1. INTRODUCTION

The number of Internet of Things (IoT) devices is rapidly increasing (Saleem et al., 2018). Almolhis et al. (2020) backed up this conclusion, adding that the rapid rise of IoT technology has the potential to affect all aspects of human life. Essentially, the manufacturing and industrial sectors are driving this IoT expansion. Nonetheless, the growth from a research standpoint has not been widely explored. The number of publications on IoT is rising, according to Kotha and Gupta (2018). They did not, however, present any proof to back up their assertion. On the other side, if this fast expansion is not adequately addressed, IoT might pose issues in terms of interoperability, communication, data processing, integration, security, and privacy (Saleem et al., 2018). Even though research in IoT-related domains began in 2008 throughout the world, the rise was seen after 2013 in numerous high-quality journals listed on the Web of Science, Scopus, and other databases (Mącik, 2017). As a result, in order to examine current developments, the research needs to look at publications published after 2013.

Nonetheless, research on individual consumers’ perceptions about IoT adoption is currently lacking in the IoT literature, which focuses more on the technology aspect (AlHogail, 2018). Understanding customer acceptability has become a must for bringing IoT solutions into everyday life. Because the IoT is still in its early phases, only a small amount of interest has been paid from the standpoint of individual consumers (Lee & Shin, 2019). Individual acceptability, according to Kim and Kim (2016), is one of the barriers to IoT technology innovation that must be addressed. This argument is backed up by Hsu and Lin (2018), who have also pointed out that influencing factors of IoT applications and services have received little attention. On the other hand, deciding whether to go with a generic or specific model is always a challenging issue (Fahmideh et al., 2020). In any case, a generic model is required to be developed in the IoT domain which will not be customized to any specific application. This generic paradigm, according to Attali et al. (2010), has major implications for acceptance since it allows for more meaningful interpretation. These models are also adaptable and transferable to other particular applications (Fahmideh et al., 2020; Rothgangel & Riegel, 2021).

Apart from this, according to Brous et al. (2017), despite the obvious prospects and optimism, IoT adoption remains low. Karahoca et al. (2018) have further added that the determination of influencing factors is vital to increasing consumer IoT adoption. In contrast, a systematic literature review (SLR) study is necessary to determine the influencing factors (George et al., 2016; Meng et al., 2021; Thuan et al., 2016). Notably, an SLR study provides a globally standard, accurate, and consistent literature review (Crocetti, 2016). Therefore, the purpose of this SLR is to enhance consumer IoT adoption by addressing the research gaps discussed above. On the other hand, the reason for performing weight analysis is that the weights represent an independent variable’s predictive power (Jeyaraj et al., 2006). Moreover, weight estimation and meta-analysis have a closer relationship, since the larger the weight of an independent variable, the more likely it is to be significant when conducting a quantitative investigation (Rana et al., 2015). Therefore, this study’s objectives are as follows:

Thus, the rest of the sections of this article are organized as follows: Section 2 contains the fundamental information related to IoT adoption, weight, and their relationships. The methodology is outlined in Section 3 along with the entire review procedure. Following that, in Section 4 the study’s findings are presented. Subsequently in Section 5, there is a discussion of the study findings. A conclusion remark, as well as the implications of this study withthe study’s weaknesses and some ideas for future research, is included in the last section.

2. THEORETICAL BACKGROUND

To have a complete awareness of the target region, the following issues will be investigated. To be specific, in this section five topics will be discussed. A brief overview of IoT and its global acceptance will be explored first. Afterward, the weight calculation process will be briefly explained. Next, the relationships between IoT adoption and weight calculation will be depicted. Finally, review studies of IoT adoption will be explored.

3. METHODOLOGY

In this study, the following steps (see Fig. 1) were followed as directed by Kitchenham and Charters (2007) and Ain et al. (2019). This study began by stating the research objectives and was completed with a discussion of the findings. In between, databases were selected, articles were searched, papers were filtered based on inclusion criteria, data were extracted from those papers, and results were demonstrated.

새창으로 보기

Fig. 1

Steps in the systemic literature review.

Since the context of this study fits our research, Alkawsi and Ali (2018) were followed to identify the literature databases for this investigation. In addition, these databases provide access to prestigious journals as well as high-quality andpeer-reviewed conference papers. Hence, the exploration was performed on five databases, ScienceDirect, Scopus, Google Scholar, IEEE Explore, and Taylor & Francis from August 5 to September 30, 2020. Our search was initiated for the papers associated with IoT adoption. Afterward, the following phrases and keywords were utilized: “IoT adoption,” “IoT adoption intention,” “IoT adoption behavior,” “Consumer IoT adoption,” “Consumer IoT model,” and “IoT adoption model,” in all possible permutations and combinations. Also, another keyword, “acceptance,” replaced the word “adoption” in all cases. In order to address the three research objectives, the following inclusion parameter in Table 1 was followed. It is worth mentioning that based on these parameters, the articles were collected from the above-mentioned five databases.

All of the articles chosen for this study were empirically evaluated in the IoT area using a quantitative approach. General context materials, technical papers, articles written in languages other than English, articles focusing on the organizational viewpoint, periodical articles, etc. were filtered out. Afterward, based on the recommendations of Suppatvech et al. (2019), the 226 downloaded papers from the above-mentioned databases were screened in these three stages (see Fig. 2). Through these steps, duplicate articles were removed from the downloaded articles. Next, these papers were screened through the support of paper title, abstract, and full text.

새창으로 보기

Fig. 2

Stages for article screening.

Here, ‘n’ denotes the number of articles that were accepted after each stage. At the accomplishment of screening, 87 papers were selected for SLR from 13 conferences and 54 journals. The flow diagram can be found in Fig. 3. This flow diagram demonstrates the procedure of how the 87 articles were selected for analysis from the initial discovery of 226 articles. This diagram further explores how the desired articles were confirmed after each step of selection.

새창으로 보기

Fig. 3

The flow diagram.

Finally, this study followed two types of analysis procedures. These are descriptive exploration as advised by Suppatvech et al. (2019), and weight evaluation as suggested by Jeyaraj et al. (2006). To illustrate, the descriptive analysis focused on the classification of papers according to the name of authors, their countries, publication year, journals, sample type, dependent variables, independent variables, etc. On the other hand, weight evaluation of the variables was performed using the procedure explained in Section 2.3. In order to perform these analyzing procedures, data were extracted from the 87 papers using an Excel worksheet. Afterward, the frequency of the variables from different theories and models was counted. Notably, the figures and tables of the next section were achieved based on the data extracted from this Excel worksheet. The fields used in the Excel worksheet are shown in Table 2.

4. RESULTS

Interesting insights into publishing patterns were found in the quantitative review of the articles.

The upsurge in the number of articles published during 2015-2018 is shown in Fig. 4. The highest number of papers is found during these two years 2018 and 2019, which accumulates 57.47% of total papers. However, no accurate picture is observed for 2020 from the graph since paper searching activities were completed at the end of September 2020 and no paper in 2020 has been contributed yet by the conferences. However, we can observe a more narrative picture in the distribution of journals across these years in the following table (see Table 3).

새창으로 보기

Fig. 4

Number of articles published per year during 2014-2020.

A total of 87 articles are distributed in 54 journals and 13 conference papers. As is seen from Table 3, all of the 13 conferences and the majority of the journals only have one paper. Importantly, the top four contributing journals are Telematics and Informatics (5 papers), Technological Forecasting & Social Change (4), Behaviour & Information Technology (4 papers), and Computers & Security (3 papers). In addition, the top 12 journals in this list have provided 32 articles, which are 36.78% of the total papers. Interestingly, no paper in 2020 is contributed yet by the conferences. In addition, we have found several theories and models in those journals and conference papers (see Fig. 5).

새창으로 보기

Fig. 5

List of theories/models. ISS: information systems success model, IDT: innovation diffusion theory, TRA: theory of reasoned action, TPB: theory of planned behavior, PMT: protection motivation theory, UTAUT: unified theory of acceptance and use of technology, VAM: value-based adoption model, TAM: technology acceptance model, HBM: health belief model, PAM: post-acceptance model, ECT: expected confirmation theory, CPM: communication privacy management, PCT: privacy calculus theory, BRT: behavioral reasoning theory, APCO: antecedents-privacy concerns-outcomes.

All the articles were categorized according to established theories and models. Notably, the extensions of models like the technology acceptance model 2 (TAM 2), technology acceptance model 3 (TAM 3), unified theory of acceptance, and use of technology 2 (UTAUT 2) were a part of the regular technology acceptance model (TAM) and unified theory of acceptance and use of technology (UTAUT) models. Further, 11.5% of articles were categorized under the ‘general’ category that did not belong to any specific theory or model but rather inherited constructs from different studies. Notably, some of the papers proposed conceptual models by integrating multiple theories. As a result, the number of the following theories exceeds the number of collected papers. TAM (40 papers) is the most widely used model in the checked articles, whereas the UTAUT (23 papers) and theory of planned behavior (TPB) (8 papers) come next. These top three theories have contributed to 62.83% of the total papers. Further, all these articles were categorized into 23 types of different technologies and services of IoT (see Fig. 6).

새창으로 보기

Fig. 6

List of Internet of Things enabled technologies and services.

Most journals (31.03%) follow general technologies and services, where these technologies and services do not specify any specific applications, rather IoT technology and services in general. Smart homes (17.24%) and health (20.68%), on the other hand, are the most popular specific technologies and services, respectively. However, most of the applications (15 out of 23) have been used only once in these papers. According to our findings, the top five contributing countries are South Korea, Malaysia, China, India, and the USA, which have contributed 42.85% of the total amount of papers.

Moreover, a total of 246 authors contributed to developing 87 IoT-related papers with an average of 2.827 or almost three authors per paper. Only 14 authors contributed more than one paper whereas 232 authors contributed only once. Nonetheless, these 246 authors have utilized 14 different types of samples in their studies (see Fig. 7). A maximum of papers (48.27%) have utilized consumers as the sampling type, followed by students (14.94%). Notably, students, teachers, faculty members, and university staff are considered the same sampling type.

새창으로 보기

Fig. 7

List of sampling types.

We have identified several theories and models utilized in the IoT domain in Fig. 5. In general, these theories and models were developed based on the combination of dependent-independent variables and their relationships. Examples of dependent variables can be Intention, Behavior, and Attitude, which were also the focus variables of this study, and independent variables performed as the influencing factors of these focus variables. As for explanation, the Trust variable acted as a predictor of Intention in 11 papers where the relationship of Trust-Intention was significant nine times. Therefore, Trust gained the weight value of 0.82 and can be categorized both as WUT and BPR (see Appendix). Importantly, in order to perform the weight analysis, we have to count the frequency of these independent variables first.

Therefore, we have counted the frequency of independent variables which were used in different theories and models in these papers (see Table 4). Our analysis has found 165 exogenous variables in different models and theories of IoT papers. Based on the occurrences, the top five variables are Perceived usefulness (48 times), Perceived ease of use (38 times), Attitude (33 times), Social influence (24 times), and Performance expectancy (18 times), which are part of TAM and UTAUT. Apart from the established models, Trust (17 times), Risk (16 times), and Privacy (16 times) are highly used.

Further, we have evaluated the weight of each predictor variable. We have three dependent variables: Attitude, Intention, and Behavior. The anticipation or present observations of behavioral advantages and negatives are referred to as Attitudes (Rodríguez-Calvillo et al., 2011). In contrast, the degree to which an individual has formed a conscious strategy to undertake or not perform certain prospective actions is defined as Intention. Finally, the user’s self-reported frequency and amount of technology use are characterized as Behavior (Amelia & Ronald, 2017). Notably, Attitude determines Intention, and Intention predicts Actual usage or Behavior, according to the TRA, TAM, and TPB models (Olushola & Abiola, 2017). Furthermore, according to Rayat (2018), a diagrammatically represented figure is meant to provide a graphical demonstration of quantitative facts to develop the concept more comprehensively. In addition, this diagram can also be utilized to evaluate the weight of variables to finalize their overall outcome (Rana et al., 2015). As a result, following Rana et al. (2015), all the significant and non-significant associations are diagrammatically displayed in Figs. 8, 9, and 10, along with their weight values and type of relations (either negative or positive). It is worth mentioning that the weight values of each variable directed towards Intention, Behavior, and Attitude are included in the Appendix in the form of tables, where the calculation procedure has been previously described in Section 2.3. In addition, the above-mentioned figures (Figs. 8Fig. 9-10) have extracted the outcomes of these tables accordingly.

새창으로 보기

Fig. 8

All associations are diagrammatically represented toward Intention.

새창으로 보기

Fig. 9

All associations are diagrammatically represented toward Behavior.

새창으로 보기

Fig. 10

All associations are diagrammatically represented toward Attitude.

A total of 94 variables are observed to have direct relationships with Intention where 12, 18, 49, and 76 of them are BPR, WUT, PROM, and EXPR, respectively (see Fig. 8). The top seven occurred variables are Attitude (37 times), Perceived usefulness (34 times), Social influence (20 times), Perceived ease of use (18 times), Performance expectancy (15 times), Effort expectancy (15 times), and Subjective norm (14 times). On the other hand, a total of 26 variables have direct relationships with Behavior where 1, 1, 17, and 25 of them are BPR, WUT, PROM, and EXPR, respectively (see Fig. 9). The maximum occurred variable is Intention (13 times), followed by Trust (2 times) and Risk (2 times), and the remaining variables exist only once.

At last, a total of 33 variables are observed to have direct relationships with an Attitude where 2, 3, 23, and 30 of them are BPR, WUT, PROM, and EXPR, respectively (see Fig. 10). The maximum occurred variables are Perceived ease of use (23 times), Perceived usefulness (23 times) which followed Risk (5 times), Trust (3 times), Privacy (3 times), Performance expectancy (3 times), and Effort expectancy (3 times). Moreover, 22 out of 33 variables occur only once.

Jeyaraj et al. (2006) advised researchers to continue employing the best predictors within their conceptual models for consumer adoption-related studies. Following the recommendations of Jeyaraj et al. (2006), a conceptual consumer adoption model has been proposed supported by the outcomes of the above three figures. According to our findings, the 12 best predictors of Intention are Perceived ease of use (0.83), Perceived usefulness (0.85), Trust (0.82), Satisfaction (1.00), Habits (1.00), Perceived value (0.88), Subjective norm (1.00), Enjoyment (0.83), Facilitating condition (1.00), Performance expectancy (1.00), Self-efficacy (0.8), and Attitude (0.92). On the other hand, Attitude has the two best predictors—Perceived ease of use (0.83) and Perceived usefulness (0.96); and Behavior has only Intention (1.00) as the best predictor. Interestingly, all of these best predictors are positively associated with the respective focus variables. To measure IoT adoption, we focused primarily on Intention and Behavior, with Attitude acting as a mediator between Perceived ease of use (0.83), Perceived usefulness (0.96), and Intention. More simply, we can say that these 12 best predictors have achieved the top weighting values, and these are the most influential factors that determine the adoption of IoT among the consumer at the individual level. Thus, the following consumer adoption model can be developed (see Fig. 11).

새창으로 보기

Fig. 11

Internet of Things adoption model using the best predictors from weight analysis.

As we discovered in our study, the correlations between independent and dependent factors are not always significant, and independent variables have sometimes failed to become an efficient predictor of dependent variables. The developed model, on the other hand, is made up of the 12 best predictors that have attained the highest weighting values. As a result, in terms of predictive or explanatory capacity, the suggested conceptual model outperforms any other evaluated theories and models in this literature study. On the flip side, since the model is generic, it may be used in a wide range of IoT areas and can aid in the development of novel and successful social applications. Furthermore, the suggested model is anticipated to serve as guidance for future studies into the aspects that drive individual IoT adoption. Importantly, factor weight patterns may be used as a guide for the subsequent variables and can be further studied to demonstrate their efficacy.

5. DISCUSSION

To comprehend the recent progression of individual consumers’ IoT adoption and the overall efficiency of the constructs, this review study performed a weight analysis of the variables from 87 articles in academic journals and conference proceedings published during the period 2014-2020. This research was carried out with three goals in mind. To be specific, the first goal of this paper was to investigate the growth of individual IoT acceptance in recent years. Therefore, the following significant points emerged from the findings:

The second objective of this study was to find the most influencing factors of the Intention and Actual usage of IoT applications using weight analysis. This objective was fulfilled through the following steps. The first step was to find the frequency of all variables used in the theories and models to determine Adoption Intention and Actual usage. The second step was to conduct a weight analysis of the variables associated with Attitude, Intention, and Behavior. Notably, we have included Attitude for weight analysis since this variable has been used as a mediator of Intention and Behavior in many papers. According to the findings, among the 165 variables from 87 articles, only 28 variables occurred more than five times, 59 variables occurred more than once, and the remaining 106 variables were attempted only once. Thus, it can be understood that most of the variables were not attempted sufficient times. Afterward, we analyzed 94 direct relationships with Intention, 26 with Behavior, and 33 with Attitude. Following that, 2, 12, and 1 best predictors were found for Attitude, Intention, and Behavior, respectively from our weight analysis. On the other hand, the numbers of promising predictors were 23, 49, and 17 for Attitude, Intention, and Behavior, respectively.

The third objective was to develop an IoT adoption model using the best predictors achieved from weight analysis. We counted the frequency and weight in all research papers on the associations between exogenous and endogenous variables. Afterward, we have found the 12 best predictors from the weight analysis and a model is developed accordingly. These best predictors are Perceived ease of use (0.83), Perceived usefulness (0.85), Trust (0.82), Satisfaction (1.00), Habits (1.00), Perceived value (0.88), Subjective norm (1.00), Enjoyment (0.83), Facilitating condition (1.00), Performance expectancy (1.00), Self-efficacy (0.8), and Attitude (0.92), which is found to be significant sufficient times in the literature. Besides this, Attitude performs as a mediator between Perceived ease of use (0.83), Perceived usefulness (0.96), and Intention. Further, we have found several promising predictors that can affect Attitude (30 promising predictors), Adoption intention (49 promising predictors), and Behavior (17 promising predictors) of IoT applications. These promising variables have the potential to become the best predictors. However, these variables have not yet been sufficiently attempted. Indeed, Jeyaraj et al. (2006) recommended studying promising predictors inside the conceptual models for individual consumer adoption-related studies.

6. CONCLUSION

This study shows the exponential rise in the number of papers published in the last seven years, which demonstrates IoT as an emerging field with tremendous potential for future years’ publications. Though consumer IoT adoption is emerging day by day, it is still in its infant stage because several issues in 20 different technologies and services are still unaddressed and people from different sectors and professionals are not equally involved. Most importantly, if the current growth rate continues, much more progress is expected in the near future. On the other hand, the developed generic IoT model can most effectively predict user IoT adoption. Furthermore, this model can be applicable to a wide range of domains and can help us develop innovative, effective social applications.

Significantly, this is one of the first SLR studies that utilize weight analysis to determine influencing factors for IoT-enabled applications. Moreover, this study’s theoretical success is also based on portraying the combined diagrammatic representation for individual IoT adoption and calculating the number of significant and non-significant correlations between the leading constructs, which are then used to assess the weight analysis. Afterward, a new IoT adoption model has been developed in our study which is generic and capable of addressing all the analyzed IoT technologies and services. In addition, this developed model is capable of enhancing theoretical knowledge of individual IoT adoption and is able to develop a solid foundation for information management literature. From a practical standpoint, consumers, communities, and businesses might all profit from this developed IoT paradigm. Additionally, similar to our findings, Lee and Lee (2015) confirm that the potential of IoT devices is enormous; however, the future uncertainty of investment remains a concern for decision-makers. Therefore, the current growth of IoT among individuals in recent years and derived influencing factors (best predictors and promising predictors) are expected to influence and assist organizations, policymakers, and entrepreneurs in the further development of their business plans. Further, the results of this study have meaningful implications for companies that provide IoT-based services and/or decide to launch new services.

This study has a few drawbacks which can be addressed in future studies. To begin, papers were collected only from five databases, as stated earlier. Hence, some other databases such as EBSCO, ProQuest, PubMed, ACM, JSTOR, Wiley, or Hindawi were overlooked. Afterward, the searching keyword used for this weight analysis might be inadequate. We could have used more keywords related to specific IoT applications such as “smart home adoption/acceptance,” health service adoption/acceptance,” etc. As a result, we may have overlooked some articles and facts. Moreover, we performed a weight analysis on the generic issues of IoT adoption, whereas specific issues might lead to some other interesting results. On the other hand, some more statistical procedures can be followed to measure the relationships between the variables, for example, standard normal deviation, average β values, and lower & upper 95% confidence intervals (Baptista & Oliveira, 2016). Last, we have addressed individual consumer IoT adoption in this paper, and organizational IoT adoption might also be an interesting topic. Afterward, it would also be possible to make a comparison between these two types of IoT adoptions which will lead to some remarkable results.