1 Introduction

In the past few years, there has been an exponential growth in the use of smart wearable devices as the number of estimated connected wearable devices worldwide is expected to jump from 100 million in 2016 to over 373 million in 2020 [1]. By 2018, smartwatches are forecast to account for about half of all wearable unit sales worldwide [1]. Wearable devices, especially smartwatches are predicted to have significant impact on the quality of people’s daily life [2]. The smart wearables exist in a variety of forms ranging from basic fitness trackers, smartwatches, smart glasses, wireless headsets, clothing, to smart and fashionable jewelry items like bracelets, necklaces, etc. [3]. Often, there is a confusion between smart fitness trackers and smartwatches because there is a heavy overlap in the type of functionalities they provide. Therefore, right at the onset we provide a working definition of a smartwatch as “a wrist-worn device with advanced computational power, that can connect to other devices via short range wireless connectivity; provides alert notifications; collects personal data through a range of sensors and stores them; and has an integrated clock” [4].

Currently, there is a lot of competition in the smartwatch sector, which is evident by the presence of the major IT and electronics giants like Google, Apple, Microsoft, Samsung, Motorola, etc., all competing against each other in this segment. This is good news for the end-users, since they have a wide variety of choice while selecting a smartwatch. Attracting a customer for the first time to use any product is relatively simpler than retaining existing ones because users always want to try out new technology, products, and services [5]. This is particularly true in the smartwatch industry where even a slight dissatisfaction can result in a change of brands due to the plethora of options available with all of them providing nearly identical functionalities. With huge advancements in the Internet of Things (IoT) technologies along with cloud computing, it is very easy for an end-user to switch brands (even across different platforms like watchOS, Wear OS, Tizen, etc.) due to the ubiquitous nature of user data in the form of easy to restore data back-ups. Therefore, it is critical for the different smartwatch vendors to understand the factors that motivate the users to be loyal to them by continuing their usage of a specific smartwatch brand. Likewise, for the success and future sustainability of this industry it is extremely important to know the exact motivations behind the users’ continuous usage of the smartwatches.

The aspect of technology adoption has been covered extensively by the information systems (IS) literature [6, 7]. Similarly, the users’ continuance technology use has also been covered in depth [8,9,10]. However, the nature of technology examined by the above theories is quite different from the IoT technology products like smartwatches. A number of studies also exist in the context of end-user smartwatch adoption [11,12,13,14]. All these works take into account various aspects of human behavior in order to predict their acceptance. Nevertheless, none of them investigates the human and technological factors that account for the users’ continuance intention to use these new devices. Clearly, a research gap exists that correlates the users’ expectations before using a smartwatch with the actual experience received, post usage. We aim to bridge this gap in this work, by answering the following two research questions:

R1

What are the driving factors behind the users’ continuous usage of the smartwatches?

R2

What is the underlying theoretical model, and how does it perform in an empirical setting?

To collect relevant data for answering the above research questions, we adopt a dual strategy: a comprehensive literature search, and a thematic analysis using an ethnographic approach. The collected data are used to build a relevant theoretical framework based upon an extended version of the expectation-confirmation model that answers the end-users continuous usage of smartwatches.

The novelty of this work can be judged by the following three aspects: First, we propose a new theoretical framework that successfully explains the users’ continuous usage of smartwatches, instead of looking into their adoption details, which current research focuses on. Second, this work considers real smartwatch users (usage experience of at least 1 year) across different brands in contrast with existing studies, most of which collect the data from potential smartwatch users instead of the actual ones. Finally, since we focus on an extensive post-usage experience of smartwatches, the results of this research can be used by the relevant stakeholders to identify the existing gaps and improve upon those aspects to ensure a long-term success. Similarly, the exact reasons behind an early abandonment of smartwatches can also be inferred and suitable corrective measures taken.

2 Related research and ethnographic study

The most critical step towards creating a successful theoretical framework is an accurate estimation of the various factors involved. In this study, we use a two-pronged approach for factor identification. To begin with, an extensive literature search is done, which is well supplemented by conducting an ethnographic study on limited number of users followed by a semi-structured interview, the results of which are analyzed using a thematic approach. We use this dual strategy with the hope of discovering new factors critical in assessing the continuous usage of the smartwatches apart from those, which are already accounted for.

2.1 Smart-wearables and end-user adoption

Investigating into the social acceptance of the smart wearable industry in general is an important issue. The smart wearable devices are generally worn external to the body, either attached as an accessory, or embedded in clothes [15]. Generally, these devices are classified based on their functional properties, appearance, proximity to the human body, and other parameters [16, 17]. They exist in a variety of forms ranging from fitness-trackers, smart glasses, smart clothing, wearable medical devices, to smartwatches [17].

Although the concepts of technology adoption have been used extensively to measure the end-user perception and social impact for a number of new technologies, its usage is very limited in the context of smart-wearables. In Table 1, we present detailed information on the current state of research about technology adoption and smart wearables (excluding smartwatches). Table 1 highlights the smart wearables application category, the underlying factors along with the theoretical model used, together with the sample characteristics. More details into the current state of the wearables and the challenges being faced are presented in [18, 19].

Table 1 Synopsis of literature review related to the adoption/use of smart-wearables
Full size table

2.2 Background work on smartwatches

Table 2 presents a concise review of the work that has been done until now exclusively related to smartwatches. In the table, we highlight the objective of each study, the factors that have been considered, and the underlying theoretical framework. We also report if the model takes into account the continuous usage scenario or not, along with the experience that the users have in using the smartwatches.

Table 2 Synopsis of literature review related to the adoption/use of smartwatches
Full size table

One unique aspect that differentiates these wearables from other technology products is the concept of ‘fashnology’, which was first proposed in [42]. The adoption of the wearables can be affected from both a technology perspective as well as fashion perspective, which focus primarily on the aesthetics of design. The concept of ‘fashnology’ has been applied in a variety of contexts, ranging from smart glasses to smartwatches and found to be a significant predictor to the usage intention in all the cases [21, 24, 33, 35].

2.3 Expectation-confirmation model (ECM)

This work is based on the expectation-confirmation model. Originally, the ECM framework had been proposed to explain the satisfaction and continuance usage intention of IS services by the end-users [8]. This framework has its roots in the expectation-confirmation theory (ECT), which is a cognitive model seeking to explain the post-purchase or post-adoption satisfaction of a particular product or service as a function of expectations, perceived performance, and confirmation/disconfirmation of beliefs [43]. Initially, the end-users have some expectations regarding a particular product or service (prior to purchase). After using it for a while, they form certain perceptions about its performance. They assess the perceived performance vis-à-vis their original expectations and determine the extent to which their expectations are confirmed. Finally, they form a satisfaction based on the confirmation level.

One drawback of ECT is that it examines the effect of pre-consumption (ex ante) expectation, but not post-consumption (ex post) expectation [8]. However, ex post expectation is extremely important for those products or services where the expectation may change over time, due to the addition of new features or design changes. This fact is particularly applicable for innovative IoT products like smartwatches, which keep on adding new features, functionalities, and design changes with each new successive product iterations. The ECM model incorporates this dynamic aspect and, therefore, best suited for the present context. Additionally, to justify the applicability of the ECM model, we review previous related works using this framework and present the results in Table 3. We identify additional factors unique to the context of smartwatch usage and extend the ECM model for the same.

Table 3 Synopsis of literature review related to the ECM
Full size table

2.4 Ethnographic Study

Ethnography is a qualitative research technique where the researchers observe and/or interact with the participants under study in their real-life environment. It is a useful technique for understanding the overall user-experience (UX) of a particular product/service, especially new and innovative ones [54]. In order to capture the behavior of the users using smartwatches in their real environments and identify relevant factors motivating their continuous usage we conducted the present ethnographic study.

A total of 42 participants took part in this study and were carefully tested and selected. All of them have at-least 1 year experience in using smartwatches and diverse in age (18–55 years). In order to generalize the experiment and remove bias towards any particular product, brand or operating system, the smartwatches involved in this study came from multiple brands (Apple, Samsung, Motorola, LG, Xiaomi, Fossil, and Garmin), different product models (Apple Watch Series 1 and 3, Samsung Gear S2, etc.), different form factors (square vs. round dial) and multiple operating systems (watchOS, Wear OS and Tizen). The study lasted for a period of 2 weeks, after which a semi-structured interview was conducted with the participants. The semi-structured interview has a certain degree of flexibility associated with it and hence chosen for this purpose [55].

The results of the interview were evaluated based on a thematic approach. Thematic analysis is a useful tool for identifying, analyzing, and reporting patterns (factors) within data [56]. The overall thematic approach is shown in Fig. 1. The entire purpose for supplementing the literature review process with ethnographic study and subsequent thematic analysis is to identify any unknown factors that have not been considered by existing research on smartwatches.

Fig. 1
figure 1

Overall process of thematic analysis

Full size image

2.5 Summary

The literature review clearly points out the research gap that exists in assessing the users’ continuous usage of smartwatches. Apart from one, all the other works assess the usage-intention, rather than the continuous usage scenario. Similarly, the majority of the studies consider potential smartwatch users against real ones that can be a source of bias. With regards to the ECM model, none of the existing works take into account the dynamic requirements of an IoT paradigm, where the product concepts can change rapidly and evolve at a very fast pace.

Six prime factors (hedonic motivation, self-socio motivation, perceived privacy, perceived comfort, battery-life concern, and perceived accuracy and functional limitations) are identified in this work, apart from the four original ones (perceived usefulness, confirmation, satisfaction, and continuous usage) that are a part of the ECM framework itself. Three out of the six factors (perceived comfort, battery-life concern, and perceived accuracy and functional limitations) are obtained from the ethnographic study. Self-socio motivation is a hybrid construct, which incorporates the concepts of subjective norm (obtained from literature review) with self-motivation (obtained from the ethnographic study). Using a mix of factors already considered together with the new ones, we aim to propose a robust model for assessing the continuous usage of smartwatches. Figure 2 shows a taxonomy of the factors that has been used.

Fig. 2
figure 2

Taxonomy of the proposed factors

Full size image

3 Theoretical framework and hypotheses

This study proposes a new theoretical framework based on the ECM model for investigating the continuance intention of using smartwatches. Below, we define every factor considered followed by the relevant framework by proposing suitable causal relationships.

3.1 Perceived usefulness (PU: ECM construct)

Perceived usefulness has originally been defined as “the extent to which a person believes that using a particular technology will enhance his/her job performance” [6]. From a motivational perspective, it is a measure of the users’ level of extrinsic motivation and outcome expectancy [57, 58]. It is a widely used construct, which demonstrates a positive relationship with the adoption intention, especially in a variety of work-related contexts [59, 60]. Since the present study emphasizes on the end-user perspective related to technology use, perceived usefulness needs to be re-defined [61, 62]. For this work, PU is defined as “the extent to which an end-user believes that using smartwatches will increase his/her personal efficiency, such as being more organized and productive” [13]. The more benefits the users get, the more satisfied they will be, and the higher the probability of their continuous usage of smartwatches will be. Hence we propose the following hypotheses:

H1a

PU has a positive effect on satisfaction;

H1b

PU has a positive effect on the continuous usage of smartwatches.

3.2 Satisfaction (ST: ECM construct)

Satisfaction is a key construct that determines the post-adoption behavior in an IS context [8, 63]. A higher satisfaction level creates a positive attitude towards a technology [64, 65]. There is a plethora of options available as far as smartwatches are concerned; hence, user satisfaction is a salient feature that will decide its continuance usage by the end-users. We define satisfaction as “a type of overall mental belief that the end-users perceive after a prolonged use of the smartwatches”. Improving the satisfaction level will promote brand loyalty and long-term usage [65]. Only the satisfied users will continue using a particular product/service [44]. We hence posit the following:

H2

Satisfaction has a positive effect on the continuous usage of smartwatches.

3.3 Confirmation (CF: ECM construct)

Confirmation can be defined as “the expectations of the end-users that have been met (or not met) after using a certain product/service” [8]. It is, therefore, a perceptual correlation between the pre-expectation and post-expectation of the users. Confirmation can be positive (when post-expectation exceeds pre-expectation) or negative (when pre-expectation exceeds post-expectation) [10]. In other words, if the system does not meet the expectations, it creates a negative post-adoption belief and the users feel dissatisfied. Several research corroborates this relationship in a variety of contexts [44,45,46, 66]. A positive confirmation increases the perceived usefulness of any system, while a negative confirmation reduces the same [67, 68]. Accordingly, we hypothesize the following:

H3a

Confirmation has a positive effect on satisfaction;

H3b

Confirmation has a positive effect on perceived usefulness.

3.4 Hedonic motivation (HM: non-ECM construct)

The concept of hedonic motivation was first introduced as a part of the UTAUT2 model in the context of IT usage [69]. For the present context of smartwatches’ hedonic motivation is defined as “the extent to which the end-users perceive the smartwatches to be a pleasure to wear and provide recreational facilities”. Several empirical studies are reported in literature that investigates the relationship between the hedonic aspect and the corresponding user perception [4, 33, 34, 70]. Other studies have linked it to perceived usefulness [71,72,73]. Some researchers suggest that hedonic motivation has a strong impact on the end-user satisfaction level, especially in case of those technologies/services that are used for fun, rather than performance improvements (e.g., smartphones, online gaming, instant messaging services, etc.) [74,75,76]. In fact, several TAM-related studies have established the importance of hedonic motivation on IT usage [69, 77, 78]. In the context of mobile and wearable devices also, this factor has been investigated in previous research and found to be a prime determinant of user acceptance [33, 38, 79]. The users may wish to use the smartwatches not only for productive-utility purposes, but also for their own entertainment. Accordingly, we propose the following:

H4a

Hedonic motivation has a positive effect on the perceived usefulness;

H4b

Hedonic motivation has a positive effect on the continuous intention to use smartwatches;

H4c

Hedonic motivation has a positive effect on the satisfaction.

3.5 Self-socio motivation (SsM: non-ECM construct)

This construct is a combination of self-motivation (intrinsic) and social motivation or subjective norm (extrinsic). Self-motivation refers to the motivation to do something due to inherent satisfaction [80, 81]. Subjective norm refers to the perceived expectations from others (close friends, relatives, family members, etc.) that influence a user to perform a particular behavior or task [82]. We anticipate that the continuous usage of smartwatches can be affected by both self-motivation and subjective norm. The role of subjective norm into technology acceptance has been addressed by many previous works [14, 39, 83,84,85]. However, in the context of smartwatches, self-motivation can be an important factor too that affects the usage scenario. This finding is attributed to the ethnographic study, which revealed that the users use the smartwatches to track their fitness and have their own set of goals (in terms of number of steps walked per day, calories burned, average running speed, etc.) that they try to surpass. Thus, they are often motivated to beat their own performance. Similarly, the users also felt motivated when after a period of inactivity (30 min or 1 h), the smartwatches kept on reminding them to walk. We confirm these from user quotations like ‘I try to use my smartwatch as much as I can, as it helps me getting better and track my overall fitness’. Also, ‘Every day in the morning while running, I try to surpass my previous record that encourages me to stay fit’. Therefore, the self-socio motivation can be defined as “the joint effect of inspirations from one’s own self and his/her social environment that motivates a prolonged smartwatch usage”. Hence, we propose the following:

H5

Self-socio motivation positively influences the continuous usage of smartwatches.

3.6 Perceived privacy (PP: non-ECM construct)

Privacy is a serious concern for any information-oriented service [86, 87]. With the advent of the IoT era, it has become an even more critical issue that is demonstrated by the recent Mirai malware, which compromised connected devices and conscripted them into a botnet, disrupting the Internet for millions of people. Smartwatches are capable of collecting, managing, monitoring, and analyzing all the personal data they gather from the users. These data contain private and sensitive information such as logs of personal activities, health status, financial information, etc. Previous studies have shown that lapses in security and privacy can jeopardize the continuance use and user retention [88,89,90]. For the present context, perceived privacy is defined as “the state of mind of the smartwatch users where they fear that their personal data will be lost and privacy will be infringed upon”. We posit the following:

H6

Perceived privacy negatively influences the continuous usage of smartwatches.

3.7 Perceived comfort (PC: non-ECM construct)

Previous research has demonstrated the effect of screen size on the user perceptions in various cognitive and affective domains [91, 92]. The same findings can be extended for smartwatches also [38]. The aesthetic appeal of smartwatches affects their usage intention, as they are a mixture of technology and fashion [35, 93, 94]. At the same time, people generally prefer wearing items that feel good, such as wearing comfortable clothes when at home, or using warm blankets while watching television [95, 96]. The same can be translated to smartwatches also, which emerged from the ethnographic study revealing it to be one of the prime issues that can affect the smartwatch continuous usage intention. The participants were particularly concerned about the weight of the smartwatches, and the quality of material used in the wristband. For example, one participant commented ‘Unfortunately, this particular model is too bulky and does not fit under my shirt’. Another one said ‘The material of the wristband creates rash on my skin and I cannot use it for long’. Normally, the users want to wear smartwatches for a prolonged time-period, and hence priority should be given to the usage comfort. Therefore, we propose that when people wear smartwatches, factors such as device weight, and size should matter. Perceived comfort is defined as “the users’ evaluation of the physical comfortability of the smartwatches (in terms of weight, and size) after their prolonged use that can affect their continuous usage”. We hypothesize the following:

H7

Perceived comfort positively influences the continuous usage of smartwatches.

3.8 Battery-life concern (BlC: non-ECM construct)

The effect that battery life can have on the user experience has been discussed in the context of smartphones [33, 97, 98]. This factor is even more important in the case of wearables like smartwatches because the users tend to wear them all the time for various types of purposes. This fact is evident from our ethnographic study. One user reported that I got to charge my smartwatch just every day. Hope it could give me at-least a week of battery life.’ Another one said ‘Maybe I will stop using my smartwatch and just go back to my cheap tracker which gives a month’s usage on a single charge’. Overall, the users seemed to be extremely dissatisfied regarding the frequent charging requirements of the smartwatches when using all the functionalities. Therefore, we anticipate that the users have a strong negative perception regarding the current state of battery-life that makes them dissatisfied and might prompt them to discontinue their use. Current technology adoption models or theories, however, do not include this factor to the best of our knowledge. Battery-life concern is, therefore, defined as “the users’ concerns regarding the battery longevity of the smartwatches (since one full single charge) when using all the features and functionalities to their full potential”. Henc, we propose the following:

H8a

Battery-life concern has a negative effect on satisfaction;

H8b

Battery-life concern has a negative effect on the continuous usage intention of smartwatches.

3.9 Perceived accuracy and functional limitations (PAFL: non-ECM construct)

We propose perceived accuracy and functional limitations as an important predictor for evaluating the continuous usage intention of smartwatches, which covers the current technological issues and lack of certain features, which are important to the users. In case of mobile devices, the users expect the technology to work smoothly without any errors [99]. However, the ethnographic study revealed something different. Some of the users were concerned because they felt that their smartwatches were recording inaccurate data and had certain disorders. For example, one of the participants said ‘Once, I received a notification that I have reached 10,000 steps, while I was driving my car- when I wasn’t walking’. Another one reported ‘I usually walk every day from home to university following the same route, but my watch displays quite a bit different data in terms of the number of steps walked’. Certain feature/functional constraints were also reported by the users. The limited extent of waterproofness was a frequent issue with complaints like ‘I cannot use my smartwatch while swimming because it is only water repellent and not waterproof’. Similarly, there were issues related to the limited recognition of fitness activities by comments like ‘It is really pretty bad that my expensive smartwatch cannot recognize my treadmill movements, which I thought would have been easy to detect’. To summarize, the user concerns were both regarding the accuracy of measurements and lack of certain features in the smartwatches. The dual effect of these two concerns can make the users unsatisfied so that might prompt them to discontinue their usage after a certain period of time. In fact in some previous studies, researchers have also found evidence of relationship between the perceived accuracy and the usage intention for applications like mobile-mapping services, location-based advertising, product recommender systems, etc. [100,101,102,103]. The users want the smartwatches to act as their all-time health companion by providing real-time health and fitness tracking data. However, incorrect and inconsistent measurement of the various vital body parameters and daily activities can limit the usefulness of the smartwatches, which can in turn negatively affect their continuous usage. The relationship between accuracy and usefulness has been established by previous research also in the context of recommender systems and Internet based e-banking [104, 105]. We anticipate that the same relationship will hold true in case of smartwatches also. For this work perceived accuracy and functional limitations is defined as “a negative perception in the users’ mind regarding the accuracy of the smartwatches in measuring various health parameters (e.g. step counts, heart-rate, etc.) and lack of certain features that the users’ consider to be important”. The following hypotheses are proposed:

H9a

Perceived accuracy and functional limitations negatively influences the continuous intention to use smartwatches;

H9b

Perceived accuracy and functional limitations negatively influences the end-users’ satisfaction level;

H9c

Perceived accuracy and functional limitations have a negative effect on the perceived usefulness of the smartwatches.

The final measured construct is continuous usage (CU) of smartwatches. Figure 3 shows the proposed research model.

Fig. 3
figure 3

The extended ECM model of continuance intention to use smartwatches

Full size image

4 Research methodology

4.1 Data collection and sample characteristics

The data have been collected through an online survey distributed to the participants via Google Forms. In order to improve the credibility of the research all the participants who were selected for the survey were actual smartwatch users (using for at least 1 year) apart from the 42 participants who took part in the initial ethnographic study. In total 330 smartwatch users took part in this survey. After eliminating 18 responses with incomplete or invalid data, a total of 312 valid responses are retained for data analysis. Table 4 summarizes the demographic statistics of the sample.

Table 4 Demographic statistics of the sample
Full size table

4.2 Measures

The research constructs were measured using validated multi-item scales adapted from previous studies as well as self-developed scales. Specifically, the constructs (PU, ST, CF, CU, HM, and PP) were adapted from previous studies with relevant modifications in the wordings of the questionnaire to fit the present research context (the relevant studies are presented in Table 5). For the SsM construct, the subjective norm aspect has been captured from previous work, while the self-motivational part has been self-developed. Likewise, for the constructs (PC, BlC, and PAFL) self-developed measures are used because we could not find any use of these in prior works. A seven-point Likert scale has been used for all the items ranging from “strongly disagree” (1) to “strongly agree” (7). The complete list of items used in this study is reported in Table 5 along with the relevant descriptive statistics.

Table 5 Measurement items and descriptive statistics
Full size table

When the initial set of questionnaires was developed, a pilot survey was conducted among the internal researchers of the university along with a few students for pre-testing and refining the questionnaire if needed. Five experienced researchers and five students took part in this pre-test, all of them long-term regular smartwatch users. Priority was given to the wording, completeness, clarity, and the appropriateness of the research instrument used. Based on the feedback obtained from the pilot-testing phase successive iterations were made in framing the questionnaire before it was finalized. Especially in the case of self-developed constructs the items that are found to create ambiguity during the feedback were modified and rephrased. During pre-test rephrasing was carried out to enhance the questionnaire clarity, though none of the items were dropped. The data obtained from the pilot-testing phase were not used for analyzing the results in order to avoid the problems of skewing.

5 Results

For this work, we have used SPSS 17.0 to conduct the Confirmatory Factor Analysis (CFA) and Smart PLS 3.0 for carrying out the Partial Least Squares Structural Equation Modelling (PLS-SEM).

5.1 Measurement model

CFA has been performed to test the psychometric quality of the constructs that are a part of our research model. In particular, we test for the internal consistency, convergent validity, and discriminant validity of the constructs. The internal consistency of the questionnaire used was measured by the Cronbach’s alpha values. These are shown in Table 6. For all the constructs, the values of Cronbach’s alpha obtained are greater than the threshold level of 0.7, which suggests a high degree of internal reliability [111].

Table 6 Results of internal consistency and convergent validity tests
Full size table

The results of the convergent validity tests have also been reported in Table 5. Two conditions were tested in relation to convergent validity:

  • The factor loading of every item measuring a particular construct was calculated and found to be greater than 0.6. This is the first minimum requirement for the convergent validity test to pass [112];

  • The Average Variance Extracted (AVE) was also calculated for every construct, and found to be greater than 0.5, which is the second test for convergent validity [113]. Therefore, the mean variance shared between the latent variable (construct) and its indicators (items) is greater than 50%. When AVE is greater than this threshold the variance explained by the items is greater than the variance arising from the measurement error.

The discriminant validity of the constructs was also tested to ensure that two measures, which are not supposed to be related, are in fact unrelated. Table 7 presents the relevant results. While examining the discriminant validity, the square root of the AVE for every construct should be greater than the correlational values that it shares with the other constructs [114]. This is exactly what happens in our case (the diagonal elements in Table 7 represent the square root of AVE).

Table 7 Correlation of the constructs and test for discriminant validity
Full size table

5.2 Structural model

In order to test for the significance level and obtaining the path coefficients, the bootstrapping procedure was followed [115]. In the bootstrapping method, sub-samples are created with randomly drawn observations from the original set of data (with replacements). The sub-sample is then used to estimate the PLS path model. The process is repeated until a random number of large sub-samples have been created. We used a maximum iteration value of 300 that gives an optimal performance [115]. In case of PLS-SEM, the goodness-of-fit (GoF) index is calculated in an alternative way as outlined in [116]. The AVE and \({R^2}\) values obtained from the structural model are used for calculating the GoF index as per Eq. (1).

$${\text{GoF}}=~\sqrt {({\text{average}} - {\text{AVE}}) \times ({\text{average}} - {R^2})} .$$
(1)

We obtain a GoF value of 0.65, which is greater than the recommended value of 0.36, thereby showing the model validity [116, 117].

Results show that apart from the effect of hedonic motivation on perceived usefulness (hypothesis H4a) all the other proposed hypotheses are true. As a part of the core ECM model, the effect of confirmation on satisfaction (H3a, β = 0.475, t value = 5.613, p < 0.01) and perceived usefulness (H3b, β = 0.646, t value = 15.129, p < 0.001) are supported. The effect of satisfaction on continuance intention is greater than perceived usefulness as evident from their respective standardized β values. Perceived accuracy and functional limitations are the greatest predictors of continuous usage (β = − 0.794) followed by battery-life concern (β = − 0.640), perceived comfort (β = 0.501), perceived privacy (β = − 0.494), self-socio motivation (β = 0.392), satisfaction (β = 0.358), hedonic motivation (β = 0.314), and perceived usefulness (β = 0.289). The user-satisfaction is affected by both hedonic motivation (β = 0.573) and perceived accuracy and functional limitations (β = − 0.396), the effect of the former being stronger than the latter as evident from the respective β values. In addition, perceived accuracy and functional limitations have a weak effect on the perceived usefulness (β = − 0.221). We also analyze the \({R^2}\) values in order to assess the explanatory power of the structural model. Overall, the model accounts for 64.9%, 58.6%, and 41.2% of the variance in continuous usage, satisfaction, and perceived usefulness respectively. The results of the hypothesis testing are summarized in Table 8 and Fig. 4.

Table 8 PLS-SEM path analysis and test statistics
Full size table
Fig. 4
figure 4

PLS results of the structural model

Full size image

6 Discussion and implications

In this work, we have developed a comprehensive research model having its roots in the expectation confirmation model that successfully explains the end users’ continuous usage of smartwatches. Unlike most of the previous works that focus on the behavioral intention of the users, we concentrate on a long-term post usage experience of actual smartwatch users in order to identify any limitations/deficiencies that these devices may have, which in-turn may affect their continuous usage. The use of an ethnographic study supplemented by a literature review makes this work unique as it enables us to identify the exact factors that correlate the users’ pre vs. post expectations of using smartwatches. This is evident from the emergence of new factors like perceived comfort, battery-life concern, and perceived accuracy and functional limitations, which are investigated for the first time in this work. In fact, the SEM results suggest that perceived accuracy and functional limitation has the maximum impact on continuous usage, followed closely by battery-life concerns, both of which are proposed for the first time in this work.

In the era of IoT and ubiquitous computing, where there is stiff market competition due to new products being launched frequently, it is extremely important to find out the decisive factors that will ensure customer retention for particular brands, as well as enable their long-term success. Several findings follow from this research, which is next discussed separately under theoretical contributions and managerial implications.

6.1 Theoretical contributions

The primary contribution of this study is the modification of the ECM framework in the context of smartwatches. The ECM framework has been very effective in the past for evaluating the continuous usage of a particular product/service among the end users (Table 3 gives the details). However, keeping in mind the dynamics of an IoT environment, where new technologies and products evolve at a very fast pace, the ECM framework needs to be suitably modified to fit the current context. By doing so, we extend the generalizability of the ECM framework from a general IS usage context to a more dynamic and volatile IoT domain.

Our proposed model includes novel factors that drive the long-term actual usage of smartwatches that can also be extended to other similar wearable devices. An extensive literature review shows certain critical gaps in the context of wearable devices (smartwatches in particular), about their continuous usage. In fact, theoretical studies assessing the continuous usage of wearables are relatively rare and this is one of the first ones that take into account the actual smartwatch users. Hence, all the factors that are involved in our model assess the actual continuous usage as against the usage intention that is critical for evaluating the long-term success of smartwatches.

We redefine the concept of usability to fit the smartwatch context. Specifically, hedonic motivation and self-socio motivation have been added to measure the continuous usage, apart from the perceived usefulness. In line with previous studies, perceived usefulness affects the continuous usage of the smartwatches in a positive way [59,60,61,62]. However, a low β value between the two factors is indicative of the fact that the users are concerned about the various functionalities provided by their smartwatches. From a utilitarian point of view, the users want to use the smartwatches for tracking and monitoring their daily fitness activities and setting up some long-term health goals in order to improve their QoL. However, the current state of smartwatches both in terms of hardware (quality of sensors used, variety of sensors, etc.), and software quality (accuracy of the prediction algorithms, quality of the applications, etc.) leaves the users disappointed. Therefore, although they have a positive frame of mind to use the smartwatches for leading a healthy lifestyle, yet it can affect their continuous usage if proper product improvement is not done. This fact is also confirmed via the negative relationship between perceived accuracy and functional limitations with the perceived usefulness that is also found out to be true.

In line with previous studies, hedonic motivation affects the satisfaction level [74,75,76,77,78] as well as the continuous usage [69, 77, 78] in a positive way. However, in contrast to previous research, hedonic motivation does not affect the perceived usefulness [71,72,73]. These observations from the theoretical model suggest that the users have a non-overlapping perception into the utilitarian and hedonic aspects of the smartwatches. To the end-users smartwatch utility and hence usefulness is related to the recording of their daily health and workout-related parameters, receiving alerts and notifications that make them depend lesser on their smartphones. The hedonic aspect is an entirely different scenario where the users enjoy changing the watch-faces, playing games or browsing through some applications. Therefore, in the smartwatch context, we can extend the previous results of a positive effect of hedonic motivation on the user intention to an actual continuous usage scenario [52, 60, 108].

Similarly, the effect of self-socio motivation on continuous usage is also found out to be true. This illustrates two facts. First, in line with previous studies subjective norm plays a positive role in shaping the user mindset [83,84,85]. Second, there is a positive effect of self-motivation on the continuous usage. The smartwatches are able to motivate the people effectively to exercise more and lead an active lifestyle.

Perceived comfort has a significant positive influence on the continuous usage. As explained above, smartwatches are effective in promoting a health- and fitness-oriented lifestyle among the users. However, in order to do so, the users need to wear the smartwatches for a prolonged duration. Therefore, device ergonomics is an extremely important issue that needs to be taken care of. Apart from the aesthetic appeal that affects the users’ psychology (established by previous research [35, 38, 93, 94]), emphasis must also be given on the usage comfort keeping in mind the intentions of the users to wear smartwatches for an extended period of time.

Like other information-oriented services, in case of smartwatches also privacy is an important determinant that negatively affects continuous usage. The users are concerned with the privacy of the health and other confidential data that the smartwatches are capable of capturing.

Battery-life concern is another aspect that negatively affects continuous usage of smartwatches. As previously mentioned, the users tend to use their smartwatches for an extended time frame. However, the current state of battery technology limits their real-life smartwatch usage either by wearing them for a shorter span (facilitating charging when required) or by limiting their functionalities (using lesser features can prolong battery-life).

Likewise, the users are also concerned about the accuracy as well as certain functionalities of the smartwatches. This is the most significant determinant that negatively affects the continuous usage (with the highest β value). Often, inconsistent health readings from the smartwatches frustrate the users and leave them unsatisfied. This is evident from hypothesis H9c that poses a negative relationship between perceived accuracy and functional limitations, and the satisfaction level. Therefore, the technology behind smartwatches (in terms of sensors that are more accurate as well as sensing algorithms) must be improved so that people can integrate them in their fitness-oriented lifestyle.

6.2 Managerial implications

Apart from one, all the hypothesized relationships are confirmed in this study, presenting empirical evidence for their applicability in a real-world perspective. There is a very stiff competition in the wearable sector and hence companies must not only address the current loopholes, but also come up with innovative strategies from product design to marketing to improve their success. Our study should help to achieve this goal as it takes into account the experiences of actual long-term smartwatch users.

More investments are needed in research by the companies to not only provide additional features and functionalities, but also improve the current measuring accuracy of the smartwatches. Nowadays, these are used extensively for healthcare monitoring and planning fitness goals. However, incorrect measurements can demotivate and discourage the users from their usage leaving them unsatisfied. Therefore, not only the sensor and prediction algorithm accuracies must be improved, but also the companies must target to market the smartwatches as ‘health companions’ that can effectively complement a physician’s role. Accurate measurements of various health parameters like heart rate, blood pressure, body temperature, steps walked, etc. can facilitate physicians to obtain useful health insights. Likewise, innovative health applications can be built, such as remote monitoring of infants or elderly people that can help to boost the product value. Creating products that can withstand the effects of dust/water (IP67/IP68 rating) and are more durable (shock and shatterproof) is also desirable. At the same time, more focus should be given on the product battery life. Keeping in mind, the end-user ergonomics the battery capacity should be maximized. This should be well supported by a holistic battery friendly software (operating system and applications) that does not create frustrations in the users’ mind.

Furthermore, the elements of utility and fun should be another guiding principle in smartwatch development. For users to continue using a smartwatch, it has to provide some forms of pleasurable experience. Therefore, a rich application store that can cater to the users’ diverse needs (both pleasure and utility) should be present. Focus should be given on the development of energy-efficient games for wearables along with other utilitarian applications like providing shopping recommendations when in the vicinity of a shopping center. The dependency of the smartwatches on smartphones for their proper functioning should be minimized as much as possible and they should be allowed to operate as independent standalone devices.

The users’ privacy concerns should also be addressed. Whenever installing any application, the users should be able to give the exact permissions that the application is permitted to have concerning its usage of the various hardware resources. Similarly, in order to gain trustworthiness of the users the manufacturers can address the privacy concerns through a transparent information policy. For example, they can reveal the identity of the third parties or any other stakeholders accessing the data, the purpose of data usage along with the objectives for which the data are used.

7 Conclusion, limitations and future work

In this work, we have presented a systematic approach to understand the continuous usage of smartwatches by the end-users. It extends the ECM framework by incorporating additional factors that capture the unique context of smartwatch usage. A dual prone strategy of literature review and ethnographic study has been used for an effective factor identification. The results reaffirm the role of perceived usefulness, confirmation, and satisfaction on continuous usage. The findings also emphasize the importance of perceived comfort, perceived privacy, self-socio motivation, hedonic motivation, battery-life concern, and perceived accuracy and functional limitations on the continuous usage.

The current study has a number of limitations. First, all the participants come from different Asian countries. However, more participants from across the globe should be included in order to test for any significant differences in continued smartwatch usage. This will also enable to capture the effects of cultural differences.

Second, the use of smartwatches may differ across different demographics, such as age. This study does not take into consideration the age differences. For example, the usage pattern of the young generation may be different compared to older generations. They are most likely to switch across different smartwatch vendors even at the slightest dissatisfaction. Future research can also investigate how usage patterns vary across users from different age groups.

Third, our model has been tested only for the context of smartwatches. Therefore, future research can be carried out to test the same model for other wearable devices like fitness trackers, smart glasses, etc. in order to increase its generalizability.