Optimizing the Location of Virtual-Shopping-Experience Stores Based on the Minimum Impact on Urban Traffic

Abstract

In order to enhance consumers’ experience of online shopping and to reduce their unnecessary car trips for offline shopping, a new mode, namely, establishing the virtual-shopping-experience store, is proposed in this paper. A bi-level programming model is then built with the aim of optimizing the location of the virtual-shopping-experience stores. The upper-level submodel is utilized to optimize the location of the experience stores, as well as the selection of virtual-reality (VR) devices purchased by the stores, by maximizing the social welfare generated from reducing the car trips for offline shopping after the establishment of the virtual-shopping-experience stores. The lower-level submodel is a binary Logit model, one which calculates the probability of consumers’ choices between online and offline shopping according to the locations of the experience stores output by the upper-level submodel. A genetic algorithm is adopted to solve the model. To validate the accuracy of the model, as well as that of the algorithm, case studies are carried out based on the real data collected in Dalian and Ningbo (two cities in China). The case study result demonstrates that the establishment of virtual-shopping-experience stores would contribute to reducing the frequency of car trips for offline shopping, as well as the distance of car trips for offline shopping and the time spent in car trips for offline shopping.

1. Introduction

In the past decade, the rapid rise of e-commerce has resulted in exponential growth of online shopping worldwide. For example, in the U.S., online shopping retail rose from 5% in 2011 to 15% in early 2020 [1]. In China, the number of online shoppers has increased from 302 million in 2013 to 675 million in 2019, and the transaction volume of e-commerce has increased from RMB 4.57 trillion in 2013 to RNB 34.81 trillion in 2019, with the number of users and the transaction volume increasing by 1.24 times and 6.62 times, respectively. While online purchasing experienced a rapid growth before 2020, the pandemic has accelerated this trend [2]. For example, in March 2020, Amazon consumers in U.S. spent 35% more than the same period in the previous year, in order to buy their essential items; this made Amazon hire 175,000 more workers to respond to the new demand [3]. In China, the transaction volume of e-commerce has grown to up to RMB 43.83 trillion in 2022, in comparison with RMB 34.81 trillion in 2019 and RMB 31.63 trillion in 2018.

Despite the tremendous progress achieved in the past decade, the development of online shopping has also faced some challenges and bottlenecks. One of the most noticeable challenges and bottlenecks lies in the shopping experience provided by online shopping. It should be noted that, with the improvement of consumers’ income and living standards, the shopping experience begins to play an increasingly important role in consumers’ choices of purchase channels (i.e., shopping online or offline). Relatively speaking, the experience of online shopping is worse than that of offline, which is obviously manifested in the fact that the online shoppers cannot try out commodities on site, resulting in their inaccurate perception of the commodities’ value, and thus generating higher purchase risks [4]. Therefore, the physical stores designed for offline shopping have regained their advantages under certain circumstances, and their sales volume has even outnumbered that of online stores. Thus, how to improve the shopping experience of consumers as much as possible while controlling the cost has become one of the key issues the online stores have been faced with so far.

Nowadays various e-commerce companies have been paying more and more attention to the shopping experiences of online shoppers. Among other tactics, utilizing virtual reality (VR) technology and embedding it into an e-commerce service has been considered as one of the feasible means for improving the shopping experience of online shoppers for the e-commerce companies. As a promising technology, VR is able to create a highly immersive and multisensory customer experience, and shopping and retail is considered to be one of the most promising application areas of VR [5]. Many retail giants have been making efforts to improve the shopping experiences of online shoppers through integrating VR technology into their e-commerce services and trying to transform the future of the shopping ecosystem, such as Amazon (VR kiosks), Alibaba (Buy + mobile VR platform), eBay (VR Department Store) and IKEA (VR kitchen showroom) [6].

However, for the VR-based shopping experience, the VR devices are very vital, while they are also generally expensive, with the cost of mainstream VR devices such as GearVR, OculusRIft and HTCVive ranging from RMB 1000 to 10,000. The consumers have to bear the cost of expensive VR devices if they want to have the shopping experience be as real as possible. However, as for most consumers, they are unlikely to spend a lot of money to purchase the expensive VR devices in order to achieve a better shopping experience. In order to solve this conflict, we propose that the virtual-shopping-experience stores could be established neighboring consumers’ residential locations in order to assist consumers in the evaluation of products online and to obtain an exciting and interactive online shopping experience through a high-immersive VR environment created in the virtual-shopping-experience stores.

Under the new mode proposed, consumers would not need to bear the cost of expensive VR devices, but only need to head to the virtual-shopping-experience stores to obtain the virtual-shopping-experience service. Even though the trips to the virtual-shopping-experience stores would still be needed, the characteristics of the trips are totally different from the trips for offline shopping. Since most physical stores are generally located in areas far away from consumers’ residential locations, such as in central business districts, the travel distance of the offline shopping trips induced is generally long. Consumers often head to the physical stores by car, which increases urban traffic and often leads to emission of more air pollutants and exacerbation of traffic congestion. However, under the new mode of establishing virtual-shopping-experience stores, consumers could head to the virtual-shopping-experience stores by walking or by bicycle, since the virtual-shopping-experience stores would be established neighboring consumers’ residential locations. Compared with the long-distance trips for offline shopping, the trips to the virtual-shopping-experience stores would be characterized by a very short distance, environmental friendliness and saved travel time. If consumers’ offline shopping trips could be substituted for with the trips to nearby virtual-shopping-experience stores as much as possible, then this mode could also facilitate a reduction in travel demand for offline shopping and facilitate the alleviation of urban traffic congestion and air pollution. Especially in this era, with the increasingly prominent threat of energy crisis and environmental pollution issues, many countries have proposed the long-term goal of environmental protection (such as the goals of carbon peaking and carbon neutrality proposed by the Chinese government), and there is a broad consensus around the world that a sustainable development mode should be pursued. Therefore, from the perspective of sustainable development, the new mode proposed, and the potential effects of this mode on the environmental welfare related to the substitution for offline shopping trips, have practical significance for responding to the long-term goals of, and the consensus on, environmental protection worldwide.

One key, among others, to the success of this mode is to optimize the locations of virtual-shopping-experience stores. The consumers would only be willing to head to the virtual-shopping-experience stores if the location scheme of the virtual-shopping-experience stores is sound. The unsound location scheme is thus unable to achieve the effect of increasing the proportion of online shopping and reducing the travel demand for offline shopping. Therefore, this paper aims to optimize the locations of the virtual-shopping-experience stores, with the objective of maximizing the effect (or the social welfare) of the substitution for the offline shopping trips with the trips to the virtual-shopping-experience stores.

This paper makes contributions to the following aspects of the question. Firstly, a new feasible mode, (i.e., establishing the virtual-shopping-experience stores) is proposed with the aim of improving consumers’ online shopping experience and decreasing car trips for offline shopping. Secondly, an integrated optimization of the number and location of virtual-shopping-experience stores is conducted, with the objective of maximizing the social welfare of the reduction in offline shopping trips induced by the establishment of virtual-shopping-experience stores, which is different from the traditional location-optimization objective focusing on maximizing the revenue/profit or minimizing the cost. Thirdly, from the micro-perspective of consumers’ choice between online and offline shopping, the change in offline shopping trips under the optimal location scheme of virtual-shopping-experience stores is analyzed to verify the effectiveness of the proposed mode of establishing the virtual-shopping-experience stores.

The remainder of this paper is organized as follows: Section 2 reviews the related work. Section 3 deals with the problem statement and the model formulation. The solution algorithm is introduced in Section 4. Case studies drawn from the practical situation in two cities of China are carried out in Section 5 to investigate the effectiveness of the new mode proposed. The conclusion is presented in Section 6.

2. Literature Review

2.1. Shopping Experience and Virtual Reality (VR) Shopping

In terms of shopping experience, compared with the e-retailers, the physical retail stores could generally provide consumers with a better shopping experience, which has been confirmed by many reports in the literature. For example, Chu and Paglucia pointed out that the physical retail stores still remain the primary points of contact with consumers, even though the role of the physical stores is evolving [7]. Mintel held that offline shopping prevails as the most popular route for consumers to buy body-involved products [8]. Kilcourse and Rosenblum proposed that offline shopping could provide the instant gratification of buying the product and experiencing the service [9]. Kim et al. held that consumers can directly and vividly perceive and experience a product’s “look and feel”, as well as the store’s exterior and interior appearance and physical atmosphere, the customer service provided by store clerks and other physical retail information [10].

By contrast, many scholars hold a negative attitude towards the shopping experience of online shopping. For example, Kumagai and Nagasawa suggested that, in the case of e-retail, it is difficult for a consumer to experience and feel these tangible and intangible store elements at a level equal to physical shopping, even when the shopping task is completed [11]. Babin et al. pointed out that a consumer is likely to more strongly perceive hedonic shopping value via physical retail than in the case of e-retail, while the utilitarian value might not vary, since a shopping task can be achieved regardless of retail type [12]. Blázquez concluded that lack of experiential information and physical interaction with the product is one of the main barriers to buying products online [13]. Merle et al. proposed that, for online shopping, the lack of direct experience may lead to less consumer enjoyment in the shopping process and a greater perception of risk [14].

However, thanks to the innovations in virtual reality (VR) and augmented reality (AR) technologies, highly immersive and multisensory customer experiences could be created and translated to the online environment [15]. The VR (or AR) technology is believed to solve limitations of space and time and to enable the replication and creation of any shopping environment, one that is accessible for consumers at any time [16]. Therefore, the VR (or AR) technology is blurring the boundaries between the in-store and online shopping experiences, assisting consumers to evaluate products online and creating an exciting and interactive online experience.

In parallel, the academic research on the application of VR and AR in the online shopping context is rapidly increasing. For example, Xi and Hamari investigated how VR affects consumer shopping psychology and behavior through a thorough literature review method [6]. Han et al. implemented a VR shopping environment to examine, by applying the SOM model and experience economy, how telepresence and interactivity affected the consumers’ usage behavior after experiencing a VR shopping environment; the findings suggest a role for VR shopping in the digitalization of SCM for sustainable management [17]. Alzayat and Lee proposed that a VR retail environment has a more positive impact on hedonic shopping value than an online retail website [18]. Meissner et al. assessed the effects of immersion on consumer choice by employing an incentive-aligned choice experiment. The experimental results reveal that consumers exposed to a highly immersive VR environment are more likely to choose more products and are likely to be less price-sensitive [19]. Alkarney and Almakki investigated the factors that influence consumer intention to shop online by means of VR in order to analyze consumer adoption behavior of VR apps [20]. Liu et al. proposed an AR- and VR-based try-on system that provides users a novel shopping experience in which they can view garments fitted onto their personalized virtual body and conducted two user studies to compare the different roles of VR- and AR-based try-ons and validate the impact of personalized motions on the virtual try-on experience. Their experimental results show that AR- and VR-based try-ons can positively influence the shopping experience, compared with the traditional e-commerce interface [21]. Pizzi et al. developed a multiple moderated mediation model and examined the potential of a VR store to elicit hedonism and utilitarianism in a quasi-experimental between-subjects design [22]. Riar et al. systematically reviewed a large body of literature on AR shopping and proposed a set of dynamics present in AR shopping as well as some future research agenda points [23]. Barta et al. investigated the impact of AR on the consumer online shopping decision process through a between-subjects experiment [24]. Kim et al. proposed that VR provides more realistic images of products in a 3D naturalistic environment than does a traditional online platform, and that the high-quality, vivid product presentation plays a significant role in enhancing the shopping experience by helping consumers obtain richer product information, as well as directly impacting consumers’ perceived usefulness and perceived enjoyment. Additionally, their findings also revealed that the boosting of perceived usefulness and perceived enjoyment has a significant impact on consumer behavioral responses; thus, VR could be considered as a selling device designed to enrich consumer shopping experiences in an interactive virtual shopping environment [25]. Farah et al. affirmed that the prominence of VR in the world of retail and its impact on the demise of physical stores can no longer be overlooked or placed at the end of the list of priorities [26].

The existing literature focusing on the application of VR and AR in the online shopping context demonstrate that VR and AR technology could improve the shopping experience of the online channel and reduce the perceived shopping risks, thereby increasing the online purchase intention, which lays a solid foundation for the proposal of establishing the virtual-shopping-experience stores described in this study. Additionally, the works in the existing VR- and AR-related literature mainly describe research conducted from the perspectives of theories, experimental methods, devices, simulated environments, etc. To the best of our knowledge, the location issue concerning the VR and AR stores has rarely been investigated, while the location decisions concerning the VR and AR stores are of great importance to the realization of the utility generated by the VR and AR technology. This paper tries to fill this research gap by proposing a location optimization method for the virtual-shopping-experience stores from the perspective of maximizing social welfare.

2.2. Facility Location

The available research content describing facility location is quite extensive. The location decisions, ranging from cities, industrial belts, economic development zones and multinational corporations, to airports, port terminals, water conservancy facilities, human settlement, sales outlets, warehouses, distribution centers, etc., are all described within the scope of facility location research. Research considering location theory was first proposed by Weber [27], then attracted the interest of scholars with the publication of Hakimi’s research [28], and since has developed into an important research topic in operations research and management science, given the improvements in digital computing power. A large amount of related research has accumulated, studies which mainly concentrate on solving the facility location problems of public departments (such as hospitals and post offices) and private organizations (such as factories, retail stores and banks). Although the location problems vary, the elements involved in the location problems are basically the same, namely, the market demands located at points or paths, the facilities, the candidate spaces, and the spatial impedance between demands and facilities (measured by the transportation time or transportation cost) [29]. The research in the early days of the field mainly focused on the formulation of theoretical models. For example, Krarup et al. discussed the simple plant location problem (SPLP) and proved theoretically that the SPLP is an NP-hard problem [30]. Revelle et al. conducted research on practical problems that had not been paid attention to in facility location, and proposed many new models which were different from traditional ones [31]. Klose et al. classified the facility location models as discrete versus continuous location models, location models without capacity constraints versus with capacity constraints, single source versus multi-source location models and static versus dynamic location models [32].

Current research mainly focuses on practical problems. For example, Alzubi and Noche optimized the number and the location of citrus collection points and hubs in order to optimize the citrus supply chain [33]. Wang et al. studied the service facility location problem by taking probabilistic risks of facility disruption, customer choices of last-mile travel path and traffic congestion into consideration [34]. Ge et al. examined the facility location problem for the U.S. fresh produce supply chain by adopting complementary modeling approaches [35]. Hu et al. developed a bi-level multi-objective model to optimize the location and size of general service infrastructure in urban areas [36]. Wolff et al. developed a mathematical model to determine the optimal locations of transport infrastructure, a question which is related to the design of renewable fuel supply chains [37]. Kang et al. examined the facility location problem while considering customer preferences [38].

The existing research on facility location has provided methods and models for reference for our study, but as a newly emerging question, the location problem of virtual-shopping-experience stores has rarely been studied. This paper tries to fill this research gap by optimizing the location of virtual-shopping-experience stores based on the impact of the location scheme of virtual-shopping-experience stores upon urban shopping trips, which is innovative in the context of present research ideas.

2.3. The Impact of Online Shopping on Shopping Trips

Ever since the rapid development of online shopping, the research on the impact of online shopping on the travel demands induced by offline shopping has been a hot topic. However, no consensus has been reached so far on whether online shopping would affect the travel demands induced by offline shopping and what kind of impact it would have. According to the differences in research findings, the viewpoints could be generally divided into four categories: (1) substitution—online shopping would reduce the trips for offline shopping; (2) complementarity—online shopping would increase the trips for offline shopping; (3) modification—the trips for offline shopping would be altered due to online shopping; and (4) neutrality—online shopping would have no effect on the trips for offline shopping. Some scholars hold that online shopping would have a substitution effect on trips for offline shopping. For example, Carling et al. compared the carbon emissions generated by online and offline shopping using the data from Dalecarlia region in Sweden, and found that online shopping would reduce the carbon emissions by 84% compared with offline shopping [39]. Shi et al. explored the relationship between online shopping for four types of commodities (clothes and shoes, electronics, food and drink, and cosmetics) and offline shopping trips, using data from structured interviews with 710 respondents in Chengdu, China. The results suggest that online shopping has a substitution effect on the frequency of offline shopping trips, and that online shopping could be regarded as a possible solution for urban congestion [40]. However, some scholars assert that online shopping would have a complementary effect on offline shopping trips. For example, Mokhtarian pointed out that offline shopping trips are one of the essential travel activities, and that more online shopping would generate more offline shopping trips [41]. Farag et al. investigated the relationship among frequency of online searching, online shopping, and non-daily shopping trips, and found that the complementarity effect between online and offline shopping is more likely than is substitution [42]. Zhou and Wang disentangled the bidirectional connections between online shopping and offline shopping trips. Their results demonstrate that online shopping increases offline shopping trips, while offline shopping trips tend to have a negative impact on online shopping propensity [43]. In addition, some scholars hold that online shopping would have a modification effect on offline shopping trips, which means online shopping would alter some attributes of offline shopping trips, such as travel time, travel distance, and travel frequency. In another example, Ding and Lu investigated how online shopping, offline shopping and other dimensions of travel behavior relate to each other, using a structural equation model and GPS-based travel activity diary data. Their results show that online shopping frequency stimulates the frequency of both offline shopping and online searching, but that online shopping may reduce out-of-home leisure trips [44]. Beckers et al. studied the impact of households’ online consumption on freight trips and found three key factors that determine the magnitude of freight traffic originated by household’s online shopping [45]. Aldo et al. examined how online shopping frequency influences the variables of shopping trip frequency and travel time. Their results demonstrate that consumers who shop online frequently tend to have a higher proportion of shopping trip frequencies in their total trips, but the share of travelled time for shopping tends to be reduced [46]. However, some scholars hold a neutral view: that online shopping has no significant impact on shopping trips. For example, Shi et al. conducted an empirical study on the relationship between online shopping and shopping trips for car owners. The results demonstrate that online shopping is unlikely to reduce shopping trip frequency for owners, as compared to non-owners, and that no significant difference could be directly found in enhancement of shopping trip frequency for owners compared to non-owners [47]. Shah et al. investigated how online shopping, teleworking and travel at a tour level interact between each other and asserted that there is no statistically significant relationship between online shopping, maintenance tours and discretionary tours [48].

The pandemic beginning from 2020 significantly changed people’s online and offline shopping behaviors, and many scholars have noticed this influence and explored the impact of the pandemic on online shopping, and the corresponding effect on shopping trips offline. For example, Ghodsi et al. investigated the effect of factors related to the COVID-19 pandemic and related demographic characteristics on shopping attitude and on shopping trips [49]. On this basis, Ghodsi et al. further studied the pandemic’s influence on online shopping and travel behavior through a structural equation-modelling approach and found that people’s inclinations toward online shopping during the pandemic would influence their travel habits [50]. Drummond and Hasnine investigated the factors that influenced online and in-store shopping behaviors in New York City during the COVID-19 pandemic, and obtained the following results: (1) the increased subway usage was correlated with in-store shopping during the pandemic; (2) higher-income individuals were less likely to shop in-store, whereas lower-income individuals were likely to shop in-store; and (3) an increase in online shopping does not necessarily lead directly to a decrease in in-store shopping [51].

As the pandemic subsides, research questions such as whether or not the influence of the pandemic on online shopping and shopping trips is expected to remain, and if so, to what extent, has gained the interest of researchers. For example, Diaz-Gutierrez et al. investigated this research question using a quasi-longitudinal survey dataset of Puget Sound residents (Washington, USA), and discovered that people’s online and in-store shopping frequency during the pandemic was affected, in both cases, by their perceived health risk, attitudes toward shopping, and pre-pandemic shopping frequencies. In addition, they also discovered that the frequency at which people expect to shop in post-pandemic periods was influenced by their attitudes toward shopping, changes during the pandemic and their pre-pandemic shopping frequency [2]. Adibfar et al. proposed that online and offline shopping after the pandemic would return to their normal trends, as they were before the pandemic. They also mentioned that the statement that online shopping can eliminate offline shopping due to the pandemic is hard to support, because people’s offline shopping desire needs to be fulfilled due to several considerations, including the joy of shopping, interactions with other people, and experiencing the products they want to purchase [3].

The literature discussed above sheds much light on the impact of online shopping on offline shopping trips before and after the pandemic, which lays a solid foundation for our study. However, the existing literature mainly focuses on analyzing the relationship between online shopping and urban traffic from a macro perspective, and few studies evaluate the impact of online shopping on urban traffic based on the perspective of how consumers make choices between online and offline shopping. In order to fill the research gap mentioned above, this paper tries to optimize the location of virtual-shopping-experience stores and explore the impact of online shopping on urban traffic under the optimal location scheme of virtual-shopping-experience stores.

3. Problem Statement and Model Formulation

3.1. Problem Statement

Online shopping has reshaped consumers’ shopping habits and shopping behavior. Correspondingly, the characteristics of urban traffic have changed significantly with the popularization of online shopping, since consumers’ offline shopping trips decrease, but freight trips for deliveries increase in their place. The trips related to different shopping modes are summarized in Figure 1.

Under the first shopping mode (i.e., only the offline shopping channel exists), consumers head to physical stores and take home the commodities purchased (except for some large items that need to be delivered). The traffic related to consumers under this shopping mode includes the passenger trip flows and the freight trip flows for delivery of large items, the latter being rather small in quantity. Under the second shopping mode (i.e., both online and offline shopping channels exist, but none of virtual-shopping-experience stores are established), a large number of consumers would terminate their visits to physical stores or at least reduce the frequency, but increase their online shopping frequencies instead. Therefore, the number of consumers’ shopping trips would decrease, but the number of trucks’ trips for delivery would increase instead. Under the third mode (i.e., both online and offline shopping channels exist and the virtual-shopping-experience stores are established), with the establishment of virtual-shopping-experience stores, the shopping experience of the online channel would improve. As a result, the frequency at which consumers headed to physical stores would further decrease, while the frequency for the visits to the virtual-shopping-experience stores would increase, together with the number of trucks’ trips for delivery. Since the virtual-shopping-experience stores are established neighboring consumers’ residential locations, the trip distance to the virtual-shopping-experience stores for consumers would be relatively short, which is more suitable for walking or riding a bicycle than by car. Therefore, compared with the second shopping mode, under the third mode, the increase in the number of visits to the virtual-shopping-experience stores would not lead to an increase in car traffic, but would instead result in a decrease, due to the further reduction in shopping trips, while the increase of truck traffic would remain very limited due to the economy of quantity for truck deliveries. In summary, the third shopping mode could not only further enhance the competitiveness of online shopping from the perspective of improvement in shopping experience, but is also likely to facilitate the decrease of traffic demand as long as the locations of the virtual-shopping-experience stores are reasonable.

The success of the third mode depends on the accessibility of the virtual-shopping-experience stores and the perceived improvement in shopping experience obtained in the stores by the consumers, which relates both to the establishment and the cost of the stores. The denser the distribution of virtual-shopping-experience stores, the better the accessibility of the stores, while the more expensive the VR devices, the higher the perceived improvement in shopping experience for consumers. That is to say, the higher the investment in establishing virtual-shopping-experience stores is, the more frequently the consumers would purchase commodities through online shopping, and the larger the sales volume of e-commerce would be. Compared with the increase in online sales volume, the profits of e-commerce companies are likely to shrink due to the high cost of establishing virtual-shopping-experience stores. If the location problem concerning virtual-shopping-experience stores is optimized by considering the profit of e-commerce companies as the objective, it is very likely that the traffic demand would not be reduced, and the purpose of alleviating urban traffic congestion and air pollution would not be achieved, which is not in line with the original intention of promoting sustainable development. Therefore, this paper optimizes the spatial distribution of virtual-shopping-experience stores (including the optimization of the number and locations of virtual-shopping-experience stores) by maximizing the social welfare generated from the reduction in offline shopping trips, and analyzes the consumers’ behavior of making a choice between offline and online shopping given the change in the spatial distribution of virtual-shopping-experience stores. Additionally, this paper also explores the changes in offline shopping trips under the optimal location scheme of virtual-shopping-experience stores, and verifies whether online shopping has a substitution or complementarity effect on offline shopping trips.

To facilitate the formulation of a mathematical model, the following assumptions are made, without affecting the economic principles and logical relationships of the problem to be studied:

 Assumption 1. 

The research region is divided into multiple traffic zones. On the premise of reasonable division of traffic zones, it is assumed that only one virtual-shopping-experience store could be established at most in each traffic zone, considering the fact that the investment in setting up a virtual-shopping-experience store is generally very substantial, and repeated construction would lead to resource waste.

 Assumption 2. 

Since one of the original intentions in setting up virtual-shopping-experience stores was to reduce the car travel demand, the virtual-shopping-experience stores are established neighboring consumers’ residential locations, and consumers generally arrive at the virtual-shopping-experience stores by walking or by bicycle. Therefore, it is assumed that the setting-up of the virtual-shopping-experience store only affects the shopping behavior of consumers in the traffic zone where it is set up, and that the shopping behavior of consumers in other traffic zones would not be affected, based on the fact that the virtual-shopping-experience store is out of the walking or cycling distance for consumers in other traffic zones.

 Assumption 3. 

Since the consumption of goods by consumers does not change significantly in a short period of time, the total shopping demand for consumers in the research region is assumed to be constant.

 Assumption 4. 

The virtual-shopping-experience stores set up in different traffic zones are assumed to be totally identical, given the consideration of economies of scale and the requirement of standardization. Specifically, the specification and type of VR devices and the number of VR devices, as well as the store-construction costs, are assumed to be the same among virtual-shopping-experience stores set up in different traffic zones.

 Assumption 5. 

Consistent with Assumption 2, it is assumed that the virtual-shopping-experience stores are established neighboring consumers’ residential locations, and consumers generally head to the virtual-shopping-experience stores by walking or by bicycle, which would not cause traffic congestion or carbon emissions.

 Assumption 6. 

The commodities consumers purchase online are transported to their destinations generally by four transport modes (i.e., road, rail, water and air), so it is assumed that the commodities consumers purchase online flow into the research region from four types of entrances, namely, the highway junctions, the railway stations, the port terminal and the airport.

3.2. Notations

The notations of the sets, the parameters, and the variables related to the model proposed are indicated in

Table 1. Notations of the sets in the bi-level model.

Sets	Notations
M	The sets of traffic zones, o ∈ M, d ∈ M
S	The sets of commodity categories, s ∈ S
Q	The sets of traffic zones where city entrances are located, m ∈ Q

,

Table 2. Notations of the parameters in the bi-level model.

Parameters	Notations
$\alpha$	The coefficient for converting the travel distance into the travel cost.
$P_{od}^{}$	The probability of shopping trips from zone o to zone d.
$P_{od}^{\prime }$	The modal split of cars for trips from zone o to zone d.
$di{s}_{od}$	The travel distance between zone o to zone d.
$N_o$	The population of zone o.
$n_q$	The average number of trips for offline shopping per person per year.
$P_1{}^{\prime }$	The probability of consumers choosing offline shopping under normal circumstances (i.e., when none of virtual-shopping-experience stores are established).
$K_s$	The consumption quantity of commodities for all consumers per year.
$M_s$	The average weight for each unit of commodity s.
$V$	The average load capacity per truck.
$P_m$	The probability that commodities enter the city from the entrance located in zone m.
$P_{mo}$	The probability that commodities are transported to zone o from the entrance located in zone m.
$di{s}_{mo}$	The distance between zone m and zone o.
$t$	The average annual working hours.
$t_{od}$	The travel time by car from zone o to zone d.
$t_{mo}$	The travel time by truck from zone m to zone o.
$C_s$	The profit obtained by the e-commerce companies from the sale of each unit of commodity s.
$n$	The number of VR devices invested in each virtual-shopping-experience store.
$u$	The store-construction cost for each virtual-shopping-experience store.
${\beta }_1$–${\beta }_7$	The parameters to be estimated.

and

Table 3. Notations of the variables in the bi-level model.

Variables	Notations
$X_o$	Decision variable, which equals 1 if the virtual-shopping-experience store is established in zone o, but equals 0 otherwise.
$v$	Decision variable, which denotes the unit cost of VR devices invested in the virtual-shopping-experience store.
$T_o$	The carbon emission cost savings induced by the reduction in the number of offline shopping trips after the virtual-shopping-experience store is established in zone o.
$G_o$	The time cost savings induced by the reduction in offline shopping trips after the virtual-shopping-experience store is established in zone o.
$C$	The cost of establishing a virtual-shopping-experience store.
$D_o$	The number by which offline shopping trips are reduced in zone o.
$y_2$	The perceived online shopping experience-value obtained in a virtual-shopping-experience store.
$P_{ijr}$	The probability of a consumer i in traffic zone j choosing shopping channel r (r = 1 for physical shopping, r = 2 for online shopping) after the virtual-shopping-experience stores are established.
$V_{ijr}$	The utility obtained by consumer i in traffic zone j by choosing shopping channel r (r = 1 for physical shopping, r = 2 for online shopping) after the virtual-shopping-experience stores are established.
$y_{ijr}$	The perceived shopping experience obtained by consumer i in traffic zone j by choosing shopping channel r (r = 1 for physical shopping, r = 2 for online shopping) after the virtual-shopping-experience stores are established.
$z_{ij}^{\left(1\right)}$	The accessibility to supermarkets and shopping malls for consumer i in traffic zone j.
$z_{ij}^{\left(2\right)}$	The accessibility to bus stations for consumer i in traffic zone j.
$w_i^{\left(1\right)}$	The gender of consumer i ($w_i^{\left(1\right)}=1$ for female, $w_i^{\left(1\right)}=0$ for male).
$w_i^{\left(2\right)}$	The monthly income of consumer i ($w_i^{\left(2\right)}=0$ for monthly income of 1000 or less, $w_i^{\left(2\right)}=1$ for monthly income of 1001–4000, $w_i^{\left(2\right)}=2$ for monthly income of 4001–8000, $w_i^{\left(2\right)}=3$ for monthly income of more than 8000).
$w_i^{\left(3\right)}$	The car ownership of consumer i ($w_i^{\left(3\right)}=1$ for owning cars, $w_i^{\left(1\right)}=0$ for having no cars).
$w_i^{\left(4\right)}$	The internet experience of consumer i ($w_i^{\left(4\right)}=0$ indicating using the internet for no more than 5 years, $w_i^{\left(4\right)}=1$ indicating using the internet for 6–9 years, $w_i^{\left(4\right)}=2$ indicating using the internet for more than 9 years).
$P_r$	The average probability of consumers choosing shopping channel r (r = 1 for physical shopping, r = 2 for online shopping) after the virtual-shopping-experience stores are established.

.

3.3. Model Formulation

(1) Upper-level submodel

The upper-level model is an optimization model for the location of the virtual-shopping-experience stores. The model formulation is as follows.

$$\mathit{max}\hspace{1em}F={\displaystyle \sum _{o\in M}{X}_o\left(T_o+G_o-C\right)}$$

(1)

$$S.T.\begin{array}{cc}& T_0=\alpha \left({\displaystyle \sum _{d\in M}{D}_o{P}_{od}{P}_{od}^{\prime }di{s}_{od}}-{\displaystyle \sum _{s\in S}\frac{K_s\left(P_1^{\prime }-P_1\right)M_s}{V}}{\displaystyle \sum _{m\in Q}{P}_m{P}_{mo}di{s}_{mo}}\right)\end{array}$$

(2)

$$G_0=\frac{GDP}{t}\left({\displaystyle \sum _{d\in M}{D}_o{P}_{od}{t}_{od}}-{\displaystyle \sum _{s\in S}\frac{K_s\left(P_1^{\prime }-P_1\right)M_s}{V}{\displaystyle \sum _{m\in Q}{P}_m{P}_{mo}{t}_{mo}}}\right)$$

(3)

$$D_o=N_o{n}_q\left(1-\frac{P_1}{P_1^{\prime }}\right)$$

(4)

$$C=u+nv$$

(5)

$${\displaystyle \sum _{s\in S}{K}_s\left(P_1^{\prime }-P_1\right)C_s}-{\displaystyle \sum _{o\in M}{X}_{o}C}\ge 0$$

(6)

$$0\le y_2=f\left(v\right)\le 1$$

(7)

$$X_o=\left\{0,1\right\}\begin{array}{cc}& \forall o\in M\end{array}$$

(8)

The objective function of the upper-level submodel (denoted by Equation (1)) represents the maximization of the social welfare generated by the reduction in the number of offline shopping trips after the virtual-shopping-experience stores are established. The social welfare generated by the reduction in offline shopping trips is measured by the sum of the carbon emission cost saved and the time cost saved, minus the cost for establishing the virtual-shopping-experience stores. Constraint 1 (denoted by Equation (2)) represents the carbon emission cost saved by the reduction in offline shopping trips after the virtual-shopping-experience store is established in zone o, which is measured by the difference between the carbon emission cost saved due to the reduced offline shopping trips and the carbon emission cost added due to the increase in truck freight trips. Constraint 2 (denoted by Equation (3)) represents the time cost saved by the reduction in offline shopping trips after the virtual-shopping-experience store is established in zone o, which is measured by the difference between the time cost saved due to the reduced offline shopping trips and the time cost added due to the increase in truck freight trips. Constraint 3 (denoted by Equation (4)) represents the number by which offline shopping trips are reduced in zone o after the virtual-shopping-experience store is established. Constraint 4 (denoted in Equation (5)) represents the cost of establishing a virtual-shopping-experience store, which consists of the store-construction cost and the investment of VR devices. Constraint 5 (devoted in Equation (6)) indicates that the profit earned by the e-commerce companies should be greater than the cost of establishing the experience stores in order to protect the interests of the e-commerce companies. Constraint 6 (denoted in Equation (7)) represents the perceived experience value obtained in the virtual-shopping-experience stores, which is a function of the investment in the VR devices, with the value ranging between zero and one. Constraint 7 (denoted in Equation (8)) is a variable constraint.

(2) Lower-level submodel

The lower-level submodel is a binary Logit model that is used to determine the individual consumer’s choice between online and offline shopping after the upper-level submodel outputs the optimization scheme (i.e., the location of virtual-shopping-experience stores, and the unit cost of VR devices installed in virtual-shopping-experience stores). The model formulation is as follows.

$$P_{ijr}=\mathit{exp}\left(V_{ijr}\right)/{\displaystyle \sum _R\mathit{exp}\left(V_{ijr}\right)}$$

(9)

$$V_{ijr}={\beta }_1{y}_{ijr}+{\beta }_2{z}_{ij}^{\left(1\right)}+{\beta }_3{z}_{ij}^{\left(2\right)}+{\beta }_4{w}_i^{\left(1\right)}+{\beta }_5{w}_i^{\left(2\right)}+{\beta }_6{w}_i^{\left(3\right)}+{\beta }_7{w}_i^{\left(4\right)}$$

(10)

$$P_r=\frac{1}{{\displaystyle \sum _j{N}_j}}{\displaystyle \sum _j{\displaystyle \sum _i{P}_{ijr}}}$$

(11)

Constraint 1 (denoted by Equation (9)) represents the probability that an individual consumer in each traffic zone would choose online or offline shopping after the virtual-shopping-experience stores are established. Constraint 2 (denoted by Equation (10)) represents the utility function of online and offline shopping, which is measured by the shopping experience, the attributes of traffic zones and the demographic attributes. Based on the study from Shi et al. [47], the attributes of traffic zones consider two elements, i.e., the accessibility to supermarkets and shopping malls, and the accessibility to bus stations, which are both similarly measured, one by the number of supermarkets and shopping malls within a buffer distance of 800 m, and the other by the number of bus stations within a buffer distance of 800 m. In terms of the demographic attributes, altogether, four elements are considered, i.e., gender, monthly income, car ownership and internet experience (measured by the years of using the internet). Constraint 3 (denoted by Equation (11)) represents the average probability that consumers would choose a certain shopping channel after the virtual-shopping-experience stores are established.

(3) Interaction between the upper-level and the lower-level submodels

The interaction mechanism between the two submodels is described in Figure 2. The upper-level submodel first calculates and outputs the initial value of the two decision variables (i.e., the location scheme of virtual-shopping-experience stores, and the unit cost of VR devices installed in virtual-shopping-experience stores) to the lower-level submodel. Then, based on the output from the upper-level submodel, the lower-level submodel updates the perceived shopping experience of the online shopping channel for consumers in each traffic zone. Then, on this basis, the utility function value and the probability that consumers in each traffic zone would choose the offline shopping channel and the online shopping channel, respectively, is obtained. Then the lower-level submodel feeds back the result of the consumers’ choice between the two shopping channels to the upper-level submodel. Based on the feedback from the lower-level submodel, the upper-level submodel then recalculates the number by which offline shopping trips are reduced and the increase in truck freight trips in each traffic zone. On this basis, the two items relevant to social welfare, i.e., the time cost saved and the carbon emission cost saved, can be determined, and then the total social welfare could also be ascertained by considering the cost added for establishing virtual-shopping-experience stores. After the total social welfare is updated under the given plan, the upper-level submodel would be operated again and a new optimization scheme would feedback to the lower-level submodel. The whole process repeats until an equilibrium is reached.

4. Solution Algorithm

To solve the above-mentioned bi-level model, a specific algorithm based on the genetic algorithm is designed. The detailed steps are as follows.

Step 0: Set $\rho =1$ ($\rho$ denotes the number of iterations), $F^{*}=0$ ($F^{*}$ denotes the objective value of the upper-level submodel for the best found solution), and $X^{*}=\varnothing$ ($X^{*}$ denotes the best solution).

Step1: Generate an initial population $X^{\left(\rho \right)}=\left\{X_1^{\left(\rho \right)},X_2^{\left(\rho \right)},\cdots ,X_{LL}^{\left(\rho \right)}\right\}$ ($LL$ denotes the number of chromosomes in the population). For each chromosome, the coding method is shown in Figure 3. The sub-steps for generating a chromosome are as follows.

Step (1-1) Generate L random numbers with the value of 0 or 1, where L denotes the number of traffic zones;

Step (1-2) Generate a random number v and place it at the end of the chromosome, where v denotes the unit cost of the VR devices investment acquires for the virtual-shopping-experience store (the value range is determined by Equation (7));

Step (1-3) According to the value in each gene, calculate the probability that consumers would choose physical shopping after the virtual-shopping-experience stores are established (i.e., the variable $P_1$) through Equations (9)–(11);

Step (1-4) Determine the validity of the chromosome. Specifically, determine whether Equation (6) holds, i.e., determine whether the cost of establishing the virtual-shopping-experience stores has exceeded the profits of the e-commerce companies. If Equation (6) holds, then go to Step 2. If Equation (6) does not hold, i.e., the cost exceeds the profit, then calculate the maximum number of virtual-shopping-experience stores that makes Equation (6) hold, and change the gene value of “one” to “zero” in reverse order until the number of the potential substitutions with a remaining value of “one” reaches the maximum number of virtual-shopping-experience stores, and then go to Step (1-3).

For example, the chromosome is set as [1 0 0 1 0 1 1 0 1 1 0 1 0 0 3.2], which represents that the virtual-shopping-experience stores are established in seven traffic zones (i.e., zone 1, zone 4, zone 6, zone 7, zone 9, zone 10 and zone 12), and the VR devices are purchased at a price of RMB 3200. If the maximum number of virtual-shopping-experience stores allowed to be established is calculated to be five according to Equation (6), then the first five zones are retained, and the last two zones are eliminated (i.e., zone 10 and zone 12 are removed, with the value of “one”, changing into “zero” at the corresponding position of the chromosome).

Step 2: Calculate the fitness value of each chromosome in the population generated. According to the probability calculated in Step (1-3), calculate the objective of the upper-level submodel, which is regarded as being the fitness value of the chromosome.

Step 3: Compare the fitness values among the chromosomes in the population, and record the maximum fitness value $F^{\left(\rho \right)*}$, and the corresponding chromosome $X^{\left(\rho \right)*}$.

Step 4: Compare $F^{\left(\rho \right)*}$ and $F^{*}$. If $F^{\left(\rho \right)*}>F^{*}$, then set $F^{*}=F^{\left(\rho \right)*}$, $X^{*}=X^{\left(\rho \right)*}$; otherwise, keep $F^{*}$ and $X^{*}$ unchanged; output $F^{*}$ and $X^{*}$.

Step 5: Set $\rho =\rho +1$; determine whether the maximum iteration number ${\rho }_{\mathit{max}}$ is reached. If yes, terminate the computation; otherwise, go to Step 6.

Step 6: Perform the operations of selection, crossover and mutation to obtain a new population. To avoid getting stuck at a locally optimal value, the roulette wheel selection is adopted, and the single-point crossover and the single-point mutation are used. Due to the constraint of Equation (6), the operations in Step (1-3) and Step (1-4) need to be performed to verify that, after crossover and mutation, the offspring chromosomes meet the constraint. Return to Step 2.

5. Case Study

5.1. Case Inputs and Basic Data

(1) Case-study regions and parameters input

In order to verify the effectiveness of the methodology proposed in Section 3 and Section 4, two cities (specifically, the urban areas) in China are selected as the case-study regions, with the added consideration of data availability. The profiles of the two case-study cities are summarized in

Table 4. Profiles of the two case-study cities.

No.	Case-Study City	Area (km2)	Population (Million)	Retail Sales of Consumer Goods (RMB, in Billions)	Size Category
1	Urban area of Dalian City, China	624	2.27	110.63	Medium
2	Urban area of Ningbo City, China	3730	3.11	296.11	Large

. The case-study regions of the two cities are divided into several grids, with a size of 2 km × 2 km for each grid. The various grids as divided are regarded as traffic zones, shown in Figure 4. The road networks of the two cities are shown in Figure 5. Based on the road network, the shortest distance between each traffic zone, i.e., $di{s}_{od}$ and $di{s}_{mo}$, can be calculated by applying the Dijkstra shortest-path algorithm.

The origin–destination (OD) matrixes among each traffic zone for the two case-study cities are estimated based on the residents’ completion of a travel survey conducted in Dalian in 2013 and in Ningbo in 2016, and the probability matrixes for offline shopping (i.e., $P_{od}^{}$) among each traffic zone are thus obtained for the two case-study cities. Additionally, based on the shortest distance between any two traffic zones in the two regions, the travel cost and the travel time of cars for offline shopping can be calculated, and then the modal split of cars for trips between any two zones (i.e., $P_{od}^{\prime }$) can be calculated by using the Logit model for the two cases. The respective matrixes for the probability that commodities are transported to each traffic zone from entrance zone (i.e., $P_{mo}^{}$) for the two cases are calculated according to the frequency of online shopping of residents in each zone.

In addition, according to the Dalian Statistical Yearbook (2022) [52] and the Ningbo Statistical Yearbook (2022) [53], the average annual expenditure on various commodities for consumers in the two case-study cities can be estimated, and then the average annual consumption of various commodities (i.e., $K_s$) for the two cases can be calculated by dividing the annual expenditure by the unit price. The average number of visits to physical stores per capita per year is determined as $n_q$ = 50.4 through survey data. Assuming that the e-commerce companies charge 5% commission for each commodity sold online, the average prices of common types of commodities, the average profit earned from sales of each unit of commodities by the e-commerce companies (i.e., $C_s$) and the average weight of each unit of commodities (i.e., $M_s$) are shown in

Table 5. Average price, profit and weight of each unit of commodities.

Categories of Commodities	Average Price (RMB)	Average Profit (RMB)	Average Weight (kg)
Clothes, shoes, hats and accessories	200	10	1
Electronic products and accessories	2000	100	2
Daily necessities, cosmetics, skin-care products	200	10	1
Food and fruit	50	2.5	5
Books and magazines	30	1.5	2
Other	100	5	1

.

Considering empirical reality, the average load capacity of trucks is assumed to be one ton, and the coefficient $\alpha =0.02$, each virtual-shopping-experience store is assumed to have purchased 30 sets of VR devices, and the store-construction cost is assumed to be RMB 120 thousand per year. The proportions in which commodities enter the city through highway, railway, shipping and air (i.e., $P_m^{}$) is calculated by the modal splits of the four transportation modes, based on the Dalian Statistical Yearbook (2022) [52] and the Ningbo Statistical Yearbook (2022) [53].

(2) Functional relationship between the perceived shopping experience and the price of VR devices

The VR devices can be classified as output devices or input devices. In terms of the output devices, four types of vision-based VR systems are commonly used, i.e., monitor, “powerwall”/projector, CAVE and head-mounted display (HMD). HMD has become the most prominent and visible piece of technology pertaining to VR; commonly used HMD devices include HTC Vive, Smartphone-based, and Oculus Rift, etc. [6]. The output VR devices can provide other sensory information, including sound, touch and smell, in addition to the visual output. In order for consumers to effectively conduct shopping in virtual reality, the input devices are also needed, so that consumers can interact with different objects in the VR shopping environment. The commonly used input devices include hand-movement-based devices (e.g., VR controller, mouse and keyboard, gamepad, etc.), eye/head-movement-based devices, general-body-movement-based devices, leg-movement-based devices, and voice input devices [6].

For the different output and input devices, the shopping experience varies with the devices used. It should be noted that it has not been categorically established as to how the shopping experience of VR devices is to be measured, and the implicit idea in the extant literature is that the shopping experience of VR devices needs to more or less cover the field-of-view (FOV), the visual resolution ratio, the comfort levels of wearing devices, the compatibility among devices and the richness of the scenarios simulated. Apart from the shopping experience, some other factors also need to be considered when selecting the VR devices. One of the factors is the durability of the devices; the devices that can withstand frequent and high-intensity use should be chosen as much as possible. In addition, it is necessary to consider the maintenance cost of the devices, and the virtual-shopping-experience stores should install VR devices that are easy to maintain and repair. Furthermore, another important factor that should be considered is the price; the devices with high cost-effectiveness should be chosen as much as possible in order to control the investment and improve the returns.

In order to measure the quantitative relationship between the price of VR devices and the perceived user experience provided, we referred to the evaluation of fifteen sets of VR devices from Zhongguancun (a well-reputed website for evaluation of computer-related products in China). We invited the users in the comment sections associated with the fifteen sets of VR devices to rate the devices they had used based on their user experience. The user experience was evaluated from five dimensions, i.e., the field-of-view (FOV), the visual resolution ratio, the comfort levels of wearing devices, the compatibility among devices and the richness of scenarios simulated. A Likert scale of 10 levels was used for the rating of each dimension. The ratings of the five dimensions from these users were finally averaged and converted into a number ranging from zero to one. The fitted result is shown in Figure 6, which demonstrates a relatively high degree of fitting.

(3) Utility functions of offline shopping and virtual experience online shopping

To clarify the preference of consumers for offline shopping and virtual experience online shopping, we conducted an online stated preference (SP) survey on consumers’ behavior of choice between these two shopping modes. The consumers were prompted to make choices under situations of different combinations of shopping experience, accessibility to supermarkets and shopping malls, and accessibility to bus stations. A total of 180 questionnaires were collected, and, in the end, 155 valid questionnaires were finally obtained, after excluding those with short answer times, high consistency of selection and lack of reported online shopping experience. Based on the questionnaire survey data, the parameters in the model for choice-making between online and offline shopping is calibrated as shown in

Table 6. Parameter calibration results for the choice model of shopping channel.

Variable		Parameter	Value	Significance Level
Shopping channel attribute	Shopping experience ($y_{ijr}$)	${\beta }_1$	6.28	0.0022
Traffic-zone attribute	Accessibility to supermarkets and shopping malls ($z_{ij}^{\left(1\right)}$)	${\beta }_2$	−0.16	0.0076
	Accessibility to bus stations ($z_{ij}^{\left(2\right)}$)	${\beta }_3$	−0.56	0.0024
Demographic attributes	Gender ($w_i^{\left(1\right)}$)	${\beta }_4$	0.89	0.0082
	Monthly income ($w_i^{\left(2\right)}$)	${\beta }_5$	−2.58	0.0086
	Car ownership ($w_i^{\left(3\right)}$)	${\beta }_6$	−1.49	0.0072
	Internet experience ($w_i^{\left(4\right)}$)	${\beta }_7$	1.63	0.0084

.

5.2. Optimization Results

The population size of the genetic algorithm was set to be 50, and the maximum number of iterations was set to be 200, and the crossover probability and the mutation probability were set to be 0.6 and 0.05, respectively.

The optimized locations of virtual-shopping-experience stores for the two case-study cities are shown in Figure 7 and Figure 8.

For the Dalian case, the optimization result reveals that 14 virtual-shopping-experience stores should be established, with VR devices purchased at around RMB 3620 each. Therefore, for each virtual-shopping-experience store, the annual investment approximates RMB 228.6 thousand, and a total of RMB 3.2 million is needed per year for the 14 stores established. From the perspective of the distribution characteristics of store locations, the optimized results demonstrate that the virtual-shopping-experience stores are distributed relatively less in the northern, western and central parts of Dalian City. This could be explained by the fact that the population density in the northern and western part is relatively small, and thus the demand for offline shopping is also small. However, as for the central part, even though the population density in this region is large, large shopping malls and physical stores are densely distributed in this region, and thus the travel distance for offline shopping in this region is relatively short, and the establishment of virtual-shopping-experience stores in this region would have little impact on the distance reduction gained by offline shopping trips.

For the Ningbo case, the optimization result shows that 22 virtual-shopping-experience stores should be established, with VR devices purchased at around RMB 5700 each. Therefore, a total of RMB 6.4 million is needed per year for the 22 stores established. The reason why the price of VR devices purchased in the Ningbo case is more expensive than that in the Dalian case could be attributed to the higher consumption capacity of consumers in the Ningbo case. In terms of the distribution characteristics of store locations, the findings reveal that the virtual-shopping-experience stores are distributed relatively densely in the northeastern and central parts of Ningbo City. This could be explained by the high population density in these parts of the region.

5.3. Discussion

(1) Impacts of this new mode on offline shopping trips

The impacts of the establishment of virtual-shopping-experience stores on offline shopping trips for the two case-study cities are shown in

Table 7. Comparison of offline shopping trips before and after establishing virtual-shopping-experience stores in the Dalian case.

	Without Virtual-Ahopping-Experience Stores	With Virtual-Shopping-Experience Stores	Reduction in Percentage
Investment (RMB per year)		3.2 million
Offline shopping frequency (times per year)	84 million	51.8 million	38.3%
Distance for offline shopping trips (kilometers per year)	407 million	367 million	9.8%
Travel time for offline shopping (hours per year)	16.96 million	13.85 million	18.3%

and

Table 8. Comparison of offline shopping trips before and after establishing virtual-shopping-experience stores in the Ningbo case.

	Without Virtual-Shopping-Experience Stores	With Virtual-Shopping-Experience Stores	Reduction in Percentage
Investment (RMB per year)		6.4 million
Offline shopping frequency (times per year)	125 million	72.8 million	41.8%
Distance for offline shopping trips (kilometers per year)	568 million	519 million	8.6%
Travel time for offline shopping (hours per year)	28.23 million	23.22 million	17.7%

.

The results in Table 7 and Table 8 reveal that the establishment of virtual-shopping-experience stores has positive impact on the reduction of offline shopping trips. Specifically, in the Dalian case, after the establishment of virtual-shopping-experience stores, the frequency of offline shopping trips for Dalian residents would decrease by 38.3%, the total distance of offline shopping trips would decrease by 9.8%, and the total time spent in offline shopping trips would decrease by 18.3%. In the Ningbo case, after the establishment of virtual-shopping-experience stores, the frequency of offline shopping trips for Ningbo residents would decrease by 41.8%, the total distance of offline shopping trips would decrease by 8.6%, and the total time spent in offline shopping trips would decrease by 17.7%.

If we further analyze the percentages of the reductions in the offline shopping trips, and assume that one car drives an average distance of 15 thousand kilometers per year, and that each resident spent an average time of 0.5 h in an offline shopping trip, then in the Dalian case, the reduced distance of offline shopping trips would be equivalent to the annual mileage of about 2670 cars, and the reduction in time spent in offline shopping trips would be roughly equivalent to the total time spent in offline shopping trips of 0.12 million residents per year. In the Ningbo case, the reduced distance of offline shopping trips would be equivalent to the annual mileage of about 3270 cars, and the reduction of time spent in offline shopping trips would be roughly equivalent to the total time spent in offline shopping trips of 0.20 million residents per year.

In summary, the results demonstrate that the establishment of virtual-shopping-experience stores could facilitate reductions in the offline shopping frequency and the distance of offline shopping trips, as well as the time spent in offline shopping trips. Additionally, it could also be deduced that the establishment of virtual-shopping-experience stores could facilitate an increase in the proportion of online shopping due to the improvement in the online shopping experience. Furthermore, due to the reductions in the offline shopping trips, the establishment of virtual-shopping-experience stores could also facilitate a reduction in exclusive car use for offline shopping, and thus lead to lower emission of air pollutants and produce more environmental welfare.

(2) Drawbacks of this new mode

Although establishing virtual-shopping-experience stores could contribute to the improvement of the shopping experience in an online context and increase the probability of online shopping, this mode still has some inevitable drawbacks.

Firstly, this mode would further strengthen the shortcomings of online shopping. For example, since physical shopping is a common way of social interaction, the decrease in physical shopping would lead to reduced social interaction. Additionally, this mode cannot provide consumers with the instant gratification of buying products, and consumers need to be able to afford shipping costs and delivery time in order to acquire the purchased products. Furthermore, it cannot be ignored that the shipping and packaging of online-shopping products would have a negative impact on the environment.

Secondly, this mode would impose an increased reliance on VR and AR technology. One of the prerequisites for the success of this mode is providing consumers with a good shopping experience in the online context through virtual-shopping-experience stores, which necessitates high requirements for the application of VR and AR technology. Additionally, most VR devices currently suffer from issues such as high screen-refresh rate and causing dizziness in users, resulting in a lack of leapfrog improvements in user experience, suggesting a need for technical breakthroughs to address these constraints.

Thirdly, this mode is prone to creating a “Digital Divide”, a significant ethical consideration. This mode would inevitably lead to negative impact on the development of brick-and-mortar retail stores. If a large number of brick-and-mortar retail stores are affected in ways such as being forced to close or downsize, then it would bring great inconvenience to the lives of those who do not know how to or are not willing to shop online. Additionally, as to this population, they cannot enjoy the benefits brought by the new technology and the new shopping mode. We need to consider the need of that group of the population and the series of consequences brought about by the digital divide.

6. Conclusions

This paper mainly studies the impact of establishing virtual-shopping-experience stores on consumers’ choice between online and offline shopping and explores the ways to optimize the location of virtual-shopping-experience stores. Distinct from traditional facility location research, which often considers the maximization of profit/revenue or the minimization of cost as the optimization objective, this paper maximizes the social welfare generated by the reduction in offline shopping trips, which works to relieve urban traffic congestion and reduce automobile exhaust pollution. To conduct the research, a bi-level programming model was established, in which the lower-level submodel was used to determine the change in consumers’ shopping behavior after the virtual-shopping-experience stores were established according to the location scheme output by the upper-level submodel, and the upper-level submodel was used to optimize the location of virtual-shopping-experience stores according to the feedback of consumers’ shopping behavior received from the lower-level submodel. Case studies were carried out to verify the effectiveness of the proposal to establish virtual-shopping-experience stores, as well as the model proposed. The case-study results show that the establishment of virtual-shopping-experience stores is able to significantly reduce the frequency of offline shopping trips, as well as the distance of the offline shopping trips and the time spent in offline shopping trips.

The implementation of this mode would still face some challenges. For example, given the impact of the pandemic, the economic recovery requires more offline consumption and travel to promote the development of the transportation and tourism industries, which is contrary to the original intentions of this mode. Additionally, as a new shopping mode, the proposal needs to consider the consumers’ acceptance, and the market cultivation requires time and a significant investment of money. Furthermore, since this mode has an increased reliance on VR and AR technology, the technical constraints and infrastructure needs should also be considered.

The limitations of the study mainly lie in the following aspects. Firstly, the model proposed is based on some assumptions which simplify the problem to some extent, though the model is still sound. For example, it is assumed that the virtual-shopping-experience stores set up would all be identical due to the considerations of standardization and economies of scales. Even though this assumption is reasonable, if it could be relaxed to allow for the differentiated size of stores; then, more interesting results may be derived. Secondly, it should be noted that the data in the case studies were collected before the pandemic. Although some research, such as [2,3], has suggested that customers’ shopping behavior would return to normal trends, as it was before the pandemic, observation of customers’ shopping behavior after the pandemic is still required to analyze the effect of online and physical retail in the future. Thirdly, the two case studies were both carried out in China, and thus, the results may only reflect China’s situation. The implementation of case studies in other countries is recommended.

Corresponding to these limitations, the directions for future study include the following aspects. Firstly, some assumptions could be relaxed in order to make the research problem more in-line with the real situation. Secondly, investigation of customers’ online and physical shopping behavior after the pandemic could be conducted to determine the various factors influencing it and how they might influence customers’ shopping behavior after the pandemic. Thirdly, case studies in other countries could be offered to reflect the effect of the mode proposed by this study in other countries.