estimations of attributes of maritime terminals from the industry expert and the results of survey for the users of these terminals - freight forwarders. First of all, we provide the analysis of expert estimations in order to understand, how ranking would look from their perspective. After that we suggest the ranking of ports for freight forwarders, using APIS. We will see further, whether there are types of customers, for which the rankings coincide.

3.3.1 Estimation of alternatives from company`s perspective

Having received the values of the attributes from industry expert, we evaluate the attractiveness of each port from the expert`s perspective, without taking into account the preferences of customers. We obtain the ranked alternatives, based on the estimations of the industry expert.

In order to evaluate the alternatives we look first at the sum of values of attributes from the Table 3.1., and then at the elements of this sum according to the grades (from 2 to 5). We can see this analysis in the table below.

Table 3.2. Estimation of alternatives from the expert

Marks	Petrolesport	First Container Terminal	Ust-Lugsky Container Terminal	Moby Dick	Sea Fish port
Sum	52	50	50	50	41
"5"	20	10	20	10	0
"4"	20	28	24	28	20
"3"	12	12	0	12	15
"2"	0	0	6	0	6
Rank	1	2	2	2	3

The table demonstrates that Petrolesport has the highest rank, which is equal to 52, followed by FCT, ULCT and Moby Dick that share the same place because they have the equal points. Sea Fish port has only 41 points and due to this fact it is places on the lowest position. It should be noticed that even if ULCT is ranked the 2nd, it has some very low estimations of the attributes (“2”), but FCT and Moby Dick doesn't have it. Sea Fish port doesn't have any highest-graded attributes (“5”). The distributions of values of the FCT and Moby Dick are the same (10 for the “5” grade, 28 for “4” grade and 12 for “3” grade).

So we can conclude that from expert perspective Petrolesport is the most attractive port to choose, followed by FCT, ULKT and Moby Dick, which share the same positions, and at last, Sea Fish port. As we have conducted this analysis of the alternatives` attractiveness from expert`s perspective, we can look at the customers` perspective and compare the results.

3.3.2 Customers` typology

We identified 5 types of the customers` preferences based on the importance of different attributes that customers have. We used the answers on questionnaire about the importance of different characteristics. Respondents had to estimate the importance of ports` attributes and also gave their preferences about the ports itself (if they have any). Customers can not give their preferences about all the characteristics, they just choose those, which are the most important for them, indicating also others, which are less important. After we look at the attributes that matter the most for the customers - these attributes form the basis of the typology. Let us present this typology in the next paragraph.

For further convenience we denote the port`s attributes as follows:

Table 3.2. Denotation of attributes

№	Attributes
X1	Technical infrastructure
X2	Special equipment
X3	Availability of storage services
X4	Timeliness of loading/unloading of containers
X5	Customs clearing procedure
X6	Qualified personel
X7	Delivery of containers in and out
X8	Developed transportation network around a port
X9	Opportunity to choose the time of cargo taking out
X10	The conditions of cargo storage in port
X11	Port reputation
X12	Recommendation to the clients
X13	Total costs

The 1st type of customers` preferences

The first type of customer put on the first places the factors connected to the service and physical conditions of a port, namely technical infrastructure (X1), special equipment (X2), and qualified personnel (X6). Moreover, for this type of the customers the recommendations from the company are important (X12).

After we put into the APIS the preferences of such type, we got the diagram, which represents the ranked attributes from the most important to the least important. The dash crossing the red horizontal line indicates the average importance of this attribute for the customer. For example, for considered type the highest importance (weight) is given for the “qualified personnel” attribute (X6), and it accounts for 0,32. The sum of all attributes` average weights equals to unity. We can see that the first 4 positions are taken by 4 attributes mentioned above. For this type of customer such characteristics as costs, conditions of cargo storage, port reputation are of the least importance. They care the most about the facilities of a port and employees, who these facilities support. Such kind of customers transport unordinary cargo, which require existence of special equipment for loading/unloading of containers.

Diagram 3.1. Ranking of attributes for type 1

Besides the information about the distribution of weights for attributes we also obtain the ranking of ports, according to the customers` preferences. The Diagram 3.2 shows us the results. We can see that First Container Terminal is the most preferable and attractive for this customer (it has 0,81 out of 1 score), followed by Petrolesport (0,75 out of 1), ULKT (0,71 out of 1), Moby Dick (0,52 out of 1), and the most non-preferred is Sea Fish port (0,11 out of 1). “The winner” dominates the second-placed alternative with 100% probability, while Petrolesport does it with just 55% probability (the blue line indicates it). Sea Fish port has significantly lower aggregate index, than the rest of ports.

Diagram 3.2. Ranking of alternatives for the type 1

Customer itself in questionnaire chose the following ranking of alternatives: first - FCT, then Petrolesport, and third, Moby Dick. The obtained from the APIS results coincide with the preferences from the questionnaire to large extent. However, the diagram shows that instead of putting Moby Dick on the third place, this customer should rather consider ULKT. In comparison to expert opinion, Petrolesport doesn`t look like a perfect alternative for this type.

The 2nd type of customers` preferences

In comparison to the first type of customers, the second one doesn`t seek for the technical infrastructure of a port. For this type costs and port reputation are the most important, followed by the customer clearing procedure and timeliness of operations.

In the diagram below the single preference indices ranking is shown, by average values of their weights. It indicates that the most important criteria is port reputation (X11), followed by expenses (X13), timeliness of operations (X4) and customer clearing procedure (X5). While port reputation has the weight equals to 0,25, the following 3 attributes have approximately the same weight around 0,13, as Diagram 3.3 indicates.

Such type of the customers is indifferent to technical characteristics, but care about simplicity, speed and costs of shipment. He carries technically simple cargo, non-perishable. What is also interesting is that for this type the existence of personal relations is important to the large extent. It is not reflected in the table below due to the limitations of the used model.

Diagram 3.3 Ranking of attributes for type 2

The results in the form of ranked alternatives are shown in the Diagram 3.4. It indicates that the most attractive alternative for such type of the customers is ULKT, which dominates the next alternative with the 71% probability. The rest of alternatives are arranged in the following order: second - FCT, with the aggregate index equals to 0,65; third - Petrolesport, with the aggregate index 0,6; the forth position is occupied by Moby Dick, and the fifth by Sea Fish port with the lowest index 0,2. Such results partly coincide with those, indicated in questionnaire - ULKT is again not taken into account, while this is the most preferred alternative. The order after ULKT is indeed chosen by the customers. If to speak about the similarities between this type and the expert opinion, there is a little similarity, as with 71% probability ULKT is better than FCT, which is, in turn, better, than Petrolesport with 100% probability.

As this type appraises the most port reputation and costs, ULKT alternative needs to be chosen. However, customers either are not offered this port or they for some unrevealed reasons prefer other ports. The attribute of distance for this type is not important, meaning that some other reasons for not choosing this port can exist.



Diagram 3.4. Ranking of alternatives for the type 2

The 3rd type of customers` preferences

The following type of customers is characterized by strong position about storage services (X3 and X10), customs clearing procedure (X5), and the costs (X13), which represent the result of efficient storage and customs services. Such type of the customers most often deals with the import cargo, implying high-quality storage services and efficient custom procedure. He doesn't pay much attention to the technical facilities and professionalism of employees. Such type of customers could transport perishable items.

The order of attributes according to its weights is presented below. The attribute of costs is put on the first place with the average weight of 0,2, followed by “availability of storage service”, “customs clearing procedure” and the conditions of cargo storage in port” with approximately the same weights, around 0,15, as the Diagram 3.5 shows.

Diagram 3.5 Ranking of attributes for the type 3

As for the ranking of all ports for such type of the customers, the results are shown in the Diagram 3.6. ULKT dominates with 100% probability over Moby Dick, which is put on the second place. The gap between them is quite prominent - ULKT has 0,89 aggregate index, while that for Moby Dick is 0,58. The third place is given to Petrolesport, with 0,4 index, followed by FCT and Sea Fish port. Sea Fish port again the most undesirable alternative, and ULKT the most preferable one.

Diagram 3.6. Ranking of alternatives for the type 3

For this type the results from APIS do not coincide with those obtained from the expert. ULKT prevails over Moby Dick with 100% probability meaning that for all of the attributes this alternative is more preferable for this type of customers, than the rest.

The 4th type of customers` preferences

The forth type of the customers care the most about before- and after- transportation service, highly appraising such attributes as convenience of delivery of containers in and out(X7), qualified personnel (X6), developed transportation network (X8) and timeliness of handling the containers (X4). Attributes X7 and X6 have the highest position in terms of importance, and are designated the same weight, equals to 0,25. Developed transportation network is also important for this type of the customers, in the same extent as timeliness of services. All the attributes are ranked and depicted in the Diagram 3.7.