It is important for business organisations to have a deep understanding of the various customer segments they target with their online business strategies (Saprikis et al. 2010). To that end, it is important to understand the determinants of online shopping behaviour (Tiwari & Joshi 2020). Market segments can be defined based on generational cohorts (Hume 2010). The generational theory was developed by Strauss and Howe in 1991, who identified five cohorts of generations (Moreno et al. 2017). Each of those cohorts is different in terms of experiences, expectations, history, attitudes, values, world perspectives and lifestyles, and each manifests different buying behaviours as consumers (Levıckaitė 2010). This study focuses on young people and more specifically on the following two generational cohorts: Generation Z (people born mid-1990s) (Levıckaitė 2010) and Generation Y (also known as the Millennials), who represent the people born in the time period 1981–1995 (Junker, Walcher & Blazek 2016).

Generation Z people were born in an era dominated by computers and the Internet and did not have to get accustomed to this technology, and as a result, they place value on technology, speed and individualism (Berkup 2014). From this description, it can be deduced that the technological nature coupled with the speed and convenience of online shopping will naturally appeal to Generation Z people. Even though a significant part of the South African Generation Z population lives in poverty, many have the ability to access the Internet and social media via mobile devices, and therefore, they present a lucrative target market for companies (Duffett 2017). This concurs with the finding of Dabija and Lung (2018) that Generation Z members in an emerging market mostly prefer online shopping using their smartphones.

Millennials can be considered a demographic generation which is also significantly influenced by technology, including the Internet, and they are considered the largest market segment currently. Millennials are also considered a very attractive niche market because of the influence members of this cohort have on the formation of purchasing trends with peers (Moreno et al. 2017). As a market segment, Millennials are also considered highly skilled in using the Internet for product searches, comparison and buying (Naumovska 2017). For the purpose of this study, Generation Z and Millennials are considered young people who constitute the ideal market segment for business organisations targeting online sales growth.

The past few studies have found many factors influencing the online shopping behaviour of young people. In a study of the online retail buying behaviour of Indian college students, Jadhav and Khanna (2016) found the main influencing factors to be availability, competitive prices, comparison, promotions, convenience, customer service, perceived ease of use, attitude, time consciousness, trust and variety seeking. In a similar vein, a study of Greek university students by Saprikis et al. (2010) found the main reasons for adopting online buying to be lower prices when compared with conventional stores, the ease of the online buying process and the wide variety of available products. The main reasons for not buying were found to be security and privacy reasons, a need to physically examine products, preference for buying from brick-and-mortar stores and the non-usage of credit cards (Saprikis et al. 2010). Similarly, a study by Zendehdel et al. (2015) on the factors driving the adoption of online practices amongst Malaysian students found that the perceived risk of online transactions is an influencing factor negating the growth of online buying in the country. The latter study confirmed cost benefits as the main perceived benefit of online buying amongst the youth in Malaysia. A study of young consumers in Australia and the United States of America found a positive attitude towards online shopping as a strong influencing factor. The study also found that young consumers are very familiar with online shopping processes (Dharmesti et al. 2019).

Online shopping is an aspect of e-commerce that can be described as the process of researching and buying goods using the Internet (Varma & Agarwal 2014). In order to understand how business tactics influence customers’ online behaviour, one needs to deconstruct the online buying process into its main phases. For this study, three key phases of online buying have been identified: product search, payment and fulfilment execution. The product search phase entails the accumulation of information by prospective buyers to consider factors, such as prize, size and availability (Riley & Klein 2019). The payment phase can be either owned by the supplier or outsourced to external providers such as PayPal (Ascarza et al. 2017). The fulfilment phase mainly concerns delivery of products to their final destinations in an efficient and effective manner (Ascarza et al. 2017) and includes product returns (Nguyen et al. 2019). Research has indicated that the delivery of products has a strong impact on customer satisfaction in e-commerce (Mofokeng 2021), and that as many as 85% of customers have selected online retailers based on their delivery service reliability and trustworthiness (UPS 2021). This research focuses on the fulfilment part of online transactions because it is an important success factor in e-commerce and because of the wide variety of related factors that influence the behaviour of consumers buying online.

Last-mile delivery can be explained as the activities required for physical delivery to the final destination (Olsson, Hellström & Pålsson 2019), and it is a critical success factor for online business. The findings of an international study conducted by Capgemini (2019) suggested that last-mile delivery is becoming increasingly important, with 40% of consumers considering delivery services as a necessary requirement when selecting food and grocery retailers for purchases. The study also found that the provision of last-mile services constitutes 41% of overall supply chain costs, which is more than double the cost of any other category of logistics costs in the supply chain (Capgemini 2019). Last-mile delivery costs are variable, implying that total last-mile delivery costs as a component of total logistics costs will increase as the volume of online sales increases (Capgemini 2019). Furthermore, delivery timeframes are shortening on the back of consumer demand, and shippers are compelled to find efficient ways to operate the last mile (Hu & Acar 2018).

Service quality variables related to last-mile delivery are wide and varied. Critical service aspects include delivery information or options that consumers want before they decide to order products and include information about carriers, shipping dates and delivery time slots (Nguyen et al. 2019). Various studies have measured the effects of various aspects of last-mile delivery on online customer behaviour. Nguyen et al. (2019) found that in their evaluation of online delivery services, consumers attribute the greatest importance to delivery fee, followed by delivery speed, time slot, delivery date and daytime or evening delivery. Rao et al. (2011) found that delivery options provided with online transactions contribute significantly to service quality and customer satisfaction. Agatz et al. (2011) found that consumer behaviour was affected by the number of delivery time slots offered.

Service quality can broadly be explained as the disparity between customer expectations and customer perceptions of the service (Luke & Heyns 2017). According to Rust and Oliver (1994), the perception of service quality is grounded in customers’ comparison of prior experiences of excellent service encounters. Bitner and Hubert (1994) support this and posits that service quality is a customer’s assessment of an entity’s overall excellence or worth. This suggests that, in essence, service quality entails a level of relativity where the consumer compares their service experience with prior experiences of similar types of services. There is an array of approaches that are used to measure service quality, including several national indexes that are centred on customer perceptions and expectations (Andreassen & Lervik 1999; Johnson et al. 2001); service quality indexes (SQIs) that are grounded on random utility theory; and discrete choice models (DCM). Service quality indexes differ from customer judgement ratings because they use choice data for their measurements (Eboli & Mazzulla 2007; Hensher, Stopper & Bullock 2003). Customer service is also measured through customer satisfaction indexes (CSIs), which assess user opinions through a numeric scale (Eboli & Mazzulla 2009; Hill, Brierley & MacDougall 2003). The SERVQUAL methodology, developed by Parasuraman, Zeithaml and Berry (1985), has also been used extensively across many industries (including retailing, tourism, banking, education and transport) to measure and compare a customer’s perceived service expectations with their perceptions of actual service experience (Barabino & Deiana 2013; Luke & Heyns 2020; Morton, Caulfield & Anable 2016). The SERVQUAL model was adapted by Zeithaml, Parasuraman and Malhotra (2002) to measure customers’ perception of the quality of certain aspects of e-services (such as website quality). Lee and Lin (2005) also developed an E-SERVQUAL framework that investigated the association between the overall service quality and specific quality characteristics; however, a limitation of these models is that they exclude last-mile distribution. Xing and Grant (2006) developed an electronic physical distribution service quality (e-PDSQ) framework that focussed on B2C e-commerce models, which revealed that price is a key online purchasing measure. A study by Xing et al. (2011) explored retailers and logistics service providers’ physical distribution service quality challenges (LSPs).

The purpose of this study was to ascertain the online shopping preferences of young people before and during the COVID-19 pandemic, gauge the fulfilment-related service quality perceptions of online consumers and ascertain their most preferred last-mile delivery options in the Gauteng region.

In order to acquire an informed understanding of consumers’ online shopping issues, a quantitative research approach was used. The research instrument, a structured self-administered online questionnaire, was developed through a review of the relevant literature and a previously developed survey (Brink et al. 2019) to assess young people’s online shopping behaviour and last-mile delivery preferences for their online purchases.

To obtain the research data, a non-probability sampling strategy was applied. In population research, convenience sampling (or opportunity sampling) is frequently employed because of the relative ease of collection, cost-effectiveness, simplicity, accessibility and the willingness of respondents to participate in the survey (Stratton 2021; Saunders, Lewis & Thornhill 2019). However, non-probability sampling is inclined to be biased, and generalised interpretations of research outcomes must be deliberated on with caution (Zikmund et al. 2013; Saunders et al. 2019). The research data were collected from two data collection campaigns that were conducted in 2019 (before the COVID-19 pandemic) and in 2020 (during the COVID-19 pandemic). The original aim of the 2019 data collection campaign was to collect research data to investigate the online shopping preferences and service quality perceptions of young people shopping online. Before this research project was concluded, there was an outbreak of COVID-19 leading to a global pandemic. The researchers decided to extend the original research design to include the potential impact of the pandemic on the online shopping preferences and perceptions of young consumers. Accordingly, a second data collection drive was conducted to gauge the online shopping behaviour and fulfilment-related service quality expectations of young consumers during the COVID-19 pandemic.

The research instruments used in both data collection campaigns consisted of four parts. In the first section, respondents were requested to provide demographic information, including gender, age group, student and employment status; the second part assessed the respondents’ online shopping behaviour and patterns; part three assessed the respondents’ last-mile delivery preferences, and the last section covered the respondents’ perceptions of the key service quality variables that influenced their online shopping decisions. However, in the 2020 questionnaire, two additional questions were included, specifically to gauge how the initial national lockdown period instigated by the COVID-19 pandemic changed the respondents’ online shopping activities and impacted the purchasing of discretionary products.

For the 2019 data collection campaign, a chain-referral sampling technique (or snowball sampling), which was initiated amongst a small group of postgraduate students at the University of Johannesburg (UJ), was used to recruit respondents by distributing the research instrument amongst their acquaintances or family members via emails or various social media platforms (such as WhatsApp). Although the sampling technique encounters some criticism, such as selection bias, low response rates, lack of external validity and generalisability, it has become popular because of its flexibility, quickness and relatively low costs (Kaliszewski et al. 2021; Parker, Scott & Geddes 2019). The target population included consumers between the age of 18 and 45 years who conducted online shopping in Gauteng, South Africa.

For the 2020 data collection campaign, a self-administered online questionnaire was distributed amongst students studying logistics and supply chain-related courses at the UJ. The survey was circulated to approximately 4500 tertiary students in June 2020 and completed by 446 respondents, with a response rate of around 10%.

The data collection phases provided a total of 461 usable responses, 268 from the 2019 survey and 193 from the 2020 survey. Although the sample size could be considered a limitation of the study, the authors are of the opinion that the results do offer an important contribution to understand the impact of the COVID-19 on the online shopping behaviour and service quality perceptions of consumers in South Africa. The research data were analysed using the Statistical Package for the Social Sciences (SPSS) for Windows, version 27.

When asked to indicate the frequency of their online purchases, most of the respondents, in both surveys, indicated that they only occasionally carry out online shopping, whilst the 2020 survey respondents revealed a noticeable increase (4%) in only shopping online once a week (see Figure 1). Based on the study conducted by Deloitte in 2020 on South African online consumer behaviour, more than 70% of the respondents (aged over 18 years) shopped online at least once a month (Deloitte 2021). According to Business Insider SA (2021), South Africa’s online retail business has more than doubled in recent years from R14 bn in 2018 to over R30 bn in 2020. Research conducted by Mastercard (2020) indicates that nearly 68% of SA consumers shopped more online since the start of the COVID-19 pandemic.

The research results indicate that the respondents mostly purchased non-perishable products (37%) and airtime or data (22%) during 2020 (see Table 2). It seems that during the early lockdown phases of the COVID-19 pandemic, respondents under the age group of 18–34 years purchased more non-perishable and airtime or data than during 2019. A considerable increase in online shopping for groceries is also noticeable in the 25–34 years age cohort. The noteworthy decrease in online purchases of electronic equipment could be the result of limited availability and inflated prices for integrated circuits (i.e. semiconductor chips) experienced globally (Casper et al. 2021). The significantly reduced spending on entertainment highlighted by the 2020 survey is probably the result of the stringent lockdown regulations imposed by the South African government, which brought most of the entertainment industry to a complete stop.

The respondents of the 2020 survey (during the COVID-19 pandemic) were requested to indicate any changes in their online shopping behaviour. Most of the respondents (32%) indicated that at that time, the COVID-19 pandemic and related restrictions did not have a significant impact on their online shopping behaviour. Approximately the same percentage of respondents (25%) increased or decreased their online shopping frequency during the COVID-19 pandemic. However, it is observed that 29% of the more mature age group made their first online purchases during this time, which was somewhat higher than the other age cohorts (see Table 3).

The Deloitte (2021) report on South African online consumer shopping insights, however, presents a slightly different picture, with 29% of respondents indicating no change in shopping frequency, 9% indicating less online shopping and 62% stating that they would increase their online purchases.

Respondents were also asked to select all the factors they considered to be important when deciding on an online service provider. The respondents of both surveys indicated delivery costs (2020% – 69.4%, 2019% – 69.8%) as the most important factor, followed by waiting time to delivery and order accuracy (as the second and third most important factors, respectively (see Figure 2). When considering the different age cohorts, 42% of the 18–24-year-old respondents from 2020 (during the COVID-19 pandemic) considered order accuracy as an important factor, compared with only 26% in 2019 (before the COVID-19 pandemic). However, when reflecting on the 25–34-year-old respondents, 51% of the respondents from 2019 and only 38% of the respondents from 2020 considered order accuracy to be important.

According to the Deloitte (2021) survey, 86% of South African online consumers also viewed low delivery fees as an important factor in choosing an online retailer. Other factors identified include effective checkout (88%), reassurance regarding returns and refunds (83%) and trust in the brand.

When asked to indicate their preferred delivery time slot, the majority (66%) of the respondents from the 2020 sample preferred to receive their deliveries in the morning (between 8:00 and 12:00), whilst the majority (55%) of the respondents from 2019 preferred to receive their deliveries between 12:00 and 18:00, as shown in Figure 3. The variance is probably attributable to the differences between the respondent profiles of the two samples, with 2019 comprising of mostly younger, full-time student respondents as opposed to slightly older, employed and part-time students and not necessarily because of the COVID-19 pandemic.

When asked to indicate their most preferred delivery choice option, the majority of the respondents, 77% and 74%, from the 2020 and 2019 samples, respectively, preferred attended delivery. The noticeable increase in 2020 could most possibly be attributed to the fact that many individuals were working from home at this time and that most students were studying online and off-campus.

Respondents were also asked to rate, on a five-point Likert-type scale (where 1 = never and 5 = always), the possibility of them making use of the different delivery options. As depicted in Figure 4, attended home or work delivery is rated the highest likelihood of usage before (M = 4.05, SD = 1.05) and during the COVID-19 pandemic (M = 4.22, SD = 0.87). The higher mean rating in the year 2020 is also attributed to the South African government’s lockdown conditions imposed at the time. The results also show that the 2020 respondents continued to be somewhat hesitant to make use of unattended delivery services (M = 2.29, SD = 1.11), with 59% of the respondents indicating that they never or rarely made use of this delivery option. All the delivery options, except attended delivery, were ranked below the mid-point value of 3, further indicating that the most desired delivery option was attended delivery.

When asked to assess on a five-point Likert-type scale (where 1 = very unimportant and 5 = very important), the importance of various delivery-related factors that could impact their decision to shop online instead of going to the brick-and-mortar shop, the respondents of both samples indicated that easy returns (2020 – M = 4.24, SD = 1.36; 2019 – M = 4.57, SD = 0.99), followed by delivery in preferred time slot (2020 – M = 4.07, SD = 1.35; 2019 – M = 4.40, SD = 1.08) were the most essential factors that influenced their decision. For both surveys, all the factors presented were considered important as they were all ranked above the mid-point value of 3. Delivery in the preferred time slot and same-day delivery seemed to be somewhat less important during the COVID-19 pandemic, which could be attributed to the fact that respondents were possibly working or studying from home at the time and thus more flexible when receiving online deliveries (Figure 5).

These results are also supported by Deloitte (2021), who affirms that the majority of the surveyed respondents believe that high delivery fees (86%) and policy on returns and refunds (85%) are key factors that influence South African online customers’ decisions.

The participants were also requested to assess on a five-point Likert-type scale (where 1 = not important and 5 = extremely important) the importance of various fulfilment-related online shopping service quality variables. The measurement scale for the service quality variables was evaluated to determine the reliability and internal consistency. The internal consistency measurement indicated overall Cronbach’s α values of 0.877 and 0.857 for 2019 and 2020, respectively, which revealed that the reliability of both surveys was perceived as good (Field 2018).

Table 4 depicts the descriptive statistical results and the importance of ranking of various service quality variables for both 2019 and 2020. The top four service quality variables for both samples were order conditions, order completeness, order accuracy and returns channel options. Likewise, the six least important service quality variables were also the same for both samples, of which most relate to tangible aspects. For both sample years, the colour of product packaging and the size of the delivery vehicle were the only two variables that were ranked below the mid-point value of three, indicating no meaningful influence on respondents’ online shopping behaviour.

The Mann–Whiney U test was conducted to assess any statistical differences between the service quality variable scores for the 2019 (before the COVID-19 pandemic) and 2020 (during the COVID-19 pandemic) samples. The results are shown in Table 5. The Mann–Whitney U test revealed significant differences in three variables, viz. colour of product packaging in 2020 (Md = 2, n = 164) and 2019 (Md = 2, n = 249), U = 17 484, z = −2.57, p = 0.01, r = 0.13; method of delivery, in 2020 (Md = 4, n = 164) and 2019 (Md = 4, n = 250), U = 17 015, z = −3.08, p = 0.02, r = 0.15; and substitutions offer, in 2020 (Md = 3, n = 160) and 2019 (Md = 4, n = 249), U = 16 811, z = −2.77, p = 0.006, r = 0.14.

An exploratory factor analysis (EFA) was completed to reduce the number of variables that potentially influence the service quality perceptions of online shoppers. Guidelines obtained from the literature to determine the appropriateness of the sample size for factor analysis vary to a great extent, and according to Beavers et al. (2013), it can be portrayed by the minimum number of cases or subjects-to-variables (STV) ratio. According to Field (2018), a sample of at least 300 would most likely provide a stable solution; however, other researchers have suggested that relatively small sample sizes could provide adequate results. The recommendations vary from a minimum STV ratio of five cases for each item (Bryant & Yarnold 1995); (Suhr 2006); (Hair et al. 2010) to a minimum sample size of at least 100 cases (Gorsuch 1983), at least 150 cases (Hutcheson & Sofroniou 1999) or at least 200 cases (MacCullum et al. 1999). Costello and Osborne (2005), who surveyed over 300 EFA articles, found that a large percentage of the observed research used relatively small sample sizes. Over 40% of the researchers conducted their analyses with an STV ratio of less than 5:1. However, according to Zhao (2009), it is difficult to state if the total sample size or the STV ratio should be used as a general rule-of-thumb to determine the minimum sample size, as it is dependent on other aspects of design, including communalities of the variables’ degree of overdetermination of the factors and item loadings. Costello and Osborne (2005:4) asserts that a smaller sample size is acceptable when there are high communalities without cross-loadings, as well as quite a few variables that load strongly on each factor. Thus, the adequacy of the sample size is reliant upon the strength and the stability of the solution, and it can only be ascertained after the analysis is concluded. Beavers et al. (2013) suggest that researchers design for at least 150 cases for initial exploratory analysis because factor analytical techniques are multivariate tools that require larger sample sizes. This is reinforced by Darlington (1997), who asserts that a clear factor structure may be obtained by 100 or more cases. The 461 cases and an STV ratio of more than 20 are thus viewed as sufficient to provide a clear factor structure.

The correlation matrix was inspected, and a sufficient number of coefficients above 0.3 were found, indicating that the data were suitable for factor analysis (Tabachnick & Fidell 2013). Bartlett’s test for sphericity and the Keiser–Meyer–Olkin (KMO) measure of sampling adequacy were generated to further assess the factorability of the data. Bartlett’s test was significant (p = 0.000), and the KMO measure was calculated at 0.907, which suggested that factor analysis could be conducted on the sample data (Pallant 2016). A value closer to 1 suggested that the correlation configurations were relatively dense, indicating that factor analysis should yield reliable factors. Field (2018:1287) suggests that KMO values between 0.7 and 0.8 are perceived to be good, between 0.8 and 0.9 are seen as great, and values above 0.9 are recognised to be superb.

The principal component analysis was conducted on the 20 items to reveal the number of factors that best characterised the underlying correlation between the variables. The initial analysis resulted in four factors with eigenvalues over the Kaiser criterion of 1 (eigenvalues > 1). These four factors explained 55.30% of the variance. The scree plot was unclear and indicated inflexion points that suggested retaining either two or four components. To avoid possible overestimation of components, a parallel analysis was performed, which indicated that only two factors with eigenvalues exceeding the corresponding criterion values for randomly generated data (461 respondents × 20 variable items) should be retained for further investigation (Pallant 2016).

Subsequently, a factor analysis with two factors was performed, and the factors were rotated using the direct Oblimin rotation method with Kaiser normalisation to obtain a clearer picture of the factor loadings. According to Tabachnick and Fidell (2013), if an oblique rotation (e.g. Oblimin) with the preferred number of factors provides correlations that exceed 0.32, there is enough variance in the data to warrant oblique rotation. The resulting correlation matrix indicates many correlations that exceed the 0.32 threshold; thus, the solution remains oblique. The initial two-factor solution explained 44% of the variance, with factor 1 contributing 33.95% and factor 2 contributing 10.05%, respectively. This two-factor solution provided a clean factor structure with no communality values with a value less than 0.3, no significant cross-loadings and no factors with less than three variables. Table 6 shows the factor loadings after rotation.

Although the results do not identify the absolute list of service quality variables which may be considered when conducting online shopping, the research does indicate that the main underlining decision variables are clustered into two latent groups (factors), namely order fulfilment and tangibles. These findings are supportive of the various international research mentioned in the literature review.