5. Empirical Analysis

A wide range of other strategic and competitive factors affect organizational performance. Therefore, researchers recommend the evaluation of the impact of IT by examining specific benefits (cp. Barua et al. (1995), p. 6; Dedrick et al. (2003), p. 21).

Cp. Sections 4.​2.​2.​3 and 4.​2.​2.​4.

Cp. Section 5.2.1.

Cp. Gläser and Laudel (2010), pp. 107–108.

Within the system use dimension, the focus on functionality choice has been set because there are not many empirical studies for this design variable. The selection of ETMs is in focus because the criteria for the selection of ETMs are assumed to be easier recollectible by experts than ETM properties which affect the rather abstract concept of commitment.

Cp. Gläser and Laudel (2010), p. 11.

The objectives include obtaining new insights, e.g. for extending the conceptual research framework. To achieve this, semi-structured interviews provide an open and flexible approach for understanding the interviewee’s perceptions and experiences (cp. Mason (2004), p. 1020). Besides leaving room for the thoughts of interviewees, however, the interviews should be centered around the concepts and ideas of the conceptual research framework. Therefore, an interview guideline has been developed and was followed.

Cp. Appendix 2 in the Electronic Supplementary Material.

An expert is a person who possesses in-depth knowledge of a certain field (cp. Gläser and Laudel (2010), p. 9). In addition, experts should also be able to apply this knowledge to identify causes of problems and principles for solving them (cp. Pfadenhauer (2009), p. 452).

Source: own representation.

The translations of the German quotes into English that will be provided in the following have been provided by the author.

Cp. http://​www.​e-werkzeug.​eu.

Cp. Section 4.​4.​3.​2.​1.

Cp. Section 4.​2.​2.​4.

Source: own representation.

Cp. Section 4.​4.​3.​2.​2.

With the term “relation”, shipper 1 refers to transportation relations (cp. Section 2.​1.​2.​2).

Usually, offering a transport order to an unrestricted supplier base also implies that it is offered to a larger number of transportation service providers compared to using a restricted supplier base. The larger number of transportation service providers should also be associated with more possibilities to increase the capacity utilization of vehicles.

Both shippers 1 and 2 emphasized that an ETM may not only be used for transactions on the spot market but that it can also be used to support transactions between transportation service providers and shippers that are conducted outside of the spot market (e.g., with the functionality electronic transport order). This has been considered for the definition of the term ETM in the main study.

Shipper 3 and Shipper 4 represent the extreme points on the ETM use spectrum: shipper 3 uses the ETM to 100% for the procurement of transportation services on the spot market with an unrestricted supplier base, while shipper 4 does not use an ETM and only perceives a potential for using spot market functionalities of about 1%. Shipper 1 and shipper 2 use spot market and contract market functionalities on an ETM. The largest part of transportation services on the ETM is sourced via the functionality electronic transport order (shipper 1: 70–80%, shipper 2: 98–99%) and a smaller share is purchased on the spot market (shipper 1: 20–30%, shipper 2: 1–2%).

Source: own representation.

Shipper 3 reports that besides the use of spot market functionalities, transportation services are purchased based on contracts without the use of an ETM. These transportation services are performed for movements of goods to a warehouse in Austria which are executed with a fixed schedule. Similarly, shipper 1 reports that contracts are used for purchasing transportation services for planned movements to Italy or Spain. In contrast, the spot market is seen to be suitable when demand uncertainty is high. Shipper 2 reports that the spot market is primarily used for sporadic transportation activities into other EU countries. In addition, shipper 4 sees a large potential for using an ETM for purchasing transportation services on the spot market which are difficult to predict (e.g., because of demand peaks or product shortages).

Shipper 3 uses an ETM only for procurement on the spot market. At the same time, the transportation services do not involve specific assets because they deal with the movement of non-food products on standardized pallets. All other shippers use the spot market to a much lesser extent, or they do not use it at all. They also reported that the transportation services involve higher levels of asset specificity. For example, shipper 2 requires special vehicles which provide a high volume for voluminous products. In addition, the drivers which perform transports for shipper 4 must have training for dealing with hazardous goods and require knowledge of special loading and unloading procedures.

Shipper 2 and shipper 4 both mentioned that there are only a few transportation service providers which can offer most of their needed services. Since shipper 4 did not make any use of the spot market and shipper 2 only uses it to a very limited extent (1–2%), this can be interpreted as an indication for a positive effect of the availability of alternatives on the use of spot market functionalities (i.e., lower availability of alternatives is associated with lower use of spot market functionalities).

Only adopters of ETMs have been asked about the selection of ETMs because they have already made such a decision. Since the company of one of the experts has not adopted an ETM, only responses from three experts could be used.

Shipper 1 mentioned that acceptance in the transportation market is important so that sufficient transportation service providers are available. Shipper 2 reports that the ETM has been selected because it was an established marketplace that is used by other large shippers within the same industry. For shipper 3 it is important to get fast offers which will be more likely when there is high market liquidity.

Shipper 1 and 2 both emphasized that the fee levels are important when selecting ETMs. They also indicated that they were not satisfied with the fee levels of the currently adopted ETM by their companies which were perceived as being too high. For system quality, shipper 2 mentioned that the ease of use of the ETM was an important aspect during the selection of the ETM. In addition, shipper 3 emphasized that the ETM allows the entry of transport orders in a short time (e.g., by pre-filling data fields).

Source: own representation.

A similar procedure was chosen by Son and Benbasat (2007). Including both adopters and non-adopters of ETMs provides a larger target group for the survey than the focus on only adopters or non-adopters. In addition, some variables within the conceptual research framework are only relevant for either adopters (e.g., commitment to the ETM) or non-adopters (e.g., intention to adopt an ETM). Therefore, including both adopters and non-adopters allows to include a broader range of variables in the main study.

Cp. Dillman et al. (2014), pp. 32–33.

Information capability was not included as a success variable because it mainly depends on the use of the ETM functionality tracking and tracing.

Cp. Section 4.​3.​2.​2.​4 and Section 4.​3.​2.​2.​6.

The non-adopters of ETMs were asked to select the three most important selection criteria from the six contextual variables of the ETM properties which have been identified in the conceptual research framework. The rank order provided by the participants can help to find out which of the variables are particularly important from the point of view of shippers. The results are discussed in the descriptive statistics (cp. Section 5.2.4.2).

Source: own representation.

Cp. Cleff (2019), p. 260.

Cp. Lux et al. (2008), p. 178.

Source: own representation.

Cp. Section 5.2.1.3.

Cp. Section 5.2.4.2.

An alternative hypothesis is a hypothesis that is actually being tested. The null hypothesis is a competing hypothesis that states that the statements formulated in the alternative hypothesis are not true. Given an alternative hypothesis, the substantive statement of a null hypothesis does not provide any information (cp. Bortz (1999), p. 109). Therefore, only the alternative hypothesis will be formulated.

Cp. Section 4.​2.​2.​4. In contrast to the use of spot market functionalities on an ETM, the use of electronic transport orders must be based on agreements the are rather long-term than short-term oriented. In addition, the electronic transport order functionality focuses only on the settlement phase. Spot market functionalities, however, support the information and negotiation phase and can improve the freight rates obtained on the market. The fact that his opportunity is forgone when using electronic transport orders can be a sign of higher relational focus.

Cp. Jap and Haruvy (2008), p. 557; Stoll (2008), p. 239.

Cp. Pu et al. (2018), pp. 1700–1702; Son et al. (2005), pp. 338–346.

Asset specificity increases transaction costs which in turn tend to make hierarchies or hybrid governance structures like partnerships more efficient. In addition, specific investments create switching costs which make the party investing in specific assets more dependent (specific investments lose value when they must be redeployed to other uses). Therefore, transportation service providers who must invest, for example, into specific vehicles for a shipper might even make such investments only when shippers offer them a long-term relationship in return (cp. Section 4.​3.​2.​2.​1).

Cp. Anderson and Weitz (1992), pp. 25–28; Aral et al. (2018), pp. 606–607; Garbe (1998), pp. 164–170; Hallén et al. (1987), pp. 28–35; Hallén et al. (1993), pp. 76–79.

Cp. Pfeffer and Salancik (2003), p. 46.

Cp. Pfeffer and Salancik (2003), pp. 143–147; Ellis et al. (2010), p. 38.

Cp. Pfohl (2018), p. 292; Rümenapp (2002), p. 19.

For example, the fee levels are important for EM adoption (cp. Johnson (2010), pp. 163–167).

Haws and Bearden (2006), p. 304.

Cp. Maxwell (2002), pp. 192–194; Xia et al. (2004), pp. 6–7.

Cp. Martín‐Consuegra et al. (2007), p. 463; Herrmann et al. (2007), p. 56.

Cp. Doong et al. (2008), pp. 146–147; Jiang and Rosenbloom (2005), p. 164.

Cp. Flavián and Guinalíu (2006), p. 611; Kim and Ahn (2007), pp. 125–126; Koufaris and Hampton-Sosa (2004), pp. 386–388; Yoon (2002), pp. 58–60.

Cp. Pavlou (2002), p. 231; Pavlou and Gefen (2004), p. 50.

Cp. Moorman et al. (1992), pp. 320–322; Morgan and Hunt (1994), pp. 29–31.

Cp. Agag (2019), pp. 401–404; Kim and Ahn (2007), pp. 125–126.

Cp. Schüler (2021), p. 170; Shih et al. (2013), pp. 400–403.

Cp. Pavlou (2002), p. 231; Pavlou and Gefen (2004), p. 50; Shih et al. (2013), p. 402.

Cp. Kaufmann and Dant (1992), p. 173; Kaufmann and Stern (1988), p. 536.

Cp. Gulati (1995), p. 93; Leimeister (2010), p. 56.

Cp. Fink et al. (2006), p. 144.

For example, previous empirical studies have shown that the use of electronic reverse auctions can have detrimental effects on business relationships (cp. Jap (2007), pp. 154–157; Moramarco et al. (2013), pp. 57–70).

For example, such advantages due to high relational orientation are attributed to single sourcing (cp. Ramsay and Wilson (1990), p. 22).

Cp. Gulati (1995), p. 93; Pieper (2000), pp. 188–195.

Cp. Garrido et al. (2008), pp. 622–624; Giunipero et al. (2012), pp. 285–287; Kassim and Hussin (2013), pp. 10–13; Mishra et al. (2013), pp. 385–386; Ranganathan et al. (2011), pp. 539–541; Wu et al. (2007), pp. 582–583.

Cp. Yu et al. (2015), p. 1065.

Cp. Section 4.​4.​3.​3.​2.

Cp. Section 4.​3.​3.​1.​1.

Cp. Teo and Wong (1998), pp. 615–619.

Cp. Li et al. (2015), pp. 276–279.

Cp. Pavlou and Gefen (2004), p. 46.

Cp. Le et al. (2004), p. 299; Pavlou (2003), pp. 116–118; Pavlou and Gefen (2004), pp. 50–51; Rao et al. (2007), pp. 1041–1042.

Three hypotheses are differential hypotheses: H1, H2a and H3a.

Source: own representation.

Cp. Section 4.​2.​2.​4.

Cp. Winter (2013), p. 83.

Frequently recurring transactions make concluding long-term contracts a more profitable proposition (cp. Picot et al. (2001), p. 53). In particular, the use of contracts reduces efforts for recurring supplier identification and selection for single transport orders because this has only to be done once for the whole contract (cp. Section 2.​3.​3). Moreover, contracts for services with a low demand uncertainty provide benefits to transportation service providers due to a contractually guaranteed asset utilization (cp. Figiel (2016), p. 143). This should also be reflected in better conditions for shippers.

Cp. Hohberger (2001), p. 31; Malone et al. (1987), pp. 486–487.

Cp. Cannon and Perreault Jr (1999), p. 444.

Cp. Weber (2001), p. 179.

Cp. Moramarco et al. (2013), pp. 57–70.

Cp. Jap (2007), pp. 154–157; Moramarco et al. (2013), pp. 57–70.

Cp. Le et al. (2004), p. 299; Rao et al. (2007), pp. 1041–1042.

Cp. Mishra and Agarwal (2010), p. 255.

Cp. Sänger (2004), p. 118.

Cp. Mishra and Agarwal (2010), pp. 261–262.

Cp. H7 in Section 5.2.1.1.2.

Cp. Spinler and Huchzermeier (2005), p. 701.

Cp. Bickenbach et al. (2002), p. 189.

In the survey, non-adopters of ETMs had to select whether they perceive a higher potential of e-tendering or spot market functionalities on an ETM for their company.

Cp. H8 in Section 5.2.1.1.2.

Cp. Section 4.​4.​3.​2.​2.

Cp. Carter et al. (2004), p. 248.

H15, H16a, H17a, and H23a are differential hypotheses.

Source: own representation.

A reduction of freight rates by a certain percentage will be larger in absolute terms if the underlying purchasing volume is higher. Furthermore, savings in process costs are dependent on cost drivers like transport orders.

Cp. Hadaya (2006), p. 177; Johnson (2010), p. 160; Soares-Aguiar and Palma-dos-Reis (2008), p. 129; Sila (2013), p. 205; Zhu et al. (2003), p. 255.

Cp. Zhu et al. (2003), p. 255.

Cp. H26 in Section 5.2.1.2.2.

Cp. Möhring and Schlütz (2019), p. 115. In the first question, participants were asked if their company uses transportation service providers for transporting goods on the road.

Cp. Arbeitskreis Deutscher Markt- und Sozialforschungsinstitute e.V. (2001); Bandilla (2016); Dillman et al. (2014); Faulbaum et al. (2009); Häder (2019); Homburg and Klarmann (2004); Lenzner and Menold (2016); Menold and Bogner (2016); Möhring and Schlütz (2019); Porst (2014); Schnell (2019).

Cp. Häder (2019), p. 216; Möhring and Schlütz (2019), pp. 117–119; Porst (2014), pp. 114–116.

Common method bias exists if the measured variance is not caused by the construct but by the measurement method (cp. Podsakoff et al. (2003), p. 879).

Cp. Chang et al. (2010), p. 180.

Cp. Weiber and Mühlhaus (2014), p. 359.

An indicator (or item) is a directly measured observation (cp. Hair et al. (2017), p. 319). Within the survey, indicators have been used to measure the variables of the research models. Using multiple indicators should improve the psychometric quality of the measurement. The use of existing indicators does not only reduce efforts for creating own items but also improves the ability to compare empirical findings (cp. Homburg and Klarmann (2004), pp. 76–77).

Indicators can be formative or reflective. For formative indicators, the direction of causality is from the indicators to the investigated variable. Reflective indicators assume an opposite direction of causality, i.e. from the investigated variable to the indicators (cp. Jarvis et al. (2003), p. 203).

Rating scales indicate (by numbers, verbal descriptions, etc.) a continuum which respondents should use to mark their subjective perception of the rated object (cp. Bortz and Döring (2006), p. 177).

Cp. Menold and Bogner (2016), p. 2.

Strictly speaking, rating scales only provide ordinally scaled data (cp. Berekoven et al. (2009), p. 68). However, it is common to assume that participants perceive the scores on a rating scale to be equidistant which allows the interpretation as interval data (cp. Berekoven et al. (2009), p. 68; Döring and Bortz (2016), p. 251; Häder (2019), p. 98). To emphasize that the scales are equidistant, the rating scales used within the survey of this work were fully and equidistantly numbered.

When rating scales are used, either all or only the endpoints can be verbally labeled. The use of fully verbalized scales (i.e. all points are verbally labeled) can be especially beneficial for people with low and moderate formal education (cp. Menold and Bogner (2016), p. 3).

The answer options have been presented in a randomized way for three questions for non-adopters of ETMs. The first question with randomized answer option presentation asked participants how transport orders are transmitted (answer options: e-mail, phone, fax, EDI, others). The second question asked whether the participants see a larger potential for their company to use spot market functionalities or e-tendering on an ETM. The third question presented the six ETM properties identified in the conceptual research framework and asked the participants to select the three most important criteria for selecting an ETM. The randomized presentation of answer options has been selected for these questions because the answer options were not measured on a rating scale in contrast to most of the other questions of the survey.

Cp. Porst (2014), p. 138.

The German measures have been translated to English by the author of this thesis.

Cp. Appendix 3 in the Electronic Supplementary Material.

When a shipper uses multiple ETMs, the ETM which is most frequently used has been selected for the following questions in the main part.

Source: own representation.

The part which must be filled out was determined based on the following logic: if only one functionality is used, the corresponding part was determined. If multiple functionalities are used, spot market functionalities and electronic transport order had a higher priority for the determination of the part than e-tendering (because there should be the largest differences in relational orientation between these functionalities). If both spot market functionalities and electronic transport order are used by a shipper, the percentage of the total purchasing volume for these two functionalities had to be provided by the participants and the part for the functionality with the larger percentage was selected.

The text elements within the indicators which are specific for functionalities reminded the participants about the functionality which is currently being evaluated by them. This has been done to avoid that shippers that use multiple functionalities will evaluate the variables with the context of another functionality in mind. In the tables which depict the indicators of the variables, the text elements depending on the ETM functionality (spot market functionalities, electronic transport order, e-tendering) are represented in square brackets with the respective functionality part of the survey in italics.

Answer options of the scale: 1 = “strongly disagree”, 2 = “rather disagree”, 3 = “neither agree nor disagree”, 4 = “rather agree”, 5 = “strongly agree”.

The self-developed indicators reflect physical asset specificity (ASSET1), human asset specificity (ASSET2), and procedural asset specificity (ASSET3).

Source: own representation.

A similar approach for adding item values has been implemented for measuring procurement process performance (cp. Mishra et al. (2007), p. 112).

Source: own representation.

To elicit the effect of using the ETM on freight rates, it is important to define the reference point for which such an evaluation should be made. A promising reference point is a comparison against a manual process without the use of an IT application (cp. Schulze (2009), p. 108; Schumann (1992), p. 54). Therefore, the participants were asked to evaluate the effect of a transition from the status quo (using an ETM) to not using an ETM on the freight rates paid to transportation service providers.

Besides a decrease in freight rates due to the use of an ETM, the item FRATE1 allowed the participants to indicate that the use of an ETM increases freight rates (which should result in a decrease of freight rates without the use of the ETM). However, no respondent indicated that the use of an ETM increases freight rates. Therefore, all responses can be used to measure freight rate reductions.

Thus, FRATE represents an unweighted multiplicative index (cp. Döring and Bortz (2016), p. 281). The indicator FRATE3 was not considered for this index because it was expected that not all participants can or want to provide the information on a percentage estimation of the effect on freight rates. Therefore, the indicator was included purely for the descriptive analysis.

Source: own representation.

In particular, the following additional variables have been collected for adopters: number of ETMs currently used, names of ETMs in use, years of using the ETM, types of transports purchased on the ETM (e.g. CEP, LTL, etc.), knowledge about transportation services and relationship to transportation service providers, percentage of all transport orders for road transportation conducted via the ETM, and percentage of all transport orders conducted via the used functionalities (spot market functionalities, electronic transport order or e-tendering) on the ETM.

Like the survey part for adopters of ETMs, the separate survey parts were used to collect data that can be used to test differential hypotheses (e.g., for the differences in demand uncertainty between transportation services to be purchased via spot market functionalities or e-tendering).

Source: own representation.

Source: own representation.

The indicators of variables that are collected in the different parts for spot market functionalities and e-tendering contained text elements with different texts for the corresponding parts. These text elements should remind the survey participants of the kind of transportation service and ETM functionality for which they should provide the answers. In the tables summarizing the indicators of variables, these text elements are depicted in square brackets with the respective functionality part of the survey in italics.

Source: own representation.

Volume flexibility refers to the ability to effectively react to changes in customer demand. The responsiveness to target markets refers to the ability to respond to the needs of a firm’s target markets. Furthermore, access flexibility denotes the ability to provide widespread distribution coverage (cp. Vickery et al. (1999), pp. 17–19).

In particular, the following additional variables have been collected for non-adopters: the current method used to transfer transport orders (e.g., EDI, e-mail, etc.) and the types of transports that are purchased (e.g., CEP, LTL, etc.).

Source: own representation.

Cp. Section 2.​1.​1.

Source: own representation.

According to the definition in EU recommendation 2003/361, these smaller companies have less than 50 employees and their turnover or balance sheet total does not exceed a threshold value of 10 million Euros (cp. European Commission (2021)).

Although large and medium-sized companies only represent 2.9 % of the total number of companies, they account for 65.7 % of the total value added within Germany (cp. European Commission (2019)). Since a larger total value added should also be associated with larger movements of goods, it can be expected that larger companies also have higher transportation volumes.

Cp. https://​orbis.​bvdinfo.​com.

The business register system (code 52111) of Destatis provides a set of statics about companies in Germany (the statistics can be accessed on the internet via https://​www-genesis.​destatis.​de/​genesis/​online). The official statistics are provided based on the national WZ 2008 classification of economic activities (cp. Statistisches Bundesamt (2008)) which is based on the EU classification NACE Rev. 2. NACE is the statistical classification of economic activities in the European Community which imposes the use of a uniform classification system within all the countries of the EU (cp. Eurostat (2008), p. 5). Both of these classifications make a distinction between 21 sections on the highest level (cp. Eurostat (2008), p. 47; Statistisches Bundesamt (2008), pp. 52–53). Two out of these sections are relevant for this work, namely, section C for manufacturing companies and section G for wholesale and retail companies.

Cp. Statistisches Bundesamt (2017a), (2017b). The most recent data at the time of the data retrieval from Destatis stem from the year 2017.

These selection criteria do not completely represent the criteria for the identification of the population. For example, the criteria based on turnover and balance sheet total for the exclusion of small and micro companies are not included. This was necessary to be able to compare the Orbis data with the data from Destatis.

The NACE Rev. 2 divisions 10–33 represent companies in section C (manufacturing) and the divisions 45–47 comprise companies of section G (retail and wholesale trade) (cp. Eurostat (2008), p. 44).

The initial dataset fulfilled the specification of the population in terms of the dimensions “country” and “industry”. However, companies that did not fulfill the requirements of the dimensions “size” and “purchasing” still had to be removed.

Source: own representation.

Companies were excluded if no data on turnover or balance sheet total were available or if either the turnover or the balance sheet total did not fulfill or exceed the threshold value of 10 million Euros.

The companies have been excluded based on the 4-digit NACE Rev. 2 code in the Orbis database. First, companies with NACE Rev. 2 codes for repair and maintenance services (3300–3320 and 4520) have been excluded. Second, further companies have been excluded because it is unlikely that these buy transportation services: car dealers (4500–4519), single stores for electrical goods (4643, 4743,4750,4754), and single supermarkets (4711). Especially for single stores of market chains for electrical goods or supermarkets, it is more likely that transportation services are purchased and coordinated centrally and not within the single stores. Therefore, companies with the corresponding NACE Rev. 2 codes have been excluded.

Cp. Bortz and Döring (2006), p. 32.

For an overview of different sampling methods, cp. Löffler and Heyde (2014), pp. 19–24.

It is common to use randomly generated numbers for creating a random sample (cp. Bortz and Döring (2006), p. 400).

The industry classification of companies is based on the intermediate aggregation A*38 developed by the System of National Accounts (SNA) (cp. Eurostat (2008), pp. 43–44).

Source: own representation.

Cp. Schnell (2019), pp. 123–144.

The participants consisted of two researchers of the HHL Leipzig Graduate School of Management who had a background in logistics, six practitioners representing the target group of the survey (working in logistics functions or functions related to logistics within companies), and three persons with university degrees but without experience in logistics.

Initially, it was planned to ask participants for responses for multiple functionalities if they are used in combination on the ETM. However, to reduce the burden of participation and survey length, it has been decided to focus only on the functionality with the highest use in terms of the percentage of transport orders.

A few participants of the pretest indicated that some pages require a lot of scrolling and reading which was perceived as negative. For some pages, four to six questions were asked on one page. When multiple questions are combined on one page, respondents must scroll through many questions which increases the chances that questions will be missed. In addition, such a design can be difficult when the survey is completed on a mobile phone. However, surveys that are constructed with one question per page often take longer to complete and participants may have difficulty remembering the context established by previous questions (cp. Dillman et al. (2014), pp. 311–315). As a compromise, the survey was therefore designed in a way that a maximum of two to three questions were presented on one page. This number of questions per page does not require extensive scrolling efforts by participants. At the same time, fewer pages are needed in comparison to the option with one question per page.

Cp. Berekoven et al. (2009), p. 107. Multiple studies have also shown that web surveys entail at least the same validity, reliability, and objectivity as traditional paper-based surveys (cp. Grether (2003), pp. 211–215).

Cp. http://​www.​limesurvey.​org.

The Orbis database contains decision-makers from different fields and one company can have multiple decision-makers. For a single company, the decision-makers have been identified from the different fields in the following order (from highest priority to lowest priority): (1) logistics, (2) IT, (3) general management.

To be able to provide the needed information of the survey, subjects within the relevant companies should possess two main characteristics. First, they should have knowledge about the transportation services which are relevant for their company and how these are purchased. Second, they should be able to provide information on the IT systems used for purchasing transportation services and they should be able to evaluate the potential use of such systems. Since it was expected that not all the invited contact persons may possess these capabilities, they were asked to forward the invitation to other employees if they are more capable to answer the survey questions.

Cp. Ernst (2001), pp. 87–96.

Cp. Bagozzi et al. (1991), pp. 424–426; Ernst (2003), p. 1250.

The access key was embedded in a link in the invitation e-mail which allowed the participants to directly start the survey.

Several recommendations for the design of the survey instrument have been followed. In particular, attention was paid to asking interesting questions (cp. Dillman et al. (2014), pp. 28–29), keeping the survey as short as possible (cp. Keusch (2015), p. 189) and limiting the length and complexity of questions (cp. Dillman et al. (2014), pp. 32–34). The invitation e-mail was kept as short as possible (cp. Schulte (2018), p. 57). In addition, the potential participants were asked for help or advice to encourage their willingness to help and increase their motivation for participation (cp. Möhring and Schlütz (2019), p. 44). Furthermore, the HHL Leipzig Graduate School of Management was mentioned as a legitimate organization which should increase trust (cp. Dillman et al. (2014), p. 29) and it was specified how the survey results will be useful to increase the intrinsic motivation of potential participants (cp. Theobald (2003), p. 406). Finally, a donation of 2 euros for a good cause per completed survey as well as the option to participate in a lottery where the participants can win wine bottles was offered as an incentive for participation.

During the phone calls, the employees of the contacted companies were asked for the person or department responsible for the purchasing of transportation services. Besides having the needed knowledge, these employees should also have the largest interest in the survey. As a result of the phone calls, 396 additional e-mail invitations have been sent whereby 53 resulted in completed surveys.

The adjusted sample takes the invalid e-mail addresses into account and consists of 1,500–47 = 1,453 companies.

Cp. Völker and Neu (2008), p. 118.

Source: own representation.

One of the responses indicated that neither spot market functionalities, e-tendering nor the functionality electronic transport order is being used on the ETM (answer ID 326). Therefore, this response cannot be used for the analysis of the variables relational orientation, asset specificity, and importance which were gathered in the corresponding parts of the survey. The other two responses could not be used because the respondents named booking portals of single transportation service providers as the names of the used ETMs (answer IDs 135 and 152). Because these booking portals are not ETMs, these responses were excluded.

Three responses of non-adopters were excluded based on the time they needed for completing the survey. One response was completed in below 5 minutes (answer ID 95). Since it cannot be expected that the survey is thoroughly processed in such a short time, the response has been excluded. In addition, two responses were completed after more than 45 minutes (answer IDs 94 and 282). For such a long processing time it is assumed that respondents were interrupted while completing the survey and did not finish it in a concentrated manner as the other participants.

Source: own representation.

This applies to one indicator for the adopters of ETMs (CMMT3) and four indicators for the non-adopters (ADOP3, UNCT1, UNCT2, and UNCT4).

Calculations were only needed for two indicators within the survey part for adopters of ETMs (BREADTH and FRATE).

Cp. Armstrong and Overton (1977), p. 396.

Cp. Fawcett et al. (2014), p. 11.

Source: own representation.

Cp. Podsakoff et al. (2003), p. 879.

Cp. Section 5.2.2.

Cp. Podsakoff et al. (2003), p. 889.

Cp. Appendix 4 and Appendix 5 in the Electronic Supplementary Material.

The industry of companies is reported based on the system of national accounts (SNA) which has developed a so-called “intermediate aggregation” consisting of 38 categories for structuring the economic activities of companies (abbreviated as SNA A*38 code). The SNA A*38 code has not been retrieved directly from the Orbis database. Instead, the NACE Rev. 2 divisions retrieved from Orbis have been matched with the corresponding SNA A*38 codes (cp. Eurostat (2008), pp. 43–44).

Source: own representation.

This is also reflected in the self-assessment of the knowledge of adopters: most adopters indicate that their knowledge of the transportation services of their company is rather good to very good (18 of 22 adopters) and that they have rather good to very good knowledge of the business relationships with transportation service providers (16 of 22 adopters).

Source: own representation.

15 non-adopters of ETMs replied that they see no potential at all to use an ETM. Since these respondents have not been asked to assess the relational orientation toward transportation service providers on an ETM, the remaining 120 responses for the variable relational orientation were evaluated (22 responses of adopters and 98 responses of non-adopters).

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Dependence techniques are used to predict or explain one or many dependent variables by independent variables (cp. Hair et al. (2019), p. 21).

In contrast, so-called first-generation dependence techniques (e.g. multiple regression) can only analyze one layer of linkages between independent and dependent variables at a time (cp. Gefen et al. (2000); Hair et al. (2019), pp. 21–24).

Cp. Backhaus et al. (2016), p. 19; Hair et al. (2011), p. 139; Hair et al. (2019), p. 608.

Source: adapted from Hair et al. (2017), p. 12.

Cp. Hair et al. (2017), p. 314.

Cp. ibid., p. 11.

Error terms are relevant for endogenous constructs (Z₃ and Z₄) and reflective indicators (e₇–e₉).

Cp. Hair et al. (2011), p. 141.

Cp. Hair et al. (2019), p. 14.

Cp. Bagozzi and Yi (1994), p. 19.

Cp. results of the normal distribution tests in Appendix 6 and Appendix 7 in the Electronic Supplementary Material.

SmartPLS v. 3.3.3

Cp. Hair et al. (2017), p. 316.

Cp. Hair et al. (2017), pp. 326–330.

Cp. Huber et al. (2007), pp. 34–39.

Cp. Hair et al. (2017), pp. 104–136.

Cp. ibid., p. 320.

Cp. Henseler et al. (2009), p. 299; Hair et al. (2017), p. 111. Nonetheless, it has been recommended to report the values for Cronbach’s alpha as the lower bound for internal consistency reliability (cp. Hair et al. (2017), p. 112). Therefore, Cronbach’s alpha will also be reported for the results of the reflective measurement model evaluation in this thesis.

Cp. Hair et al. (2011), p. 145; Hair et al. (2012), p. 429.

Cp. Hair et al. (2017), pp. 113–114.

Cp. Fornell and Larcker (1981), p. 46; Hair et al. (2017), p. 115.

Cp. Henseler et al. (2015), 115–135.

Cp. Hair et al. (2017), pp. 119–120.

Source: own representation.

Cp. Hair et al. (2017), p. 191.

Cp. Huber et al. (2007), p. 108.

Cp. Hair et al. (2011), pp. 146–147.

Cp. Henseler et al. (2009), pp. 303–304.

Cp. Hair et al. (2017), pp. 195–196.

Cp. Hair et al. (2017), p. 198.

Cp. Chin (1998), p. 323.

Cp. Chin (2010), p. 675.

Cp. Cohen (1988), pp. 413–414.

Cp. Hair et al. (2017), p. 202.

Cp. Chin (1998), p. 318.

Source: own representation.

Cp. Janssen and Laatz (2007), pp. 334–335.

Since the respondents of the survey of this work provide data only once and are independent of each other, no dependent samples are obtained. However, the respondents belong to certain groups (e.g., adopters and non-adopters of ETMs) which represent independent samples.

Cp. Cleff (2019), p. 262; Sarstedt and Mooi (2019), p. 161; Thulasingam and Premarajan (2018), p. 123.

The Shapiro-Wilk test can be used to test the normal distribution of data, whereby a statistically significant result indicates that the data are not normally distributed (cp. Janssen and Laatz (2017), p. 249). The results of the normal distribution tests which have been performed for the variables which are subject of differential hypotheses indicate that they have a non-normal distribution (cp. Appendix 6 and Appendix 7 in the Electronic Supplementary Material).

Cp. Sarstedt and Mooi (2019), p. 164.

Cp. Toutenburg and Heumann (2008), p. 174.

Cp. Schäfer (2011), p. 144.

Cp. Rasch et al. (2014), p. 100. One-tailed tests are performed for directed hypotheses, two-tailed tests for undirected ones (cp. Bortz (1999), p. 116). The results for a one-tailed test can be retrieved by splitting the significance value for a two-tailed test in half (cp. Janssen and Laatz (2017), p. 656).

Cp. Eckstein (2016), pp. 133–135. SPSS provides an asymptotic and exact significance. The asymptotic significance is based on theoretical distributions and is only an approximation. The larger the sample size, the better this approximation. For small or unbalanced samples, tests for the exact significance should be performed. These are not based on theoretical distributions but on the probability distributions of the variables which are specifically calculated for a given sample (cp. Janssen and Laatz (2007), p. 813). The use of the exact significance is recommended when the total sample size of both groups is below 30 (cp. Universität Zürich (2020)).

Cp. Universität Zürich (2020).

Cp. Cohen (1988), pp. 79–80.

Cp. Hair et al. (2017), p. 23.

For example, this sample size is still sufficient to detect a minimum R² value of 0.5 for a significance level of 0.01 and for a maximum of 3 arrows pointing at a construct (cp. Hair et al. (2017), p. 26).

For example, the study of Majchrzak et al. (2005) had a sample size of 17 and was published in a highly ranked journal.

The quality criteria can only be reasonably evaluated for constructs that have been measured with more than one indicator. Therefore, the table does not include the quality for criteria of the constructs which have been measured with a single indicator (i.e., breadth of ETM use, freight rate reductions, business process performance, and fairness of ETM fees).

Cp. Section 5.2.5.1.1.

The indicators RISK3, CMMT3, and FEES2 had to be removed. Since the construct fairness of ETM fees has been measured by two indicators, the elimination of the indicator FEES2 resulted in the operationalization of this construct with a single item. Given the simple structure of the construct, this was not considered to be problematic for the measurement model.

It is recommended that only indicators with a value below 0.40 should be directly eliminated. Indicators with outer loadings between 0.40 and 0.70 should be closely examined and only eliminated when this helps to reach the suggested threshold values for composite reliability or AVE (cp. Section 5.2.5.1.1). Since the threshold values for composite reliability and AVE already exceed the threshold values, it was decided to retain the indicator SECU4.

Source: own representation.

Source: own representation.

Cp. Section 5.2.5.1.2.

Source: own representation.

Source: own representation. The results are based on a bootstrapping procedure with 5,000 bootstrapping samples and a blindfolding procedure with an omission distance of 7.

Source: own representation.

Three companies use an ETM mainly for e-tendering. These are not considered for the group comparisons via Mann-Whitney U tests.

For a total sample size below 30, exact significances should be reported and considered instead of the asymptotic significance (cp. Janssen and Laatz (2007), p. 797; Universität Zürich (2020)).

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

The indicators UNCT3, CMPL1, and SOL2 have been eliminated.

The outer loadings for the indicators UNCT4 (0.661), SOL1 (0.682), and RISK1 (0.637) do not exceed 0.70. However, the outer loadings of these indicators are still above a threshold value of 0.40 and the constructs exceed the threshold values for composite reliability and AVE. Therefore, it was decided to not eliminate these indicators.

Source: own representation.

Source: own representation.

Source: own representation.

While the path coefficient from asset specificity to the depth of ETM use is statistically significant (P < 0.05), the exogenous variable cannot explain much of the variance of the endogenous variable depth of ETM use (R² = 0.038). Therefore, H18 is only weakly supported. H24 is only weakly supported because the statistical significance of the path coefficient from relational orientation to freight rate reductions is not strong (P < 0.1).

Source: own representation. The results are based on a bootstrapping procedure with 5,000 bootstrapping samples and a blindfolding procedure with an omission distance of 7.

Source: own representation.

It was not expected to get a valid evaluation of the potential relational orientation toward transportation service providers on an ETM when the respondents see no potential at all to use an ETM.

Source: own representation.

Source: own representation.

P < 0.001 for the indicators CMPL1, CMPL2 and CMPL4 and P < 0.01 for CMPL3.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.

Source: own representation.