Insertion of Sustainable Practices in Small and Medium-Sized Companies: Analysis of the Main Barriers in the Brazilian Metalworking Sector

: The main objective of this study was characterised by analysing the barriers associated with the insertion of sustainable practices in small and medium-sized companies in the metalworking sector, considering the Brazilian reality. Thirteen barriers were previously listed from the literature, and information about them was collected from 24 experienced managers who know the realities of small and medium-sized companies in the sector. Data were analysed using descriptive statistics. The frequency distribution showed that in each barrier, there was more than 50% of the responses allocated in the upper range, that is, medium or intense observation, demonstrating that SMEs experience many difﬁculties in this theme. When comparatively analysed via Fuzzy TOPSIS, difﬁculties associated with lack of knowledge and ﬁnancial resources/incentives that can support SMEs in adopting sustainable practices are highlighted. The main contribution of this study is to provide robust information that company managers and other researchers can use. In addition, the provided information can support more complex debates for structuring public policies.


Introduction
Regardless of their size, companies are key factors in searching for a more sustainable future [1]. By adopting sustainable practices, they can contribute to the development of countries' economies, innovation, and job creation, thus contributing in different aspects to the development of society [1].
The contributions of the companies to sustainable development are even more evident when the Sustainable Development Goals (SDGs) 9 and 12 are highlighted. They have direct relations with industry, production and sustainable consumption, and their targets present terms essential for sustainable development such as regional development, human wellbeing, employment generation, sustainable and inclusive economic growth, technological evolution, efficient use of resources, and encouragement of sustainable consumption, among others [2].
Focusing on Small and Medium Enterprises (SMEs), their role towards the development of nations should be highlighted. Generally, they are in greater quantity than the large companies in a country, and they contribute with a significant share of wealth In the first stage, bibliographic research was carried out to identify the barriers to the insertion of sustainability in small and medium-sized companies. To this end, the following terms were used with their possible combinations: barriers, difficulties, small companies, medium-sized companies, SMEs, sustainability, and sustainable practices. It should be noted that some synonyms for some words were used, as an example, for the company (enterprise, organisation). The main scientific bases were consulted and as result of this search, it was possible to obtain Table 1, which presents 13 barriers associated with the adoption of sustainable practices in SMEs.

B6
Difficulty to measure the risks associated with adopting sustainable practices [15,22,24,36] B7 Deficiencies in organisational communication for the dissemination of sustainable practices [22,24,30,35,36,38] B8 Lack of market demands related to sustainable aspects in the development of new products [18,19,21,22,25,26,30] B9 Intense competition in the sector, leading SMEs to reduce their costs, compromising possible resources for sustainable practices adoption [18,19,21,28,30,32,33,35,38,39] B10 Employees' resistance to change regarding sustainable practices adoption [22,25,30] B11 Lack of interest from companies' managers regarding the adoption of sustainable practices [14,17,19,20,[24][25][26][27][28][29][30][31]34,36,37] B12 Immediate vision of the company, making sustainable practices adoption difficult, which in general require more time and planning [20,25,26,31,34,40] B13 Difficulty in establishing partnerships (with mutual benefits) between the company and its suppliers [15,18,20,22,25,35] Considering the 13 barriers presented in Table 1, a questionnaire was structured and used to survey 24 managers of SMEs in the Brazilian metalworking sector (stage 2). The first part of the questionnaire focused on sample characterisation and the second part was regarding the analysis of the barriers. For each barrier, considering the metalworking sector as a whole, the respondents should assign one of the following options: "barrier not observed" (NO), "barrier observed subtly" (OS), "barrier observed on an average level" (OA), and "barrier intensely observed" (OI). It should be noted that, in Brazil, research involving people must be assessed by a research ethics committee before data collection, and this was done. The above ethics committee approved the research. In the first stage, bibliographic research was carried out to identify the barriers to the insertion of sustainability in small and medium-sized companies. To this end, the following terms were used with their possible combinations: barriers, difficulties, small companies, medium-sized companies, SMEs, sustainability, and sustainable practices. It should be noted that some synonyms for some words were used, as an example, for the company (enterprise, organisation). The main scientific bases were consulted and as result of this search, it was possible to obtain Table 1, which presents 13 barriers associated with the adoption of sustainable practices in SMEs.

B9
Intense competition in the sector, leading SMEs to reduce their costs, compromising possible resources for sustainable practices adoption [18,19,21,28,30,32,33,35,38,39] B10 Employees' resistance to change regarding sustainable practices adoption [22,25,30] B11 Lack of interest from companies' managers regarding the adoption of sustainable practices [14,17,19,20,[24][25][26][27][28][29][30][31]34,36,37] B12 Immediate vision of the company, making sustainable practices adoption difficult, which in general require more time and planning [20,25,26,31,34,40] B13 Difficulty in establishing partnerships (with mutual benefits) between the company and its suppliers [15,18,20,22,25,35] Considering the 13 barriers presented in Table 1, a questionnaire was structured and used to survey 24 managers of SMEs in the Brazilian metalworking sector (stage 2). The first part of the questionnaire focused on sample characterisation and the second part was regarding the analysis of the barriers. For each barrier, considering the metalworking sector as a whole, the respondents should assign one of the following options: "barrier not observed" (NO), "barrier observed subtly" (OS), "barrier observed on an average level" (OA), and "barrier intensely observed" (OI). It should be noted that, in Brazil, research involving people must be assessed by a research ethics committee before data collection, and this was done. The above ethics committee approved the research.
Stage 3 corresponded to the execution of the survey with 24 managers of SMEs in the Brazilian metalworking sector. Considering the argument of Chen [40], regarding the uncertainties on the answers of respondents of a survey, this scale was transformed into triangular fuzzy numbers (TFN) as presented in Figure 2.
Stage 3 corresponded to the execution of the survey with 24 managers of SMEs in the Brazilian metalworking sector. Considering the argument of Chen [40], regarding the uncertainties on the answers of respondents of a survey, this scale was transformed into triangular fuzzy numbers (TFN) as presented in Figure 2. Analysing the characteristics of the 24 respondents, it was possible to notice that they are at different position levels in their companies (from engineers/coordinator to director), they have different levels of education (from technician to postgraduate), and different levels of experience (ranging from 14 to 45 years of experience in business management). Thus, they were classified into three levels, L3, L2 and L1, in which L3 is the level for those with the greatest ability to infer about the barriers presented for the sector and N1 is the level for those with the lowest ability to infer about the barriers presented. The authors made the classification of this article analysing the characteristics of the respondents, but as mentioned by Chen [40], this process also presents uncertainties; thus, the classifications were also considered using fuzzy triangular numbers, as shown in Figure  3. Data analysis was conducted through descriptive statistics and the Fuzzy TOPSIS technique (stage 4), according to the guidelines of Chen [40]. For this, the first step was to define the Matrix , using the scores attributed by the respondents using TFN format and the vector that represents respondents' levels of qualification also using TFN format.
[ , , ] = (Matrix 2) It is necessary to normalise the Matrix in the next step. For this, data was normalised by the greatest value to make the greatest barrier the first in the rank. The Equation (1) was used in this step. From this procedure, the Matrix (matrix 3) is obtained. Analysing the characteristics of the 24 respondents, it was possible to notice that they are at different position levels in their companies (from engineers/coordinator to director), they have different levels of education (from technician to postgraduate), and different levels of experience (ranging from 14 to 45 years of experience in business management). Thus, they were classified into three levels, L3, L2 and L1, in which L3 is the level for those with the greatest ability to infer about the barriers presented for the sector and N1 is the level for those with the lowest ability to infer about the barriers presented. The authors made the classification of this article analysing the characteristics of the respondents, but as mentioned by Chen [40], this process also presents uncertainties; thus, the classifications were also considered using fuzzy triangular numbers, as shown in Figure 3.
Stage 3 corresponded to the execution of the survey with 24 managers of SMEs in the Brazilian metalworking sector. Considering the argument of Chen [40], regarding the uncertainties on the answers of respondents of a survey, this scale was transformed into triangular fuzzy numbers (TFN) as presented in Figure 2. Analysing the characteristics of the 24 respondents, it was possible to notice that they are at different position levels in their companies (from engineers/coordinator to director), they have different levels of education (from technician to postgraduate), and different levels of experience (ranging from 14 to 45 years of experience in business management). Thus, they were classified into three levels, L3, L2 and L1, in which L3 is the level for those with the greatest ability to infer about the barriers presented for the sector and N1 is the level for those with the lowest ability to infer about the barriers presented. The authors made the classification of this article analysing the characteristics of the respondents, but as mentioned by Chen [40], this process also presents uncertainties; thus, the classifications were also considered using fuzzy triangular numbers, as shown in Figure  3. Data analysis was conducted through descriptive statistics and the Fuzzy TOPSIS technique (stage 4), according to the guidelines of Chen [40]. For this, the first step was to define the Matrix , using the scores attributed by the respondents using TFN format and the vector that represents respondents' levels of qualification also using TFN format.  Data analysis was conducted through descriptive statistics and the Fuzzy TOPSIS technique (stage 4), according to the guidelines of Chen [40]. For this, the first step was to define the Matrix G, using the scores attributed by the respondents using TFN format and the vector E that represents respondents' levels of qualification also using TFN format.
It is necessary to normalise the Matrix G in the next step. For this, data was normalised by the greatest value to make the greatest barrier the first in the rank. The Equation (1) was used in this step. From this procedure, the Matrix R (matrix 3) is obtained. , In the sequence, Matrix R is weighted by the vector E, and the Matrix V (Matrix 4) is developed. Once the Matrix V is obtained, the distances from each element to the positive and negative ideal solutions are calculated. In this case, the following positive and negative ideal solutions were used respectively. The calculus of the distances is calculated through Equation (2).
The total positive d * i and negative d − i distances concerning each alternative (in this study, "the barriers for sustainable practices adoption in SMEs of Brazilian metalworking sector") is obtained through the sum of the partial distances, as evidenced in Equations (3) and (4). Finally, the closeness coefficient (CCi) that enables rank the alternatives is calculated using Equation (5).
The sensitivity analysis of the ordering created by Fuzzy TOPSIS was based on the guidelines developed by Memari et al. [41]. In the particular case of this research, respondents were classified according to their experiences in classes N1, N2 and N3. Regarding the traditional Fuzzy TOPSIS proposed by Chen (2000), in this study, the categories N1, N2 and N3 assume the role of the "criteria". Thus, for the sensitivity analysis, three additional scenarios were analysed, in which the impact on the ranking was analysed in case one of the categories was totally omitted. Subsequently, the discussion and conclusions are done at the end of the article (stage 5).

Results and Debates
The first result to be presented here refers to the classification of respondents concerning their greater ability to infer on the barriers for inserting sustainable practices in SMEs. As previously mentioned, it is understood that in the extreme, those respondents in board positions that have postgraduate training and many years of experience in the sector present a greater knowledge on the subject and, consequently, they have a greater capacity for inference (they are at level L3). Using this logic for the other relationships, the 24 respondents were carefully analysed and classified according to the levels presented in Table 2. Performing a frequency distribution analysis for each of the analysed barriers (see Table 3), it is possible to notice that more than 50% of the responses measured by the sample are in the "average observed" or "intensely observed" range, clearly showing that practically all barriers are present in the daily lives of SMEs in the Brazilian metalworking sector when they aim to introduce sustainable practices. In a general way, this result can be considered in line with statements presented by Ahmad et al. [10] when they argue that SMEs face many barriers to adopt sustainable practices in developing countries. Through the application of the Fuzzy TOPSIS technique, it was possible to carry out a comparative analysis between the barriers studied, considering the ability of each respondent to assess the topic and the uncertainties associated with their allocation at levels N1, N2 and N3 and the uncertainties inherent to the process of measuring responses in the survey, as previously mentioned [40].
Due to the matrices ∼ G, R and V size, which present several lines and columns, it is not possible to present them here. Thus, it is presented the calculation of the distances of each of the elements of the matrix V in relation to the positive and negative ideal distances. These distances are presented in Tables 4 and 5. In these tables, the total positive d * i and negative d − i distances for each barrier are also presented. Considering positive d * i and negative d − i total distances for each barrier, it was possible to calculate the closeness coefficients CC i using these coefficients, it was possible to rank the barriers according to their difficulty level, based on respondents' answers. The coefficients CC i calculated for each barrier are presented in Table 6 and the rank of the analysed barriers is presented in Table 7.
To perform the sensitivity analysis, three scenarios were structured considering the combination of two classes of respondents and the exclusion of others. Therefore, the scenarios structures were: Scenario 1 (only data from respondents classes N2 and N3); Scenario 2 (only data from respondents classes N1 and N3); Scenario 3 (only data from respondents classes N1 and N2). As mentioned in Section 2, this sensitivity analysis procedure was based on the guidelines developed by Memari et al. [41]. As a result, it is possible to have a better idea of the influence of each class on the responses. The results of these scenarios are shown in Table 8.
Analysing the results obtained, via Table 7, in the first positions, there are difficulties B2 and B1 regarding the lack of knowledge and financial resources/incentives that can support SMEs in adopting sustainable practices. These two barriers are broadly aligned with what the literature points out, not only for the adoption of sustainable practices, but for improvements in all aspects of management for SMEs. With small and overloaded work teams, it is not easy to achieve time to assimilate new knowledge, whether associated with innovation, sustainability, or any other concept. Furthermore, one of the characteristics of SMEs is that they have a more volatile reality, with more uncertain financial resources. In this way, managers end up prioritising resources for those needs they consider every day. It is also worth remembering, as mentioned [3,4], the COVID-19 pandemic further highlighted the uncertainties discussed above.  Next, the barriers associated with employees appear, especially those related to resistance to change and engagement; government deficiency in not encouraging the adoption of sustainable practices through regulations, difficulty in measuring risks and establishing partnerships. In the sequence, it is possible to observe a block composed of barriers that result from the daily consequences of SMEs and, finally, a block already has more specific characteristics.
When the sensitivity analysis is performed, the removal of the N1 category of respondents has very little influence on the ordering, only with a change in the positions between barriers 9 and 12. When removing the N2 category of respondents, it is observed that barrier 7 has the relevance increased. With the elimination of N3, the B6 barrier gains evidence. Overall, B2 and B1 are evidenced in all situations. As in the study of Memari et al. [41], the scenarios analysis is an essential part of Fuzzy TOPSIS analysis.
A suggestion to overcome many of the obstacles presented in this study would be to create a governmental program to train employees of the metalworking sector regarding sustainability knowledge and encouraging SMEs to adopt sustainable practices. As an example, an online government platform could be created for the training mentioned above and a tax incentive program. Small and medium-sized companies that encourage their employees to learn about sustainability and invest resources in sustainable practices in their processes could deduct part of these investments in taxes.

Conclusions and Final Considerations
The main objective of this study was to analyse the barriers associated with the insertion of sustainable practices in small and medium-sized companies in the metalworking sector. Considering the Brazilian reality and face the results presented, it is possible to note that it was achieved.