The major distinguishing factors of the 4IR are rapid technological advances (Syam & Sharma 2018), increased digitisation (Schwab 2016), and increased connectivity (Chukurna & Konak 2018). It is fast reshaping a new business environment (Philbeck & Davis 2018), and thus becoming a major economic issue (Delfmann et al. 2018; Tohanean & Toma 2018). It is altering the competitive landscape by increasing the availability of substitute services (Schwab 2016). As an illustration, in the tourism industry, this era has increased the range of possible choices of destinations (Inversini, Cantoni & De Prieto 2014; Komsic & Dorcic 2016). For marketers, it has delivered tools to build better relationships with customers, and has led to social media marketing measures, which are a necessity to reach more customers effectively, reliably and efficiently (Chary 2014). It is also generally accepted that the 4IR affects customer expectations, customer perceptions, customers’ intentions (Cervellon & Galipienza 2015; Kwok & Yu 2016; Lillqvist & Louhiala-Salminen 2014) and, ultimately, affects their actual behaviour (Schwab 2016). Of interest to the present study, ICT has given rise to online destination reputation (Gössling, Hall & Andersson 2016), which also influences people’s choices (Harvey & Groutsis 2015). This concept (firm’s reputation) has received attention from marketing researchers (Petersen & Lemke 2016; Vizcaíno-González, Iglesias-Antelo & Romero-Castro 2019). For the tourism industry, destination reputation captures tourists’ beliefs about the quality and features of a destination (Fu et al. 2016; Kock et al. 2019), affects customers’ service choice, overall attitudes about services and trust, and influences their visit intentions (Duygun & Mentes 2015). Destination reputation, therefore, has a big impact on tourists’ destination choices, visiting experiences and perceived value (Fu et al. 2016). People share information about their experiences (Cantallops & Salvi 2014; Chan & Guillet 2011), and others refer to online reviews to obtain information on the quality of services provided by a tourism destination (Casalo et al. 2015; Zhu & Zhang 2010). Observational learning theory teaches us that people’s decisions are shaped by other people’s experiences (Chen, Wang & Xie 2011; Miklós-Thal & Zhang 2013). Under the 4IR era, destination centres can develop both good and bad reputations without direct contact with customers. Tourists’ generated content can confirm or refute the online reputation of tourist destinations (De Moya & Jain 2013). In line with Harvey, Groutsis and Van Den Broek’s (2018) classification, reputation for the present study will mean that of Vilakazi Street in Soweto, Johannesburg, in South Africa. Competitiveness and reputation mutually influence each other (Assaker et al. 2015; De Moya & Jain 2013). Therefore, ICT can affect the competitiveness of a destination by enhancing its online reputation:

When customers, directly or indirectly, encounter a brand (Alloza 2008; Brakus, Schmitt & Zarantonello 2009), an experience that can affect their behaviour (Prasad & Jha 2014) is bound to happen (Kumar, Dash & Purwar 2013). It can, for instance, lead to positive word of mouth (WOM, Khan & Rahman 2015), and to an increase in consumers’ trust and willingness to use that brand (Dolbec & Chebat 2013). These impressions and behavioural responses can become a tool to attain a competitive advantage (Clatworthy 2012). For that reason, marketers often attempt to influence them (Kharat & Kharat 2017; Van Der Westhuizen 2018). In the 4IR environment, people can also encounter brands virtually (Pantano 2015). Digital technology can plunge people into a new world of real context (Guerra, Pinto & Beato 2015). This 4IR-enabled virtual experience can generate emotions and social interactions (Guerra et al. 2015), and elicit the same reactions and emotions as those felt during a real experience (Garrozzino & Bergamasco 2010). Today people can experience the content with higher vividness and realism (Bailey et al. 2015; Hsu, Tseng & Kang 2018), which can stimulate a sense of presence in the real world, and thus facilitate the pre-purchase experience (Gül & Gül 2018). Since services can also be experienced pre-consumption (Kharat & Kharat 2017), impressions about them can also be formed pre-purchase (Ismail, 2010):

Behavioural intentions are intentions to visit or revisit Vilakazi Street, recommend Vilakazi Street to friends or others, and say positive things about it and the services it offers (Kiatkawsin & Han 2017; Kim et al. 2018; Lo & Wu 2014; Loureiro et al. 2013). The 4IR can act as both an inspiration and an enabler for tourists (Moeketsi 2019). Law, Buhalis and Cobanoglu (2014) state that travellers who use more information and communication technology have greater purchase intentions, and their online review and subjective experience could affect visits by other people to a destination (Assaker & Mueller 2015). In that way, a destination’s online reputation can influence tourists’ future visit decisions, the intention to further spread positive e-WOM (Chen & Petrick 2015; Li & Petrick 2010), and, ultimately, enhance their loyalty (Chen & Petrick 2015; Oliver 2010). Kim et al. (2018) affirm the same in stating that a destination’s positive reputation has a positive effect on intentions to revisit. Studies have shown that there exists a significant relationship between the reputation of a tourism destination centre, tourists’ decision-making (Altunel & Kocak 2017; Suhartanto et al. 2018) and their behavioural intentions (such as initiating a visit or revisit) (Kaplanidou & Gibson 2010; Oliver 2010). In addition, these intentions can be used as a predictor of future consumption behaviour (Jin, Lee & Lee 2013; Kim 2018; Lai & Chen 2011). Thus, in association with the literature cited above, this study posits the following hypothesis:

Future changes in the tourism market will be under the strong influence of new technologies (Petkovic, Pindzo & Agic-Molnar 2018). This technology will increasingly play a significant role for both consumers and service providers (Gül & Gül 2019; Opara & Onyije 2013).

In the tourism industry, tourists’ visit or purchase intention (PI) is the most vital aspect (Tsiotsou 2006), because it shows the probability, willingness and likelihood to visit an establishment (Rashid 2009), thus providing the best predictor of their planned behaviour (Al-Jubari, Hassan & Linan 2018). It includes consumers’ attitudes, and their motivations (Chaniotakis, Lymperopoulios & Soureli 2010; Shah, Shahzd & Ahmed 2012). Further, PI also offers the likelihood to engage in positive WOM, and a willingness to pay more for services (Hau et al. 2009; Zainor & Mohd 2018). However, tourists’ visit choices involve subjective judgment about destination areas (Blackwell, Miniard & Engel 2001). Consumer behaviour studies show that tourists aim to maximise their utility by choosing an option that best meets their individual travel and vacation needs (Kock, Josiassen & Assaf 2016; Prayag & Ryan 2012). In the absence of real physical experience, the destination’s online presence can influence their choices and visit intentions. Rodrigez-Diaz and Espino-Rodriguez (2018) state that recommendations by friends can be a good variable for perceived value. Tourists can vicariously assess the destination using technology (Fombrun 2012; Walker 2010). Online positive or negative messages about a tourism centre can be a proxy for judging their likely experience , should they visit the destination. Therefore, technology can determine tourists’ mobility decisions:

Satisfaction represents a positive consumer or tourist feeling towards a brand or a destination; therefore, from a psychological perspective, consumer or tourist engagement has a positive effect on satisfaction (Bryce et al. 2015; Zhou, Zhang & Edelheim 2013). Moreover, loyalty represents the willingness of consumers or tourists to re-buy or revisit a brand product or destination consistently (So et al. 2016). Consumer or tourist engagement demonstrates deep commitment to and connection with the brand or destination, which can enhance the consumer or tourist’s feelings and behaviours (Hollebeek 2011; So et al. 2014). An engaged tourist with a destination develops a more favourable attitude, and is, therefore, expected to be more interested to revisit and be more loyal towards a destination (Brodie et al. 2012). Understanding customer perceptions is crucial to attracting customers to purchase (Millar & Baloglu 2008):

Customer satisfaction is the customers’ service interaction-related pleasurable response, based on a comparison of the service outcomes and a reference standard (Lechner & Paul 2019; Oliver 2010). Customers’ encounter with a service can either lead to satisfaction or dissatisfaction (Hulsheger & Schewe 2011; Lechner & Paul 2017). In turn, perceived satisfaction also influences purchase intention (Jaafar, Lalp & Mohamed 2013). People who are dissatisfied with the service would be disloyal (Millar & Baloglu 2008). Further, customer satisfaction has a positive effect on customer retention and profitability (Gupta & Zeithaml 2006), which is also the strongest predictor of repeat purchases (Hussain 2017). Su et al. (2016) identify strong links between service quality, customer satisfaction and customer-company identification, which positively influences repurchase intention. It can, therefore, be concluded that service quality directly influences satisfaction (Han & Hyun 2018) and behavioural intentions (Su et al. 2016). Consequently, this study hypothesises that:

The hypotheses (H1-H6) are represented in the form of a conceptual framework in Figure 1.

The survey used originally designed questionnaires after an extensive literature review (Ashrafi & Mueller 2015; Huang 2003). Measurement items were selected and context-adapted to suit the present study purpose and to ensure the validity and reliability of the measures. Six constructs were included, and all the survey questions measured these constructs. To minimise measurement errors, the researchers utilised multi-item measures for each construct (Churchill 1979). All scale items were scored on a five-point Likert scale ranging from 1 to 5, with 1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree and 5 = strongly agree. For the list of the constructs, operational definitions and the sources they were adapted from, see Table 1. Questionnaire development was in English, as participants were assumed to be English literate. The total number of items in the questionnaires was 37 (see addendum).

To minimise possible bias due to interviewer-participant interaction, participants were told that their partaking is voluntary and anonymous and they were encouraged to state their own personal opinion as truthfully as possible, in accordance with proper academic practice (Podsakoff et al. 2003). To mitigate the biasing effect of the research instrument, the questions asked were simple, specific and concise (Hulland, Baumgartner & Smith 2018). Different variables measuring disparate items were separated across the survey (Tang et al. 2018). Further, the cover letter was structured in a way to make it appear that the measurement of the predictor variables is not related to outcome variables. The researchers also avoided the use of words with many meanings, in line with scholarly recommendations (Podsakoff et al. 2003). Throughout the project, the researchers maintained high standards of values and norms. They obtained permission from the major businesses in the area before data gathering could commence. No data were collected to identify individual participants, and no incentives were offered that could possibly encourage their participation. Issues that might have affected the number of positive responses are the trust level between tourists and research assistants, the costs (in terms of time) and the nature of the research instrument (physical paper-based) (Koundinya et al. 2016). Also, there were no possibilities for using reminders to complete the questionnaires after contact sessions (Bosnjak et al. 2008; Saleh & Bista 2017).

The target number of participants was 400. However, a sample size of 235 was achieved, representing a 59% response rate. This sample size was deemed sufficient after taking into account the analysis method applied in the study. The number satisfies the recommended figure required for SmartPLS analysis purposes, and it provides more than sufficient (a minimum of 200) statistical power for analysis (Hoe 2008). Table 2 summarises the profile of the participants. The majority of participants were male, as evidenced by the 52.8% representation. From the table, it is clear that the people who visited the Vilakazi Street precinct were mostly the aged and highly educated individuals (about 50% of them with master’s degrees). This has a bearing on how to target this segment and to tailor the advertisement messages. The income levels also show that the average visitor is a high net worth individual. Unfortunately, a large number of them preferred not to state their income levels. Many also chose not to state the amounts they spend on tourism.

This study applied partial least squares structural equation modelling (PLS-SEM) for data analysis, due to its numerous advantages over other approaches. The SEM approach can include latent constructs in the analysis while accounting for measurement errors in the estimation process (Ledermann & Kenny 2017) and, most importantly, it provides support for examining and validating hypotheses of causal relationships due to its ability to do away with bias and distortion (Avkiran 2018; Naor & Punj 2015). In particular, PLS-SEM is a more appropriate approach for complicated models or multi-group analysis than traditional SEM based on covariance (Hair et al. 2012). Smart PLS 3.2.6 was applied to analyse both the measurement and the structural models (Kim et al. 2018; Ringle, Wende & Becker 2015), in accordance with Gerber’s two-step approach (Zhaoa, Xub & Wang 2019). Confirmatory factor analysis (CFA) and path analysis were performed simultaneously (Davari & Rezazadeh 2015). Since CFA allows the researcher to test the hypothesis that a relationship exists between the observed variables and the underlying constructs (Suhr 2006), it, therefore, evaluates the measurement model’s reliability and validity (how well the latent variables are measured by the observed variables) (Li et al. 2016). Path analysis is used to investigate the causal relationships among unobserved variables (Chin 1998; Mueser et al. 2017).

The measurement model was first evaluated for reliability, convergent validity and discriminant validity as per scholarly recommendations (Ashrafi & Mueller 2015). Cronbach’s alpha coefficients and composite reliability (CR) values formed the basis for testing the reliability of the measurements. Diedenhofen and Musch (2016) state that Cronbach’s alpha values greater than 0.6 are reliable. However, other researchers use 0.7 as a cut-off value (Cho, Hong & Hyun 2009). All the Cronbach’s alpha values exceeded the recommended higher threshold value of 0.7 (4IR-Env = 0.871; Pre-Exp = 0.906; Tech-Int = 0.875; Reput = 0.891; Sat-Dis = 0.904; Loyal = 0.848). A CR index greater than 0.7 is deemed acceptable (Diedenhofen & Musch 2016). From the results output, the lowest CR value was for loyalty (Loyal) with a value of 0.850, while the highest CR was for satisfaction (Sat-Dis) with 0.900. High values for both Cronbach’s alpha and CR (shown in Table 3) mean that all the measurement instruments used in the study are reliable.

Factor loadings generated with linear structural relations (LISREL) were also used to calculate the average variance extracted (AVE), the values used to check for convergent validity (Fornell & Larcker 1981; Igbaria 1993). The AVE values are expected to be greater than the acceptable threshold of 0.5 (Hamouda & Junoh 2018). In this model, the lowest AVE is 0.5673, which satisfies the limitation.

To check for discriminant validity, the researchers used the inter-construct correlation matrix as recommended (Chinomona & Preterms 2011). Discriminant validity is proved when the value for correlation between variables is less than 1.0 (Hox et al. 2017). Table 4 indicates the inter-correction values for all the variables and they are all less than 1.0, hence confirming a good discriminant validity.

After the reliability and validity checks of the measurement model, the next step was to assess the fit between the data and the proposed structural model. R-square values, t-values and factor loadings were used to assess the CFA model before using the model fit indices. According to Hair et al. (2017) R-square should be greater than 0.25, t-statistics should be greater than 1.96 and the factor loadings should be greater than 0.5. The structural model fitness was assessed using the Chi-square (χ²/df), the normed fit index, comparative fit index, incremental fit index, goodness of fit index, relative fit index and random measure of standard error approximation. Table 5 provides the acceptable threshold values, values for the model and the decisions after comparisons. All the model fit indicators met the acceptable threshold and the data were considered to fit the conceptualised model.

After model fit tests, the next step was to examine causal relationships among latent variables by path analysis (Hair et al. 2017; Henseler, Hubona & Ray 2016; Lefcheck 2016).

Hypotheses are significant at a 95% or higher level (≥ 95%) (Hair et al. 2010; Hastie, Tibshirani & Friedman 2009). The t-statistics have to be greater than 1.96 for the proposed relationship to be acceptable. Table 6 gives a summary of the outcomes and shows the hypotheses, t-statistics, path coefficients and the decisions taken on whether to accept or reject the hypothesis (Shan et al. 2018). The path coefficients demonstrate the strength of the relationships between the dependent and the independent variables (Hsu 2008). Higher path coefficients indicate strong relationships among latent variables (Chinomona et al. 2010). The outcomes of the analysis demonstrated that all six hypotheses postulated were significant with the lowest t-statistical value being 5.480. For the complete list, see Table 6.