Key Criteria in the Choice of IoT Platforms in Spanish Companies

: The level of implementation of IoT technology in Spanish companies is not low for a novel technology, but there are concerns that hinder its level of implementation from being higher: initial costs, security, and privacy. These concerns could be solved with the choice of the appropriate IoT platform, but it must be taken into account that there are numerous criteria that participate in this decision that inﬂuence each other. This could give rise to criteria not being considered at ﬁrst, even though they should be taken into account in this choice. Existing studies do not consider this. Thus, the objective of this study was to ﬁnd out what are the key criteria to be taken into account by Spanish companies when choosing an IoT platform. The criteria were located through a literary review, and classiﬁed according to their level of importance in this decision through a causal study carried out with the methodology of fuzzy cognitive maps. The results achieved showed that there were key criteria beyond the criteria directly implied by the concerns, such as stability, scalability, ﬂexibility, and interoperability, and these new criteria acquired the same or greater importance.


Introduction
The Internet of Things (IoT) is expected to revolutionize the business sector as it develops as a technology (providing new solutions) and as companies bet on its implementation. It is said that the revolution of the arrival of the IoT to companies is and will be similar to the revolution caused by the arrival of computers in that sector, as its arrival will change the way in which companies do business, since it will create new ways of communication between companies with customers, and increase companies' productivity and efficiency [1][2][3][4][5][6].
The IoT is a technology that allows the interconnection between devices in order to collect, transmit, and store data. Its value in the business sector goes beyond the sensors and devices by which most users know this technology; its true value arises from the data it collects and the use of that data through different applications [7].
For companies, this technology can have the following advantages [8][9][10][11]: -Savings in energy consumption, thanks to intelligent infrastructures and their sensors. -A reduction in costs, since it can predict failures in machinery and production lines, eliminating downtime caused by these failures, thanks to the possibility of monitoring these machines and production lines. -Greater production efficiency, since it allows evaluation of demand in real time. -Better use of changing markets, since it allows improvements in strategies thanks to the collected data to be marked. -A greater understanding of customers, in terms of their behavior. The collected data make it possible to identify preferences and trends among customers, which can help companies to provide more personalized services and products to their customers and detect business opportunities.
Hybrid Cloud: Capacity to adapt to existing technological systems hosted on company premises. This makes it possible for the most critical processes for the business to be carried out locally, while other processes are carried out through the platform [19,33,34].
Platform Migration: Capacity to provide the necessary tools for any possible migration to other IoT platforms [17,19].
Previous Experience: Previous experience of the supplier in similar jobs to those required by the company, within the same scope [20].
Edge Intelligence and Control: Capacity of the IoT platform to support new topologies and use edge intelligence [35], as the future of these platforms is moving towards distributed, offline, and edge intelligence [36], and decisions need to be made based on local data (in addition to cloud data).
Bandwidth: Capacity to provide high bandwidth, since the transmission of information between the processing components requires efficient high bandwidth communication [19,37].

Objectives and Hypothesis
After this analysis, it is necessary to emphasize again that the main objective of this research is to know what the key criteria are to be taken into account by Spanish companies when choosing an IoT platform. This will facilitate this selection process for future Spanish companies that want to implement this technology and may encourage other companies that did not have this technology among their plans to adopt it.
For this purpose, we formulated the following hypotheses that must be ratified or rectified in the course of this study: Hypothesis 1. The difficulties or concerns that Spanish companies encounter when betting on the implementation of the IoT are not the only criteria to take into account when choosing an IoT platform.

Hypothesis 2.
The difficulties or concerns that Spanish companies encounter when betting on the implementation of the IoT are not the most relevant criteria or key criteria in the choice of IoT platforms, since there are other criteria to take into account when choosing IoT platforms with equal or greater importance.

Methodology
Thus, the next steps in this study were to confirm whether all the criteria found in the literary review should be taken into account in the choice of the IoT platform and to classify these criteria according to their level of importance in this decision.
These steps were focused on Spanish companies, and made it possible to identify the key criteria in the choice of IoT platforms in Spanish companies thanks to the methodology of the fuzzy cognitive maps. According to each country, companies will have different concerns or difficulties when implementing this technology (depending on the countries' technological, regulatory, and/or economic level); thus, companies in each country will have different key criteria.
The methodology of fuzzy cognitive maps is a methodology used to identify causeeffect relationships between the elements involved in decision making [38][39][40]; it is based on the judgment of experts (it uses expert knowledge, obtained through interviews, as a source of information for the detection of cause-effect relationships between the elements involved in decision making, and as a source of information for the assessment of these cause-effect relationships based on the degree of influence that each element involved exerts on the other elements) [41] and allows (with the use of the appropriate software) to classify these elements according to their level of importance in the decision making, basing their level of importance on the degree of influence exercised and received by each elements involved in the decision making [42].

Process
In each interview, each expert confirmed whether the criteria detected in the literary review were all the criteria involved in the choice of an IoT platform by Spanish companies, and indicated whether any criteria had to be eliminated and/or added. For this, as observed in other studies that used this same methodology [43], each expert received a table with these criteria and their definitions (these definitions can also be seen in the section "literature review" above).
In addition, each expert had to assess the level of influence existing between the criteria involved in this choice (i.e., the cause-effect relationships) with an assessment comprised of values within the interval [−1,1]: considering the value of −1 when a criterion exerts a strong negative influence on another criterion, considering the value of 0 when a criterion does not exert influence on another criterion, and considering the value of 1 when a criterion exerts a strong positive influence on another criterion [44,45].
Thus, an adjacency matrix (such as the one shown in Table 1) was filled in for each expert; each criterion involved received a row and column in the matrix. This adjacency matrix had to be filled in with the evaluations provided by each of the interviewees regarding the influences that each criterion in the rows exerted on each of the criteria in the columns. For example, based on what can be seen in Table 1, the first row had to be filled in with the evaluations provided by each of the interviewees regarding the influences exerted by the scalability criterion on each of the criteria of the columns: scalability, flexibility, data analytics and visualization tools, redundancy and disaster recovery, stability, security, data ownership, protocol, system performance, time to market, extent of legacy architecture, attractive interface, pricing model and business case, ownership of cloud infrastructure, interoperability, application environment, hybrid cloud, platform migration, previous experience, edge intelligence and control, and bandwidth. The same had to be done with the other criteria that appear in the other rows.
It should be borne in mind that the level of influence that a criterion exerts on another criterion does not have to be the same level of influence that it receives from that criterion. It is even possible to find a criterion that influences another criterion but is not influenced by that criterion, and vice versa.
In this case, criteria were not eliminated from those provided in the literary review, since the experts believed that all of the criteria presented were involved in this decision making, and no new criteria were added by the experts. Therefore, the adjacency matrix to be filled in by each of the experts was the one found in Table 1.
Once an adjacency matrix was obtained by each of the experts, a global adjacency matrix had to be obtained (which was formed by the average evaluations of these experts). This global adjacency matrix was introduced to the FCMappers software, which analyzed these assessments and classified the criteria according to their level of importance in the choice of an IoT platform by companies, basing their level of importance on the degree of influence exerted and received among the criteria involved in this choice [46].
FCMappers was first released as a beta version in early July 2009, and is the first fuzzy cognitive map analysis tool available based on Microsoft Excel. For its operation, only the global adjacency matrix achieved with the use of the fuzzy cognitive maps methodology must be entered, and later this software automatically classifies the criteria by the degree of influence they exert on each other, by the degree of influences they receive from the other criteria, and by the degree of participation or influence in decision making.

Participants
The choice of experts is key in studies that use this type of methodology to add significance to the results achieved [47], and thus in this study, two different profiles were chosen: -Researchers in technological implementation who are university professors with training in technologies and knowledge of how to apply technologies in a useful and effective way in companies, have a doctorate, and among their lines of research are technologies in companies. The experts questioned in this study are all teachers of business technology subjects in the areas of Business Administration, Organization and Management, with expertise in the use of these technologies from a business point of view, and have more than 10 years of teaching and research experience. These researchers all belong to different Spanish universities, and only one renowned researcher per university was interviewed. -Professionals responsible for Information Technology (IT) in Spanish companies that have implemented IoT platforms in their companies, without taking into account the sector to which these companies belong (since this study is an analysis of the Spanish business sector in general) and without taking into account the size of these companies (since the study aims to make a generalized analysis, although most of the companies that have implemented these technologies in Spain can be considered large or medium-large companies, depending on their billing level and the number of their employees).
The opinion of both profiles received the same weight in the global adjacency matrix, since the opinion of both is equally important due to the two very different points of view that they contributed: one profile contributed the scientific-theoretical point of view in the implementation of this technology (in addition to their experience and knowledge in similar studies to the current one) and the other profile contributed a real-world point of view of the Spanish business sector when implementing this technology. The mean of both points of view was considered the most significant, since no great differences were found Appl. Sci. 2021, 11, 10456 7 of 14 between the evaluations provided by the experts of the two types of profiles (this helped verify the significance of the results).
Regarding the number of participants in studies that use this type of methodology, this is determined by the appearance of new criteria in each of the interviews that are carried out; thus, the number of participants is considered optimal when the number of criteria involved stabilizes as interviews are conducted and when new criteria no longer appear as interviews are conducted [48]. This means that there are studies with this type of methodology with 45 [49], 41 [50], 40 [51], 29 [49], 8 [52][53][54][55], 7 [43], and 4 participants [56].
The number of participants in this study was set at 40 (20 researchers in technological implementation and 20 IT managers in Spanish companies), since none of the participants added new criteria in the interviews (Figure 1). This number of participants could have been lower, although it was decided to reach this number so that the results achieved would be the most significant possible.

Results
This section shows the results after applying the methodology of fuzzy cognitive maps and the subsequent use of the FCMappers software, showing (in the global adjacency matrix) the average valuations of the existing influences between each of these criteria and classifying the criteria by the degree of influence they exert on the other criteria, by the degree of influence they receive from the other criteria, and by the degree of participation or influence in the choice of IoT platforms in Spanish companies (which is calculated in each of the criteria as the sum of the degree of influence it exerts on the other criteria and the degree of influence it receives from the other criteria) [57].

Global Adjacency Matrix
As mentioned in the previous section, each interview generated an adjacency matrix with the information provided by each expert on the level of influence existing between the criteria and, subsequently, a global adjacency matrix was created with the average evaluations of these experts (obtaining the existence of 248 cause-effect relationships between the criteria involved in this decision) ( Table 1).
This global adjacency matrix, reached with the use of the FCMappers software, allowed the criteria to be classified according to their level of importance in the choice of an IoT platform in companies, basing their level of importance on the degree of influence exerted and received among the criteria involved in this selection.

Criteria Most Influencing Other Criteria
After the analysis of the global adjacency matrix with the FCMappers software, it was identified that the criteria that most influence the rest of the criteria are (from greatest to least influence): previous experience, data ownership, stability, pricing model and business case, protocol, flexibility, interoperability, security, scalability, edge intelligence and control, system performance, data analytics and visualization tools, hybrid cloud, application environment, redundancy and disaster recovery, extent of legacy architecture, ownership of cloud infrastructure, bandwidth, attractive interface, platform migration,

Results
This section shows the results after applying the methodology of fuzzy cognitive maps and the subsequent use of the FCMappers software, showing (in the global adjacency matrix) the average valuations of the existing influences between each of these criteria and classifying the criteria by the degree of influence they exert on the other criteria, by the degree of influence they receive from the other criteria, and by the degree of participation or influence in the choice of IoT platforms in Spanish companies (which is calculated in each of the criteria as the sum of the degree of influence it exerts on the other criteria and the degree of influence it receives from the other criteria) [57].

Global Adjacency Matrix
As mentioned in the previous section, each interview generated an adjacency matrix with the information provided by each expert on the level of influence existing between the criteria and, subsequently, a global adjacency matrix was created with the average evaluations of these experts (obtaining the existence of 248 cause-effect relationships between the criteria involved in this decision) ( Table 1).
This global adjacency matrix, reached with the use of the FCMappers software, allowed the criteria to be classified according to their level of importance in the choice of an IoT platform in companies, basing their level of importance on the degree of influence exerted and received among the criteria involved in this selection.

Criteria Most Influencing Other Criteria
After the analysis of the global adjacency matrix with the FCMappers software, it was identified that the criteria that most influence the rest of the criteria are (from greatest to least influence): previous experience, data ownership, stability, pricing model and business case, protocol, flexibility, interoperability, security, scalability, edge intelligence and control, system performance, data analytics and visualization tools, hybrid cloud, application environment, redundancy and disaster recovery, extent of legacy architecture, ownership of cloud infrastructure, bandwidth, attractive interface, platform migration, and time to market (Figure 2). These evaluations were provided by the software based on the expert evaluations regarding the level of influence existing between the criteria involved in this choice. The criterion with the highest value is previous experience (18.00) and the criterion with the lowest value is time to market (1.00); between them, 10 criteria are equal to or above the average of these two criteria (9.50), and the other nine are below.

Criteria That Receive the Most Influence from Other Criteria
After the analysis of the global adjacency matrix with the FCMappers software, it was identified that the criteria that receive the most influence from other criteria are (from highest to least received influence): scalability, stability, security, flexibility, redundancy and disaster recovery, hybrid cloud, interoperability, pricing model and business case, extent of legacy architecture, data analytics and visualization tools, edge intelligence and control, application environment, ownership of cloud infrastructure, system performance, time to market, protocol, attractive interface, platform migration, bandwidth, data ownership, and previous experience (Figure 3).  These evaluations were provided by the software based on the expert evaluations regarding the level of influence existing between the criteria involved in this choice. The criterion with the highest value is previous experience (18.00) and the criterion with the lowest value is time to market (1.00); between them, 10 criteria are equal to or above the average of these two criteria (9.50), and the other nine are below.

Criteria That Receive the Most Influence from Other Criteria
After the analysis of the global adjacency matrix with the FCMappers software, it was identified that the criteria that receive the most influence from other criteria are (from highest to least received influence): scalability, stability, security, flexibility, redundancy and disaster recovery, hybrid cloud, interoperability, pricing model and business case, extent of legacy architecture, data analytics and visualization tools, edge intelligence and control, application environment, ownership of cloud infrastructure, system performance, time to market, protocol, attractive interface, platform migration, bandwidth, data ownership, and previous experience (Figure 3). These evaluations were provided by the software based on the expert evaluations regarding the level of influence existing between the criteria involved in this choice. The criterion with the highest value is scalability (18.60) and the criterion with the lowest value is previous experience (0.00); between them, 13 criteria are above the mean of these two criteria (9.30), and six are below.
was identified that the criteria that receive the most influence from other criteria are (from highest to least received influence): scalability, stability, security, flexibility, redundancy and disaster recovery, hybrid cloud, interoperability, pricing model and business case, extent of legacy architecture, data analytics and visualization tools, edge intelligence and control, application environment, ownership of cloud infrastructure, system performance, time to market, protocol, attractive interface, platform migration, bandwidth, data ownership, and previous experience (Figure 3).

Key Criteria in the Choice of IoT Platforms in Spanish Companies
After the analysis of the global adjacency matrix with the FCMappers software, it was identified that the key criteria in the choice of IoT platforms in Spanish companies are (from highest to lowest importance): stability, scalability, security, flexibility, pricing model and business case, interoperability, hybrid cloud, edge intelligence and control, redundancy and disaster recovery, data analytics and visualization tools, system performance, application environment, protocol, previous experience, extent of legacy architecture, ownership of cloud infrastructure, data ownership, time to market, attractive interface, bandwidth, and platform migration (Figure 4). These evaluations were provided by the software based on the expert evaluations regarding the level of influence existing between the criteria involved in this choice. The criterion with the highest value is scalability (18.60) and the criterion with the lowest value is previous experience (0.00); between them, 13 criteria are above the mean of these two criteria (9.30), and six are below.

Key Criteria in the Choice of IoT Platforms in Spanish Companies
After the analysis of the global adjacency matrix with the FCMappers software, it was identified that the key criteria in the choice of IoT platforms in Spanish companies are (from highest to lowest importance): stability, scalability, security, flexibility, pricing model and business case, interoperability, hybrid cloud, edge intelligence and control, redundancy and disaster recovery, data analytics and visualization tools, system performance, application environment, protocol, previous experience, extent of legacy architecture, ownership of cloud infrastructure, data ownership, time to market, attractive interface, bandwidth, and platform migration (Figure 4). As U. Ö zesmi and S. L. Ö zesmi [48] say, the key criteria in a decision are those that have the greatest degree of participation or influence in decision making. In this case, this degree of participation is determined by the influences exerted and the influences received.
Thus, this classification was obtained by the software based on the experts' evaluations. The criterion with the highest value is stability (29.65) and the criterion with the lowest value is platform migration (5.80); between them, 13 criteria are above the mean of these two criteria (17.725), and six are below. As U. Özesmi and S. L. Özesmi [48] say, the key criteria in a decision are those that have the greatest degree of participation or influence in decision making. In this case, this degree of participation is determined by the influences exerted and the influences received.

Discussion
Thus, this classification was obtained by the software based on the experts' evaluations. The criterion with the highest value is stability (29.65) and the criterion with the lowest value is platform migration (5.80); between them, 13 criteria are above the mean of these two criteria (17.725), and six are below.

Discussion
The business sector has previously gone through great revolutions caused by technological evolution and had to adapt to these revolutions in order to survive and to achieve new opportunities. Currently, a new revolution is taking place with the emergence of the IoT and its opportunities for use in companies, which affects the way in which data are collected and the way companies use and take advantage of these data.
This revolution can be observed in the global increase in companies that are beginning to use IoT in their operations and services [12], which is also reflected in the Spanish business sector [13].
However, although the level of implementation of this technology in Spanish companies is not low for a novel technology, it is true that there are a number of concerns expressed by this sector [12] that make its level of implementation not much higher. These concerns affect the choice of the IoT platform (the software necessary for the use of the IoT).
Today, the number of IoT platforms on the market is large and choosing one of them becomes a complicated process when a company decides to implement this technology. For this reason, this study located the criteria to be taken into account by Spanish companies when choosing an IoT platform (criteria that some platforms met and others did not) and classified them according to their level of importance in this decision (in order to know which of them companies should pay more attention to when choosing an IoT platform, since not all platforms comply with all criteria and this was a very difficult process for companies, which caused reluctance when implementing this technology).
Thus, at this point and by way of comparison and discussion of the results achieved in this study with respect to the results of other studies analyzed in the introduction and literature review sections, this section is developed based on the two hypotheses raised: Hypothesis 1. The difficulties or concerns that Spanish companies encounter when betting on the implementation of the IoT are not the only criteria to take into account when choosing an IoT platform.
The difficulties or concerns that Spanish companies encounter when betting on the implementation of the IoT are the initial costs, security, and privacy [12], but these should not be the only criteria to be taken into account by Spanish companies in the choice of the IoT platform.
This was already analyzed in the study by Ullah, Nardelli, Wolff, and Smolander [16] for companies in general worldwide, where 21 criteria were found: scalability, flexibility, data analytics and visualization tools, redundancy and disaster recovery, stability, security, data ownership, protocol, system performance, time to market, extent of legacy architecture, attractive interface, pricing model and business case, ownership of cloud infrastructure, interoperability, application environment, hybrid cloud, platform migration, previous experience, edge intelligence and control, and bandwidth.
In this study, all these criteria were ratified by the experts interviewed as the criteria to be taken into account by Spanish companies when choosing an IoT platform.
Before these two studies, as can be seen in the literary review, there was research that involved the participation of some of these criteria in this decision, but none grouped all the criteria.

Hypothesis 2.
The difficulties or concerns that Spanish companies encounter when betting on the implementation of the IoT are not the most relevant criteria or key criteria in the choice of IoT platforms, since there are other criteria to take into account when choosing IoT platforms with equal or greater importance.
The existing research on this problem discussed in the literary review reflects that these investigations tried to find out key criteria by independently assessing the degree of influence of each of these criteria on this decision and without taking into account all of these criteria and that the criteria that influence this decision influence each other, which can cause changes in the relevance of each of these in decision making [47]. An example of this can be seen in the study by Gallego Gómez [12], a study focused on the field of Spanish companies and the problems studied here.
This has been taken into account in this study, meaning that criteria that do not seem relevant or worrying at first to Spanish companies could be and should be taken into account in this choice. Thus, the classification of the criteria according to the level of importance when choosing an IoT platform would be (from highest to lowest importance): stability, scalability, security, flexibility, pricing model and business case, interoperability, hybrid cloud, edge intelligence and control, redundancy and disaster recovery, data analytics and visualization tools, system performance, application environment, protocol, previous experience, extent of legacy architecture, ownership of cloud infrastructure, data ownership, time to market, attractive interface, bandwidth, and platform migration.
With this classification, it was observed that two of the six most important criteria in this decision coincided with the main concerns of Spanish companies when implementing this technology (initial costs, security, and privacy [12]), in addition to criteria that initially did not stand out as being key criteria. This was due to the fact that the methodology used in this study (fuzzy cognitive maps) takes into account that the criteria that influence a decision influence each other [47].
In this way, the additional criteria that became key criteria in this selection (and that were not considered as key before) are stability, scalability, flexibility, and interoperability (which were added to the security and pricing model and business case, which were those that directly implied the main concerns of the companies: initial costs, security, and privacy); they directly and indirectly influence the main concerns of Spanish companies when implementing this technology, to such a level that their importance is higher than that for some criteria directly implied by the main concerns of companies (security and pricing model and business case).

Conclusions
With all this, this study glimpsed what criteria should be taken into account by Spanish companies when choosing an IoT platform (scalability, flexibility, data analytics and visualization tools, redundancy and disaster recovery, stability, security, data ownership, protocol, system performance, time to market, extent of legacy architecture, attractive interface, pricing model and business case, ownership of cloud infrastructure, interoperability, application environment, hybrid cloud, platform migration, previous experience, edge intelligence and control, and bandwidth), and identified the key criteria in choosing one platform over another (stability, scalability, security, flexibility, pricing model and business case, and interoperability). This evidenced that there were key criteria beyond the criteria directly implied by the concerns of Spanish companies when implementing this technology, such as the stability, scalability, flexibility, and interoperability criteria, which have the same or greater importance.
This was due to the fact that this study took into account that the criteria that influence a decision influence each other, and this could give rise to criteria not being considered at first, even though they should be taken into account in this choice. The existing studies on this problem did not consider this.
In this way, the true key criteria relevant to the choice of one or another IoT platform were revealed, in order to facilitate this process for Spanish companies that want to implement this technology in the future, and in order to encourage other companies that did not have plans to implement this technology because they encountered too much difficulty in this process to do so.
The incorporation of this technology into companies will generate great value in these, thanks to the data that it can collect and the use of that data. This value is seen in the benefits it brings to companies: savings in energy consumption, reduction in costs due to failures in machinery and production lines, greater production efficiency, better use of changing markets, greater understanding of their customers, better service to their clients, greater security for their employees, greater security and privacy of company information, and better decision making and strategy planning, among others.
This research was limited to the study of key criteria in the choice of IoT platforms in Spanish companies, since the main concerns of Spanish companies when implementing this technology have been identified (through previous studies). However, this territorial limitation could be eliminated in future research by identifying whether these concerns are the same in other territories or not, and if this is not the case, carrying out this same study in the territory of the companies to be analyzed, with experts in technological implementation and professionals responsible for information technology in companies in that territory. Further progress could be made in the studies level of detail if the analyses were made based on the size of the company, if a considerable number of companies of various sizes were included.