Spatial-temporal distribution characteristics and influencing factors of new retail stores: a case study of Freshippo stores in Shanghai

Abstract

In the era of new retail, the integration of data and technology facilitates cross-regional connections in the food retail industry, enabling precise spatial distribution of stores and active interactions with consumer markets. Some studies view new retail as an extension of omnichannel retail, noting that it integrates the advantages of both online and offline channels and reshapes the urban commercial landscape. This study focuses on Freshippo stores from 2016 to 2024, revealing the spatiotemporal changes and regional differences in new retail stores within urban settings against the backdrop of retail model transformation. Through an analysis based on multi-source data and the integration of qualitative and quantitative methods, this research systematically explores the layout preferences of new retail physical stores in terms of costs, regional environments, and service industry levels. The findings indicate that the layout of new retail physical stores exhibits a trend of expansion from city centers to suburbs, eventually covering most urban communities. Multiple clusters of new retail stores have formed in the main urban areas, suburbs, and outer suburbs, with suburbs and outer suburbs gradually becoming high-value core regions. This deviates from the traditional site selection model that relies on city shopping centers or commercial districts, expanding into down-market areas and avoiding homogenized competition. Additionally, the geographic market layout of new retail physical stores within cities is significantly influenced by multiple factors such as rental costs, accessibility, business environment, community density, and living service facilities. These factors manifest as single-factor, dual-factor, and multi-factor influences. In different local areas, the driving mechanisms behind these factors also exhibit considerable variability, closely related to the service positioning of the stores, indicating that new retail store locations demonstrate unique preferences for various regional factors. Our research findings further enrich retail geography and commercial location theory, providing new perspectives for reshaping urban commercial spaces and significant references for site selection decisions regarding fresh food, new retail stores, and sustainable development planning for urban commercial spaces.

Introduction

As online retail continues to dominate, the retail sector is progressively transitioning into a more rationalized and mature phase of development. In this process, the integration of online and offline channels has significantly increased market saturation, thereby elevating operational risks for certain retailers predominantly reliant on physical sales and leading them to adopt e-commerce strategies to expand their market reach (Mende and Noble, 2019; Huebner et al., 2016; Jones and Livingstone, 2018; Li et al., 2022; Li and Long, 2024; Forman et al., 2009; Bell et al., 2014). Concurrently, numerous online retail enterprises are rapidly expanding their physical storefronts to further consolidate their market position (Daniel and Hernandez, 2024; Zhang et al., 2019; Ge et al., 2020; Hendershott et al., 2000). In this context, leveraging digital technologies such as the internet, the Internet of Things (IoT), and Artificial Intelligence, the retail model has achieved deep integration of online and offline channels, forming what is known as “new retail” (Wang and Coe, 2021; Vhatkar et al., 2024; Gao et al., 2025). As the first company to propose the concept of “new retail,” Alibaba defines it as an omnichannel retail model centered on consumer experience, highlighting the reconfiguration of relationships among consumers, products, and retail spaces through digital technology. Scholars further argue that new retail is not a simple combination of traditional retail and e-commerce, but rather an integrated service model combining online, offline, and smart logistics systems, facilitated by the seamless integration of data and logistics between online and offline channels, thereby improving both operational efficiency and customer satisfaction (Yin and Ye, 2024; Zhang et al., 2024; Berry et al., 1988; Gao et al., 2025). Building on this logic, new retail exhibits spatial and organizational features that distinguish it from traditional retail. Online platforms extend consumer access, offline stores serve as sales centers, experience hubs, and online order fulfillment nodes, and smart logistics systems connect virtual and physical spaces to achieve efficient coordination and real-time matching of supply and demand.

Given the innovative strategies adopted by retail enterprises in various market demands and competitive environments, new retail can be divided into three main categories. The first category involves traditional retail enterprises integrating offline resources and expanding online operations to meet online challenges (Mancuso et al., 2023; Zhou et al., 2024; Li et al., 2024). Typical examples include ALDI, Walmart, Carrefour, Suning, and Sam’s Club. The second category includes online retail enterprises expanding offline, using online channel advantages to open physical stores and enhance consumers’ physical experience (Gao et al., 2025; Zhang et al., 2019; Ge et al., 2020), represented by companies like JD.com and Tmall. The third category, native new retail enterprises, have focused on seamlessly integrating online, offline, and smart logistics from the start, enhancing connections between people, products, and places through innovative supply chain management and smart technologies, improving operational efficiency and market responsiveness, with Freshippo as a typical representative (Wang and Coe, 2021; Zhang et al., 2024; Yin and Ye, 2024). It is evident that there are significant differences between new retail stores, e-retail stores, and traditional retail stores.

Scholars and industry experts concur that these new retail models are leading a transformation in the retail industry by offering consumers more convenient and intelligent services, and view new retail as an extension of the omnichannel retail model (Tang and Chen, 2023; Zhang et al., 2024). Specifically, information technology supports the construction of new retail’s “virtual-real integrated” consumption scenarios, including precise site selection for stores, smart order pushing, and optimization of offline consumption scenarios, enhancing interaction with consumers and expanding new consumer scenarios on the existing market foundation (Kamoonpuri and Sengar, 2023; Kumar et al., 2024; Cui et al., 2024). Additionally, smart logistics effectively bridges the production and sales ends, offering robust support for the seamless integration of online and offline channels (Dey et al., 2023; Xu et al., 2024). This “online + offline + smart logistics” retail model generates economic benefits, inspiring other retailers to emulate and driving the urban commercial environment toward a smarter and more efficient transformation (Lima et al., 2024).

For retailers, a precise site selection strategy is crucial for enhancing market competitiveness. This strategy not only maximizes the attraction of targeted consumer traffic but also enables dynamic adjustments to operational strategies based on regional environments and changes in purchasing power, thus ensuring a strong alignment between store locations and target consumer demographics (Liang et al., 2024; Guan et al., 2025). Furthermore, by integrating smart logistics and information technology, retailers can respond more effectively to changes in consumer demand, securing a sustainable competitive advantage in the new retail environment (Zhang et al., 2024). Despite the critical importance of precise site selection for a retailer’s competitiveness, its implementation faces several challenges. Firstly, significant differences in market demands and consumer preferences across regions necessitate flexibility in site selection strategies (Reed, 2024; Forman et al., 2009). Secondly, limited data analysis capabilities can lead to significant deviations in site selection decisions (Li et al., 2024). Thirdly, the confidentiality of corporate operational strategies hinders learning from successful industry practices, thereby increasing decision-making complexity (Uddin et al., 2024). Additionally, intense market competition and ongoing changes in urban planning add uncertainties to site selection strategies (Brooks and Meltzer, 2024; Daniel and Hernandez, 2024). These challenges underscore the need for retailers to comprehensively evaluate multiple factors and adapt their strategies flexibly in response to environmental changes when applying precise site selection strategies to enhance market competitiveness. Therefore, in-depth research into the key factors affecting precise site selection for retail stores not only assists retailers in better addressing these challenges but also provides theoretical guidance and practical insights for the industry, supporting stable, long-term development in a competitive market environment.

The rapid expansion of new retail formats in urban spaces has garnered widespread attention from the academic community and has facilitated in-depth research on the spatial distribution characteristics of physical stores and their influencing factors. Since Jack Ma proposed the concept of “new retail” in China in 2016, this new retail model—based on information technology and logistics, integrating online and offline channels, and using demand data to guide supply—has been widely adopted and refined across various retail sectors (He et al., 2020; Zhang et al., 2024). In terms of research focus, department stores, chain stores, and supermarkets have become major areas of academic research. At the macro level, retail store site selection exhibits a clearly uneven distribution pattern (Bateman Jerram et al., 2020; Zhou et al., 2024); at the meso level, store distribution displays a multi-center clustering pattern (Xu et al., 2023; Wang et al., 2024); and at the micro level, reasonable product layout significantly improves store operational efficiency (Edirisinghe and Munson, 2023). Some scholars have analyzed the impact of factors such as operational models (Kita and Cvirik, 2024; Guan et al., 2025), brand effects (Cohen et al., 2024; Park et al., 2023), supply chains (Ghosh et al., 2023; Sun et al., 2024), and business environments on retail stores (Lima et al., 2024), finding that effective decision-making requires considering not only consumers’ diverse needs for product categories but also shopping convenience and the appeal of surrounding facilities. In addition, some scholars have used big data to analyze consumer preferences, finding that the application of Internet technologies in logistics and sales chains not only influences consumer behavior but also facilitates the transformation and upgrading of traditional retail into new retail formats (Wu et al., 2024; Yang et al., 2024; Nam et al., 2025). Studies on retail store location selection have found that factors such as consumer coverage, product pricing, and market coverage significantly impact competition in urban retail spaces (Tang and Chen, 2023; Ge and Zhu, 2023; Lu and Menezes, 2024). Given the advantages demonstrated by new retail enterprises in precise site selection strategies in urban commercial spaces, some scholars have proposed that this strategy can effectively mitigate the problem of store stagnation resulting from the “hollowing out” phenomenon in urban commercial centers (Li and Long, 2024).

In the era of new retail, precise site selection for physical retail stores, considering driving factors in diverse urban environments, is crucial for the geographical market distribution of retailers, profoundly impacting their market competitiveness and profitability. Moreover, studying representative new retail formats can provide retailers with successful site selection strategies and effective pathways for market entry. Over the past 10 years, China’s reform and opening up have provided significant opportunities for retail industry development, improving market mechanisms, fostering continuous business model innovations, and leading to the emergence of various retail formats, thereby offering consumers diversified shopping experiences (Han et al., 2024). In 2023, China had 1.092 billion internet users, with online retail sales totaling 15.42 trillion yuan, accounting for 27.6% of the total retail sales of consumer goods, indicating a vast growth potential and an enormous consumer base for new retail formats. China aims to establish a “dual circulation” model, encouraging the use of new technologies to enhance industry quality, improve service functions, expand shopping experiences, and build closer connections between retail, markets, and consumers. With the increasing consumer preference for organic food, the application of new retail models in the fresh food supply chain is particularly important. Keywords such as “fresh,” “green,” “safe,” “organic,” and “healthy” reflect consumer demands and modern consumption patterns. In light of these considerations, this study chooses Shanghai as the primary research area and takes Freshippo (China) Co., Ltd. as a typical representative of the new retail format (Feng et al., 2022; Shyu et al., 2023; Huang et al., 2024).

The rapid expansion of new retail is significantly reshaping the urban commercial landscape; however, the spatiotemporal distribution patterns and underlying drivers of retail store spatial distribution remain underexplored. Limited integration of online and offline channels may restrict access to new retail services among diverse urban populations, while over-centralization of retail store locations may result in uneven service distribution and inefficient resource allocation. These challenges pose several key scientific questions: What are the spatiotemporal distribution patterns and evolutionary mechanisms of urban new retail stores? How do socioeconomic factors, urban infrastructure, and community characteristics jointly influence the spatial distribution of new retail stores? How can the spatial distribution of new retail stores be optimized to enhance operational efficiency and ensure equitable service accessibility?

This study reveals the spatiotemporal evolution characteristics of Freshippo stores by quantifying the number of stores. The study first collected data from Shanghai, including spatial data such as Freshippo stores, socioeconomic factors, rent levels, commercial infrastructure, service facilities, and the number of communities. Additionally, vector data such as roads, rivers, and administrative divisions were collected to build the required database. We used ArcGIS software and GIS spatial analysis techniques to reveal the spatial distribution characteristics and relocation trends of Freshippo stores, and compared these characteristics with various urban elements through visual layer overlays. We then employed geographical detectors and a Binary logistic regression model to systematically explore the factors influencing Freshippo store distribution and their driving mechanisms. Finally, based on the research results, we proposed recommendations to optimize site selection for fresh food new retail formats in urban areas. This study aims to enrich empirical research on commercial location theory and retail geography theory, provide references for the spatial distribution of other retail sectors in urban areas, and guide the government in optimizing the distribution of commercial centers.

Materials and methods

Research object and region

In 2016, Alibaba launched Freshippo, a new retail platform driven by data and technological innovation, and simultaneously introduced the cross-sector Freshippo model (He et al., 2020). This model redefines the entire lifecycle of agricultural products, from production to storage, transportation, and sales, by integrating online and offline channels with smart logistics. The goal was to restructure the entire supply chain to achieve a seamless connection between urban and rural supply and demand. In terms of retail, Freshippo integrates experiential dining, fresh food logistics, and multi-functional services, including takeout, restaurant dining, and shopping through supermarkets and convenience stores, through its supermarket model to meet diverse consumer needs (Shyu et al., 2023). Notably, during the COVID-19 pandemic, the Freshippo model significantly reduced unnecessary human contact, effectively lowering the risk of virus transmission. Its “online + offline + smart logistics” model ensured the efficient delivery of fresh food during urban lockdowns. Additionally, in 2023, Freshippo achieved a milestone with an annual gross merchandise value exceeding 59 billion yuan. By 2024, Freshippo continued to adjust its operational strategies, focusing on creating localized consumption scenarios and expanding into 36 major cities across China.

Shanghai is a significant economic, financial, trade, and shipping hub in China, and ranks as one of the country’s largest cities. According to 2023 data, the city boasts a permanent population of 24.8745 million, with an average disposable income of 84,834 yuan and average consumer spending of 52,508 yuan. The urbanization rate is 91.30%, and the total retail sales of consumer goods amounted to 1.85 trillion yuan, placing them among the highest in the nation. Shanghai is at the forefront of advancing new retail formats, with Freshippo stores constituting 56.51% of the city’s market share. For regional comparative studies, based on urban planning schemes and related research (Wu et al., 2022; Zhang et al., 2024), Shanghai’s 16 districts are categorized into urban central districts, urban periphery areas, and outer suburban districts (Fig. 1).

Fig. 1

Spatial distribution of Freshippo stores in Shanghai in 2024 (map review number: GS [2024] 0650).

Research data

Our research team has been continuously tracking Freshippo stores in Shanghai’s new retail environment since 2016, collecting their spatial and attribute data. From 2016 to 2023, on December 31st of each year, we collected store data from the official Freshippo website, www.freshhippo.com, and as of September 20, 2024, we have gathered data in nine batches. Using Python and the Amap API, we extracted spatial coordinates and points of interest (POIs) of the stores. Additionally, housing and rental information was sourced from real estate websites like Anjuke and Fang.com. All collected spatial attribute data underwent rigorous name and location verification. Vector data for Shanghai’s roads, administrative boundaries, and rivers were obtained from the National Geographic Information Public Service Platform (https://www.tianditu.gov.cn/), remote sensing imagery was provided by the Luojia One website (http://59.175.109.173:8888/app/login.html), and other economic and social data came from the Shanghai Municipal People’s Government and Statistics Bureau.

Research methods

This study employed a combination of field research, spatial analysis, and both quantitative and qualitative analyses. Field research involved semi-structured questionnaire interviews to analyze market planning, operational models, and product layout. Spatial analysis, employing kernel density estimation (KDE) and standard deviational ellipse techniques, precisely measured the clustering of Freshippo stores and visualized changes in their spatial patterns. By integrating quantitative and qualitative analyses, including a literature review, the application of geographical detectors, and the use of Binary logistic regression models, this research systematically identified the key factors and their mechanisms affecting the spatial distribution of new retail physical stores.

Kernel density estimation

KDE is the representative of modern nonparametric statistical methods, which is often used to detect the event density (also known as intensity) that can be measured at any regional site (Zhang et al., 2024; Zheng et al., 2024). KDE is used to analyze the spatial distribution form and density characteristics of spatial data such as Freshippo stores, residential quarters, and commercial facilities in Shanghai. The specific calculation formula is:

$$f(x)=\frac{1}{n{h}_{n}}\mathop{\sum }\limits_{i=1}^{n}k\left(\frac{x-{x}_{i}}{{h}_{n}}\right)$$

(1)

Where n equals the total number of POI data of a certain type; h_n is the bandwidth named search radius. Where the higher the kernel density value is, the higher the probability of event occurrence and the denser the elements are.

Nearest neighbor index

Nearest neighbor index is suitable for the analysis of many point data with irregular distribution. It measures the actual point distribution by the random distribution. We use this method to analyze the form and change characteristics of the spatial agglomeration of Freshippo Stores in Shanghai from 2016 to 2024. The calculation formula is as follows:

$$R=\frac{{\bar{r}}_{1}}{{\bar{r}}_{{\rm{e}}}},\,{\bar{r}}_{{\rm{e}}}=\frac{1}{2\sqrt{n/A}}=\frac{1}{2\sqrt{D}}$$

(2)

Where R is the nearest index; ${\bar{r}}_{1}$ is the actual average distance of the professional Freshippo stores; ${\bar{r}}_{{\rm{e}}}$ is the theoretical average distance of Freshippo stores under random distribution; A represents the study area, n is the number of Freshippo stores, and D is the density of Freshippo stores. When R < 1, Freshippo stores are clustered; When R > 1, Freshippo stores are evenly distributed; And when R = 1, Freshippo stores are randomly distributed.

Standard deviational ellipse

The standard deviation ellipse (SDE) method is a classic technique for analyzing the directional characteristics of spatial distribution. The size of the ellipse reflects the concentration of overall spatial pattern elements, while the azimuth (major axis) indicates the dominant direction of the pattern (Zhang et al., 2023; Liu et al., 2024). This paper uses this method to measure the central point, long and short axis, azimuth, and other distribution indicators of Freshippo stores in Shanghai. It can analyze the central position, distribution direction, spatial distribution range, and temporal change trend, which can quantitatively describe the spatial-temporal distribution and evolving characteristics of Freshippo stores, and use ArcGIS to visualize the SDE of Freshippo stores in different years.

The standard deviation ellipse is calculated as follows:

$$SD{E}_{x}=\sqrt{\frac{\mathop{\sum }\nolimits_{i=1}^{n}{({x}_{i}-\bar{X})}^{2}}{n}},\,SD{E}_{y}=\sqrt{\frac{\mathop{\sum }\nolimits_{i=1}^{n}{({y}_{i}-\bar{Y})}^{2}}{n}}$$

(3)

Where SDE_x and SDE_y are the axis lengths in the x and y directions of the standard deviation ellipse, the major axis is the direction with the most spatial distribution, and the minor axis is the direction with the least spatial distribution; x_i and y_i are the coordinates of Freshippo stores; $(\overline{X},\overline{Y})$ is the average center of the spatial distribution of Freshippo stores; and n is the total number of stores.

Geographical detectors

Geographic detectors are an effective tool that can detect the spatial distribution of geographical phenomena and their driving factors. They are widely used to solve scientific problems in geography, planning, sociology, and other disciplines (Song et al., 2020; Hu et al., 2024; Li et al., 2024). Compared with other traditional models, Geographic detectors have the advantages of multi-collinear immunity, avoiding the mutual causality between independent variables and dependent variables. With the number of Freshippo stores in the grid as the dependent variable Y, the value ranges from 0 to 2, and the driving factors x_i, as the independent variable, the calculation formula for exploring the significant driving factors of location selection of Freshippo stores is constructed as follows:

$$q=1-\frac{{\sum }_{h=1}^{L}{N}_{h}{\sigma }_{h}^{2}}{N{\sigma }^{2}}$$

(4)

Where q is the explanatory ability of the driving factor X_i, and the value range is [0,1]. The larger the value, the greater the influence. L is the stratification of dependent variables or independent variables, classification, or partition. N_h and σ^h₂ are the number and variance of layer grid cells, respectively. N and σ² are the unit number and variance of the whole study area, respectively.

Binary logistic regression

Binary logistic regression is a statistical analysis method often used in the regression analysis of binary dependent variables. It is a nonlinear model. In recent years, this method has been used to study the distribution analysis of urban geography, commercial geography, and the retail industry (Zhao et al., 2024; Jiao et al., 2024). By assigning values to geographical units after a grid division, the dependent variable Y is assigned 1 and 0, respectively, according to whether there are Freshippo stores in the grid. With each driving factor x_k as the independent variable, the regression equation for exploring the driving factors and driving mechanism of Freshippo stores is constructed as follows:

$${\mathrm{ln}}\frac{{P}}{1-{P}}={\rm{b}}_{0}+{\rm{b}}_{1}{{x}}_{1}+\cdot \cdot \cdot +{\rm{b}}_{{k}}{{x}}_{{k}}$$

(5)

Where b₀ is a constant that represents the natural logarithm of ratio when the independent variable values are all 0; Parameters b₁, b₂,…, b_k are the logistic regression coefficients, indicating that when the values of other independent variables remain unchanged, the increase of one unit in the value of independent variable will cause the change of the natural logarithm of odds ratio (OR); P is the probability of the occurrence of an event Y, with a value range of 0 to 1, and 1-P is the probability when the event will not occur.

Characteristics of the spatial‑temporal evolution of Freshippo stores

Using ArcGIS software for spatial analysis, we visualized the spatial distribution of Freshippo stores in Shanghai from 2016 to 2024, illustrating their spatiotemporal evolution (Fig. 2). We also calculated the nearest neighbor index R using the nearest neighbor algorithm (Table 1). The analysis indicates that there was a significant change in the spatial distribution of Freshippo stores over this time frame.

Fig. 2: The spatial-temporal evolution of Freshippo stores in Shanghai from 2016 to 2024.

a–e indicate the sequence of different years.

Table 1 Nearest neighbor index of Freshippo stores in Shanghai from 2016 to 2024.

Spatiotemporal clustering characteristics

As shown in Table 1, the annual nearest neighbor index (R value) of Freshippo stores has consistently remained below 1, indicating a pattern of clustered spatial distribution. Since 2016, when the Z value was −2.58 with a P value of 0.032, the Z values have consistently stayed below −2.58 with P values under 0.01, emphasizing a continuous intensification of spatial clustering at Freshippo stores. Time series analysis from 2016 to 2024 indicates that the R value initially rose and subsequently fell, depicting a dynamic distribution pattern in urban space that initially expanded and later contracted. Significantly, since 2022, the increased clustering intensity has consistently filled market gaps and strategically ensured a more balanced alignment between supply and demand for Freshippo stores.

As shown in Fig. 2 and Table 1, the spatial-temporal distribution of Freshippo stores has evolved from a dispersed pattern to a single-core clustering pattern, eventually leading to a multi-core clustering structure. In 2016, the number of Freshippo stores was relatively small, exhibiting a predominantly dispersed distribution. From 2017 to 2020, the stores expanded outward from the central urban districts to adjacent peripheral areas and then further into outer suburban districts, forming a clustering structure centered around the urban core and its peripheral districts, while stores in outer suburban districts remained sparse and scattered. In 2021, Freshippo actively explored the sinking market and community economies, resulting in a significant increase in store density in both the urban periphery and outer suburban areas. By 2024, the spatial distribution pattern of Freshippo stores in Shanghai exhibited a multi-core clustering structure, with the urban periphery and outer suburban districts emerging as major high-density clusters, accounting for 83.62% of the total number of stores. This spatiotemporal evolution is closely linked to shifts in the consumer market environment and infrastructure development. Transportation accessibility, consumer demographics, and the business environment are critical factors influencing new retail store location decisions.

Spatial migration trends

The perimeter and area of the standard deviation ellipse effectively reflect changes in the spatial diffusion range of the research subject, while the centroid and rotation angle reveal trends in spatial migration. As shown in Table 1, from 2016 to 2024, the perimeter and area of the standard deviation ellipse have continuously increased, with the area of the ellipse in 2024 being 4.15 times that of 2016, indicating a significant trend in spatial expansion. However, from 2020 to 2024, the growth rate slowed. Regarding the shape of the ellipse, the ratio of the long axis to the short axis decreased from 1.89 in 2016 to 0.91 in 2024, indicating that the ellipse has gradually become more circular. This suggests that the spatial diffusion of Freshippo stores is gradually shifting toward a uniform distribution, possibly because its offline sales network has essentially covered all geographic markets in Shanghai, leading to a slowdown in the momentum of expansion.

The centroid and orientation angle of the standard deviation ellipse have changed significantly (Fig. 2e and Table 2). From 2016 to 2024, the variation in the orientation angle of the ellipse has consistently remained within 90°, showing a trend of first increasing and then decreasing. Starting from the central area of Huangpu District in 2016, the centroid moved to the northern part of Xuhui District in 2018, then to the western part of Huangpu District in 2020, followed by a move to the northeastern part of Xuhui District in 2022, and finally settled in the western part of Xuhui District in 2024. Overall, the centroid of the standard deviation ellipse has moved in a southeast direction, which is directly related to the increase in the number of stores in certain areas. Specifically, in 2018, due to the increase in stores in Songjiang District, Minhang District, and Jiading District, the centroid shifted southwest; in 2020, the increase in stores in Pudong New District, Baoshan District, and Huangpu District caused the centroid to shift eastward; and from 2021 to 2024, the increase in stores in suburban and outer suburban areas further moved the centroid southwest.

Table 2 Standard deviation ellipse of Freshippo stores in Shanghai from 2016 to 2024.

However, our field research found that the changes in the characteristics of the standard deviation ellipse are not only due to the opening of new stores but also to the closure or transformation of a few stores. For example, the Changliu Road store and the Songbin store in Pudong New District have been transformed into outlet stores, while the Zhuqing Chengnan store has been converted into a mini store. The F2 Platinum Bay store in Hongkou District opened in 2018 and closed in 2020. These changes reflect the influence of the regional market environment and commercial demand, indicating that some stores may undergo transformation or eventually close. Therefore, we believe that the commercial expansion of new retail formats aligns with the “industry life cycle theory,” which suggests that after a certain stage of development, they may face decline, disappearance, or the need for innovation and redevelopment.

Preference analyses of the spatial distribution of Freshippo stores

To explore scientific issues, we utilized the comprehensive advantages of multi-source big data to analyze the spatial preferences of Freshippo stores in 2024. Existing studies demonstrate that factors such as regional business environments, community density, rental levels, and transportation conditions significantly influence the location decisions of physical stores. Consequently, to more precisely delineate the interrelationships between key urban factors and the spatial distribution of Freshippo stores, we conducted an in-depth visualization analysis using ArcMap 10.8 and ArcScene 10.8. We then applied statistical methods to assess the range of values for these factors within the store location areas, further investigating Freshippo store’s preferences for various levels of urban factors. We are thankful for your feedback, which is vital for enhancing the quality of our research. To effectively conduct the analysis, we employed the natural breaks method to categorize the data into three levels: high-value areas (levels I and II), medium-value areas (levels III and IV), and low-value areas (levels V and VI).

Spatial correlation of the business environment

Urban business districts are symbols of urban commercial culture, carriers of urban commercial prosperity, and important embodiments of urban construction improvement and growth (Yang et al., 2024; Lima et al., 2024). As places for consumers to shop, business districts are highly concentrated spaces for businesses and properties, representing the temporal and spatial distances that consumers must traverse to engage in consumption activities, which are both local and hierarchical. Business districts radiate and expand in certain directions with commercial stores at their center, and the “business district effect” weakens as the distance increases (Han et al., 2024; Yiu et al., 2024; Guan et al., 2025). The density of commercial facilities and the level of human activity at night are important factors determining the spatial location of urban business districts, reflecting the degree of improvement in commercial infrastructure and social activities. Commercial stores serve as windows for the urban retail industry to conduct business activities and are the main venues for providing sales services to consumers. Based on POI data for commercial facilities and the visualization results from night light remote sensing data, the primary distribution areas of Shanghai business districts are comprehensively identified. This analysis is overlaid with the spatial points of Freshippo stores and the kernel density visualization layer. The results of the spatial overlay analysis indicate that: A comprehensive analysis (Figs. 3b, d and 4) shows that the majority of Freshippo stores are located in areas where the kernel density of commercial infrastructure is between level II and level V, and the night activity index light intensity is above level IV. These stores are primarily concentrated at the intersection of high-value and low-value areas. From a localized perspective (Fig. 3b, d), approximately 85% of Freshippo stores in the central urban area are concentrated in level II and level III kernel density areas, with a small number of stores located at the edges of level I areas. About 88% of stores in the suburban areas are situated in level III and level IV areas, with very few in level II and level V areas. Although the commercial aggregation in the outer suburban areas is relatively low, Freshippo stores are still mainly located in the more prosperous level IV and level V areas. Therefore, the spatial distribution of Freshippo stores demonstrates a clear preference for the commercial environment, indicating that a favorable commercial environment is crucial for the operation of new retail enterprises and market demand.

Fig. 3: The current spatial distribution of Freshippo stores and the spatial correlation of main factors.

a–f are only used to indicate the sequence of different years and are not presented separately in thespecific analysis within the text.

Fig. 4

The spatial distribution of Freshippo stores in business districts.

Spatial correlation of community distribution

Regional consumer purchasing power is a key factor in ensuring the effective operation of physical retail stores (Aversa et al., 2024; Zhang et al., 2024; Qiu et al., 2023; Shankar et al., 2021). Urban communities, with a significant number of stable consumers, have become a primary market for fresh products and consistently provide a steady flow of customers to physical stores. The use of technologies such as big data, cloud computing, and the IoT in the retail sector not only enhances the understanding of both community values and market potential but also supports the development of new retail markets (Vhatkar et al., 2024; Uddin et al., 2024). Based on the overlay analysis of Freshippo’s geographic data and community spatial density data (Fig. 3c), Freshippo’s site selection demonstrates an “inverse U-shaped” relationship with community density. Among these stores, 12% are located in high-value areas, 63% in medium-value areas, and 25% in low-value areas. The local area analysis shows that Freshippo stores are mainly located at high-density intersections, which are seldom found in core commercial areas, helping to lower rental costs. Specifically, stores in urban areas are concentrated in second-tier areas, while those in suburban areas are primarily distributed across third and fourth-tier clusters, and those in outer suburban areas are primarily concentrated in fourth and fifth-tier clusters. This layout strategy reflects Freshippo’s demand for commercial space and the balance between commercial and residential land use.

Spatial correlation of rent level

The intensity and cost of commercial activities are crucial factors in the site selection process for retail stores, especially for retailers specializing in fresh products, as high rents increase operational costs and restrict profit maximization (Brooks and Meltzer, 2024; Ossokina et al., 2024). The rental levels of commercial stores typically reflect the purchasing power of local consumers in those areas (Xu et al., 2022; Akansha et al., 2022). By overlaying the spatial distribution data of Freshippo stores with rental price estimates from a random forest model (Fig. 3e), we found that 65% of Freshippo stores are located in moderate-rent zones, while only 2% are located in high-rent zones. This indicates a spatial misalignment with high-rent zones. Freshippo stores are primarily situated within commercial clusters in suburban and outer suburban areas (Fig. 3b, e), suggesting that this strategy aims to cater to diverse consumer groups and enhance brand presence. The Zhonghai Huanyu Hui, Freshippo’s first mini store, is located in high-rent districts and the Central Business District (CBD), adopting an omnichannel sales strategy to enhance its brand presence.

Spatial correlation of the traffic environment

The service range of commercial stores is influenced not only by the regional business environment, community density, and investment costs but also by the convenience of surrounding road traffic (Merten and Kuhnimhof, 2023; Wang et al., 2024). Transportation and production costs play a decisive role in commercial activities, significantly expanding the scope of these activities (Dey et al., 2023; Vhatkar et al., 2024). In the “new retail” model, the efficient transportation of goods is key, and overall transportation convenience is a core factor in reducing related costs. Furthermore, transportation accessibility and convenience not only affect the costs incurred by consumers when visiting stores but also directly impact their shopping experiences (Kim and Wang, 2021; Zhang et al., 2024). Shanghai boasts an efficient and diverse transportation network, and optimizing transportation services has greatly alleviated the critical “last mile” logistics challenge. Therefore, commercial stores focusing on fresh products tend to prefer locations with convenient transportation. As shown in Fig. 3f, the distribution ratios of Freshippo stores in high-value zones (levels I and II), medium-value zones (levels III and IV), and low-value zones (levels V and VI) of the transportation accessibility layer are 29%, 46%, and 25%, respectively. This indicates that the spatial distribution of Freshippo stores is primarily concentrated in areas with high to medium accessibility, aligning with conventional trends in site selection for commercial stores. Thus, it is evident that the spatial distribution of Freshippo stores in Shanghai is closely related to the transportation environment.

Based on the above analyses, the spatial geometric composition of Freshippo stores covers the main business districts, significant community-concentrated areas, hotspots of human activities at night, areas with gradient concentrations of rental price levels, and areas with high road density in Shanghai. Stores are mainly located in the core and secondary business districts, which exhibit middle to high levels of community density, nighttime lighting, rental prices, and traffic accessibility. Therefore, the spatial distribution of Freshippo stores has a strong correlation with four factors: the business environment, community distribution, rental prices, and traffic accessibility.

Results

In its early stages, Freshippo relied on markets developed by similar and related industry clusters. It gradually emerged as a typical example of innovation-driven new retail models in China’s new retail industry. For the site selection of service commercial stores, the development level of regional modern service industries is also a crucial reference factor. This level serves as an important indicator for measuring the degree and development level of regional production socialization, reflecting the economic development stage and quality of life of local residents (Rana and Paul, 2017). Therefore, based on existing research, we introduced three variables: the density of medical facilities, life service facilities, and accommodation service facilities, to reflect the development level of regional modern services (Song et al., 2020; Imtiaz et al., 2021; Wild et al., 2021). Subsequently, based on the coupling analysis in Chapter 4 on the relationships between Freshippo stores and factors such as rental levels (Akansha et al., 2022; Merten and Kuhnimhof, 2023; Brooks and Meltzer, 2024), business environment (Aversa et al., 2024; Lima et al., 2024; Yang et al., 2024), transportation distance (Kim and Wang, 2021), and community density (Wang et al., 2024; Han et al., 2024), we conducted collinearity and significance analyses using data from multiple databases. We eliminated 12 non-significant factors and validated the rationality of analyzing the correlation of Freshippo stores with these four factors in urban settings. Finally, after verifying that the selected urban factors passed the spatial autocorrelation test, we found that the spatial distribution of all relevant factors exhibited strong clustering patterns, and the null hypothesis of random distribution was rejected. This indicates that the factors influencing the location selection of Freshippo stores themselves exhibit significant spatial heterogeneity (Table 3).

Table 3 Explanation of influencing factors and results of spatial autocorrelation tests.

Therefore, we constructed an indicator system based on costs, regional environment, and service industry dimensions and processed the data using Geographical detectors and Binary logistic regression methods.

Result by using geographic detectors

The results of the geographical detector model demonstrate that at the urban level (Table 4), the spatial distribution of Freshippo stores is significantly influenced by several factors. These influences can be classified into single-factor effects, two-factor interactions, and multi-factor interactions. The specific results are shown in Table 4, where the explanatory ability of single factor from high to low is regional accessibility x₂ (0.170) > living service x₇ (0.160) > rent level x₁ (0.159) > community density x₄ (0.151) > healthcare service x₆ (0.113) > business environment x₅ (0.109) > accommodation service x₈ (0.106) > transportation distance x₃ (0.023). Each factor is significant at the level of 1%, which indicates that each factor in the indicator system has a strong influence on the dependent variable, and the single factor detection result of x₂ (0.170) has the strongest explanatory ability. Given the significant spatial heterogeneity of these factors, they clearly vary in influencing the distribution of Freshippo stores across different geographical locations. Based on the results of single-factor detection, the factors with significant influence are interactively detected. Among the two-factor interaction results, regional accessibility (x₂) has the strongest influence on other factors, with the density of living service facilities (x₇) coming second. The interaction q value has also increased more than times compared with the single-factor interaction, and the other two-factor interaction q value has significantly increased compared with the single-factor interaction. Based on the results of the geographical detector experiment, we draw the following conclusions: Under the comprehensive influence of multiple factors, the combined detection results of single-factor detection and double-factor interaction can better explain the location preference of Freshippo stores. After the interaction of various influencing factors, a total of 28 types of interaction results were produced, mainly showing the interaction type enhanced by two factors. Following the interaction of multiple factors, a nonlinear enhancement effect is observed, meaning that the individual effect of any two factors is weaker than their combined effect, resulting in a “1 + 1 > 2” effect on the spatial distribution of Freshippo stores. These findings from the two-factor interaction analysis clearly demonstrate how the spatial clustering characteristics of Freshippo stores are comprehensively shaped by the diversity of multiple factors.

Table 4 Single-factor and double-factor interactive detection results of Freshippo stores in Shanghai.

Result by using binary logistic regression

To overcome the limitations of the Geographical detector in quantitatively demonstrating the effects of influencing factors on the spatial distribution of Freshippo stores, we adopted a Binary logistic regression model for further analysis. By quantifying the effects of each factor, this method enhances the robustness and precision of the analysis, thereby offering a stronger scientific foundation for the research conclusions. First, we assign values to the grid units of the urban central districts, urban periphery areas, and outer suburban districts of Shanghai. These assigned difference indicators are then incorporated into the Binary logistic regression model (Table 5). The forward LR method is used to obtain the regression model after fitting with the default maximum number of iterations (20 times). If the partial regression coefficient B > 0 and OR > 1, it indicates that the independent variable and dependent variable are positively correlated. Conversely, it indicates a negative correlation between the independent variable and the dependent variable. The knife-cutting method uses a 50% prediction probability as the dividing point to determine the correct classification ability. Results show that:

Table 5 Binary logistic regression result of distribution and factors of Freshippo stores in Shanghai.

There are significant differences in the factors that affect the spatial distribution of Freshippo stores in the whole region, urban central districts, urban periphery areas, and outer suburban districts of Shanghai. Among them, according to the city’s model I results, rent level (x₁), regional accessibility (x₂), transportation distance (x₃), business environment (x₅), and accommodation service (x₈) are significant at the level of 1%, while community density (x₄), healthcare service (x₆), and living service (x₇) are significant at the level of 5%. The results of model II in the urban central districts show that the healthcare service (x₆) and living service (x₇) are significant at the level of 5%. The results of model III in the urban periphery areas show that regional accessibility (x₂) and business environment (x₅) are significant at the level of 1%, while rent level (x₁) and community density (x₄) are significant at the level of 5%. The results of model IV in the outer suburban districts show that rent level (x₁), regional accessibility (x₂), and business environment (x₅) are significant at the level of 1%, while transportation distance (x₃) and community density (x₄) are significant at the level of 5%.

Comparing the results of the models shows that some factors are significant only in the overall city model, not in the local regional models. This supports the rationality of the “overall-local” analysis framework, effectively revealing key factors influencing the spatial distribution of new retail stores. Moreover, the prediction accuracies of the four models are 95.4%, 80.5%, 94.3%, and 97.3%, respectively, each significantly exceeding 50%. Such high accuracy underscores the excellent fit and stability of the model operations, confirming the appropriateness of the selected factors. When other influencing factors remain unchanged, the change probability of store location preference is determined by the odds ratio (OR value) as a single significant independent variable increases or decreases.

Comprehensive discussion of the model results

By leveraging the advantages of both the Geographical detector and Binary logistic regression models in data processing, we can more accurately reveal the significant impacts of individual factors, two-factor interactions, and multi-factor interactions, as well as their driving mechanisms on the spatial distribution of Freshippo stores (Fig. 5). The following findings can be drawn:

Fig. 5

The spatial distribution of Freshippo stores in communities.

(1) With other variables held constant, each individual factor exhibits significant differential effects. According to the OR value analysis, keeping other conditions constant, an increase of one value level of any individual factor, such as rent level (x₁), regional accessibility (x₂), community density (x₄), business environment (x₅), or living service (x₇), will increase the likelihood of opening Freshippo stores. This results in the odds of opening events being 1.091, 1.010, 1.011, 1.200, and 1.009 times higher than those for not opening, respectively. In contrast, an increase in any of the factors, including transportation distance (x₃), healthcare service (x₆), or accommodation service (x₈), will decrease the probability of opening stores, resulting in the odds of opening events being 0.329, 0.979, and 0.972 times compared to not opening, respectively.

(2) The combined effect of two factors generates a stronger driving effect, and the interaction between any two factors has a more significant impact on the spatial clustering characteristics of Freshippo stores. For example, the interactions of factors x₂∩x₁ (0.883), x₁∩x₄ (0.741), x₂∩x₃ (0.772) significantly enhance the probability of site selection, while the interactions of factors x₃∩x₆ (0.414), x₃∩x₈ (0.361), x₆∩x₈ (0.333) significantly weaken the probability of site selection.

(3) Through the comprehensive validation of the dual models presented in Tables 4 and 5, we reveal the significant impacts of various factors. With other variables held constant, increasing any favorable factor or decreasing any unfavorable factor can increase the likelihood of opening stores; conversely, reducing favorable factors or increasing unfavorable factors will decrease the likelihood. Therefore, the analysis shows that when multiple favorable factors associated with opening stores are enhanced simultaneously, the likelihood of successfully opening stores significantly increases, and vice versa.

(4) Based on the comprehensive analysis of the detection results presented in Table 5 for Shanghai, the spatial distribution of Freshippo stores is significantly influenced by cost, regional environment, and service industry factors. Specifically, the detection results indicate that Freshippo stores tend to select areas with high values of factors x₁, x₂, x₄, x₅, and x₇, while avoiding areas where factors x₃, x₆, and x₈ are predominantly high. Local detection results reveal that the urban central districts are significantly influenced by factors x₆ and x₇, while the urban periphery and outer suburban districts are significantly affected by factors x₁, x_2, x₃, x₄, and x₅. This diverse spatial distribution underscores Freshippo stores’ specific preferences for various urban elements in their commercial space selection. This differentiated impact mechanism underscores Freshippo’s distinct preferences for various urban elements in different regions, which is revealed by the distinctly stratified spatial distribution of Freshippo stores in urban centers, suburban areas, and outer suburban districts.

The adaptation of consumer demand to new retail formats

Through qualitative analysis of the respondents’ statements, it is evident that the diverse sales methods provided by Freshippo stores cater to consumers’ needs in different scenarios. When consumers need to quickly obtain products or experience offline shopping, going to the nearest store is the quickest and most direct choice, allowing them to get the required items immediately and avoid waiting for delivery. These consumers are more inclined to choose stores located in areas with convenient transportation. For those who are unwilling to go out, they can place orders at any time through the app and benefit from real-time logistics delivery services to better manage their time. These consumers particularly favor stores offering real-time delivery services, especially during adverse weather conditions, epidemic prevention periods, or peak traffic times, making this model their preferred choice. Additionally, when consumers wish to save time on both offline shopping and delivery, they are more likely to place an order on the app and then pick up the pre-packaged items at a convenient time. Convenience-oriented consumers expect Freshippo to open stores in community areas with parking facilities to meet the demand for quick pickup. These qualitative analysis results are generally consistent with the conclusions of existing studies on consumer behavior in the new retail format (Wang and Coe, 2021; Zhang et al., 2024; Gao et al., 2025; Wang et al., 2025; Huang et al., 2024; Feng et al., 2022).

Discussions

This study systematically explores the spatiotemporal distribution characteristics and influencing factors of Freshippo stores, and analyzes the evolution of this innovative retail model introduced by Alibaba. Particularly in the Chinese food retail market, the deep integration of online and offline channels is progressively reshaping consumer shopping experiences and consumption patterns. With significant advantages in data creation, collection, and analysis, Freshippo has rapidly established a leading position in a highly competitive market environment. The comprehensive analysis reveals that the spatial distribution of Freshippo stores is significantly influenced by factors such as cost, location environment, and service industry levels. Based on the results of qualitative and quantitative analyses, this research constructs a systematic discussion framework (Fig. 6) to reveal the key mechanisms behind the site selection strategy of Freshippo stores.

Fig. 6

The spatial distribution of Freshippo stores in areas with different rent levels.

Spatial evolution and location logic of new retail stores

Traditional retail stores primarily cater to consumers who prefer physical shopping, focusing on creating a comfortable and pleasant environment to attract customers and maximize direct sales through foot traffic to achieve operational goals. This leads traditional retail stores to prefer locations with large spaces, high centrality, and high rents in prime locations, which align with central place theory and rent theory (Gao et al., 2025; Berry et al., 1988). E-commerce stores, as a form of retail conducted through digital channels, serve only internet users who are willing to shop in virtual spaces. In contrast, new retail stores must cater to both online and offline consumer segments. They must not only ensure a high-quality in-store shopping experience but also meet the demands of online consumers for real-time logistics and delivery services, and consider how to enhance brand effectiveness to attract more customers (Quinones et al., 2023; Li and Shi, 2024). Specifically, the new retail model combines the advantages of traditional retail and e-commerce models, optimizing inventory management between different stores through internet technology, real-time tracking of demand data between warehouses and consumer markets, and improving supply chain efficiency (Gao et al., 2025; Zhang et al., 2024; Wang and Coe, 2021). These factors make the location strategy for new retail stores more complex and comprehensive. Based on the differences in the target audiences of traditional retail stores, e-commerce stores, and new retail stores, it is evident that Freshippo is a typical representative of the new retail format.

Unlike the traditional retail practice that relies on urban commercial centers, the new retail model attempts to address the long-standing issue of geographic market isolation by optimizing the layout strategy of offline stores. For a long period, retailers focused their market competition strategies on the heavily homogenized urban commercial centers. The application of information technology in retail has led some retailers to shift consumer market demand toward the suburbs (Zhou et al., 2024; Ballantyne et al., 2023; Zhang et al., 2024; Roggeveen et al., 2020). A data-driven approach is used to determine the general area for store location selection and the scale of stores that match regional market demand. This method not only relies on geographic factors such as traffic convenience and commercial prosperity but also takes into account the specific needs and purchasing power insights of the potential consumer market within the region (Xu et al., 2023; Tang and Chen, 2023; Ge and Zhu, 2023; Lu and Menezes, 2024). Additionally, widespread internet access has led to a rapid increase in the number of Chinese internet users, resulting in a shift in consumer habits toward more convenient and mobile online shopping (Gao et al., 2025). Freshippo, based on the new retail model, rapidly expanded in a short period to cover all districts of Shanghai and presented a significant “hierarchical layout” in urban space. The hierarchical layout is reflected in the fact that new retail stores choose different scales and service types based on the consumption characteristics and market maturity of different urban areas, forming a complementary geographic market (Wang and Coe, 2021; Yin and Ye, 2024; Zhang et al., 2024). For example, in densely populated and high-consumption urban center areas, Freshippo chooses to open large, comprehensive stores, focusing on providing consumers with a shopping experience that integrates “online + offline + logistics delivery.” This setup allows consumers to select fresh ingredients, ready-to-eat foods, beverages, etc., and offers various service types such as physical experiences, logistics delivery, and “online + self-pickup” services. In suburban and outer suburban areas, Freshippo emphasizes the “neighborhood” concept by opening small convenience stores that support self-pickup services, primarily meeting daily fast consumption needs.

Drivers of a new retail store location

Currently, a large number of linear models are used in related fields of research, but they can only present the results of single-factor driving after excluding the effects of multicollinearity and cannot reveal the interactions of influencing factors on retail location (Gao et al., 2025; Zhou et al., 2024; Xu et al., 2023; Reed et al., 2023; Formánek and Sokol, 2022). This study, by integrating geographic detectors and binary logistic regression models, extends beyond the limitations of traditional linear analysis in explaining the factors affecting new retail store site selection, further revealing the complex roles of factors in new retail store location decisions. Geographic detectors can detect the explanatory power of individual factors and their two-factor interactions, while Binary logistic regression models quantify the degree and direction of the influence of these factors, further enhancing the stability and scientific nature of the experimental results. For example, the rent level and regional accessibility show significant explanatory power in the site selection of Freshippo stores, and the two-factor interaction results indicate that the interaction of these two factors has a nonlinear enhancing effect. The quantification results of the binary logistic regression model further reveal that these factors have a significant positive impact. This method reveals the driving effects of single, double, and multiple factors on the dependent variable and provides a more comprehensive experimental analysis framework, which has not been reported in the relevant literature. This finding offers valuable technical means for enriching the case study approach in retail geography.

By comparing the research findings of traditional retail stores, this study reveals a clear differentiation trend in the site selection strategies of new retail stores (Berry et al., 1988; Jones and Livingstone, 2018; Brooks and Meltzer, 2024; Ossokina et al., 2024). Traditional retail stores usually ensure a stable flow of customers by selecting areas with high community density. However, the distribution of Freshippo stores shows an “inverted U-shaped” relationship with community density, indicating a preference for medium-density areas (Fig. 7). With the help of real-time logistics in new retail, Freshippo is able to expand the geographic coverage while avoiding high rents and maintaining stable customer flow, thus achieving an innovative “optimal location selection.” Rent theory suggests that rent levels not only directly affect business profitability but also reflect the purchasing power of regional consumers (Xu et al., 2022; Akansha et al., 2022). However, in local areas, Freshippo stores are typically located in regions with moderate-rent levels (Fig. 8), rather than in the high-rent areas favored by traditional retail. This choice helps balance the relationship between rental costs and final business benefits (Feng et al., 2022). Retail geography indicates that costs related to geographical factors can bring operational challenges. High labor and transportation costs between distant warehouses and stores can hinder the flow of goods from storage to sales, thus weakening profits (Dey et al., 2023; Vhatkar et al., 2024). Traditional retail stores generally rely on manual stocktaking and fixed inventory management systems, with location selection primarily considering foot traffic generated by business districts. However, site selection strategies now also need to take into account product distribution, logistics, and customer convenience in reaching the stores, which is why new retail stores are mainly located in areas with better traffic conditions (Fig. 9).

Fig. 7

The spatial distribution of Freshippo stores in areas with different road accessibility.

Fig. 8

The factors of spatial distribution of Freshippo stores.

Fig. 9

Discuss the framework.

In different urban areas, Freshippo uses a hierarchical spatial layout strategy to cover different geographic markets in central urban areas, suburban areas, and outer suburban areas (Fig. 4), which contrasts sharply with traditional retail stores’ preference for highly concentrated commercial areas (Ballantyne et al., 2023; Zhang et al., 2024). Additionally, when assessing the specific impact of service facility density, the site selection strategy of new retail stores retains some characteristics similar to traditional retail stores (Xu et al., 2022; Gao et al., 2025; Formánek and Sokol, 2022). Freshippo stores tend to select areas with high living service facility density and low healthcare and accommodation service facility density (Tables 4 and 5), a fact related to the target consumers of different types of service facilities. The differences in the spatial site selection factors between traditional and new retail stores indicate that the spatial layout strategy of new retail stores places more emphasis on detailed consideration of various urban factors. The integration of internet technology, real-time logistics, and retail not only provides strong support for new retail stores to achieve a new “optimal location selection,” but also lays the foundation for providing consumers with “online + offline + logistics delivery” services.

Comparative analysis of the hard-discount format

Building on the preceding discussion of retail models and their locational and strategic drivers, this study introduces ALDI, a key competitor of Freshippo in the Shanghai retail market, as a case study examining Freshippo’s spatial logic and operational positioning. Previous studies indicate that ALDI, a globally renowned hard-discount retailer, has consistently prioritized low prices as its core business objective. Its store network is characterized by compact formats, integration with local communities, and a focus on accessibility. In terms of site selection, ALDI favors locations with low rents and convenient transportation, often situated on the outskirts of residential areas or along major roads. Its stores are typically equipped with parking facilities for shoppers arriving by car, thereby attracting consumers who are primarily price-sensitive (Spanjaard and Freeman, 2023; Hökelekli et al., 2017; Uttke, 2011; Cleeren et al., 2010). In contrast, Freshippo adopts a coordinated multi-format strategy, establishing service-oriented store formats across Shanghai and leveraging its online platform to develop a multi-tiered, complementary spatial distribution that offers broad coverage and high adaptability among diverse consumer groups (Wang and Coe, 2021; Zhang et al., 2024). Notably, the Freshippo outlet store, as a component of the broader system, serves as a price-focused format within the firm’s diverse portfolio (Zhang et al., 2024). From an operational perspective, it mirrors ALDI stores: both attract price-sensitive consumers by offering low prices and tend to choose sites with low rents, convenient transportation, and proximity to target consumer markets. This approach helps reduce operating costs and improve accessibility. In addition, the Freshippo outlet store specializes in selling excess inventory, near-expiry products, and low-cost procurement items sourced from other formats within the Freshippo system, with the aim of quickly clearing inventory at discounted prices. Overall, this comparison suggests that within its multi-format portfolio, Freshippo’s introduction of price-oriented formats such as the outlet store enables effective penetration into lower price segments, expands its consumer base, and enhances its capacity to mount an effective response to direct competition from hard-discount retailers. However, this strategy also requires a more nuanced balance, as Freshippo must carefully navigate trade-offs among brand image maintenance, cost control, and differentiated competition.

Conclusions and suggestions

Conclusions

This study adopts an interdisciplinary approach to examine the spatiotemporal characteristics of new retail store expansion and explores the key factors influencing the site selection of new retail stores. The main contributions of this research are as follows: Firstly, it systematically analyzes the evolution of new retail stores within urban spaces, using Freshippo stores in Shanghai (2016–2024) as a case study. Secondly, by qualitatively comparing the findings from existing studies on traditional retail stores, this study analyzes the critical factors and targeted site selection strategies for new retail stores. Thirdly, compared to existing quantitative models, this study leverages the strengths of Geographic detectors and Binary logistic regression models to quantify the effects of individual factors, two-way interactions, and multifactorial influences, thereby enriching the methodological toolkit for case studies in retail geography. By expanding the theory of business location and conducting empirical analysis in the field of retail geography, this study deepens the understanding of the dynamic evolution of retail spaces under new retail conditions. The main conclusions of the study are as follows:

The precise location strategy of new retail stores aligns their spatial distribution characteristics with the “optimal location choice” principle of commercial site selection theory. Unlike traditional retail stores, which heavily rely on CBDs, new retail enterprises emphasize comprehensive coverage across diverse geographic markets. The overall scale of Freshippo stores continues to expand, with increasingly significant spatial clustering characteristics, evolving from “small-scale clusters” to “large-scale clusters.” Additionally, the number of stores in the main urban area, suburban areas, and outer suburban areas has changed significantly at various stages, with the spatial distribution center gradually shifting from the main urban area to suburban and outer suburban areas, further indicating Freshippo’s ongoing expansion into new geographic markets. The differentiated layout strategy based on regional environments not only facilitates inter-regional linkages in commercial activities but also effectively breaks through the physical constraints of traditional retail service areas, gradually redirecting consumer traffic originally concentrated in traditional commercial centers to residential areas and surrounding regions, thereby enhancing the overall balance and efficiency of market coverage.

Through a comprehensive study using multidisciplinary models, we have validated the robust and effective identification and selection of influential factors within the index system, providing a quantitative analysis of the intensity and modes of their effects. Results from Geographic detectors show that the explanatory power of each factor regarding the spatial distribution of Freshippo stores varies significantly. Moreover, the interactions among these factors significantly influence the distribution patterns of new retail stores. Regional accessibility exerts the greatest influence, and its interactions with other factors yield the greatest combined effect. The Binary logistic regression model demonstrates both the significance and direction of the independent variables’ effects on the dependent variable. Furthermore, the mechanisms through which various factors influence the spatial distribution of new retail stores differ significantly across regions. Specifically, the impact strength of factors such as cost, locational environment, and the level of service industry development varies by region, which highlights economic, social, and transportation differences, thereby underscoring their importance in new retail store site selection. The findings suggest that retailers, when devising strategies for new retail network layouts, must consider the characteristic and contextual differences across regions to ensure effective site selection decisions.

New retail enterprises that comprehensively cover consumer markets are reshaping the urban commercial landscape and profoundly affecting traditional commercial spaces. Through precise site selection strategies and market coverage, new retail enterprises have not only revolutionized the shopping experience for consumers but also enhanced the overall competitiveness of the business environment. Empirical research on Freshippo stores shows that their reliance on core business districts is gradually weakening as they shift towards diversified and expansive market layouts. Simultaneously, the commercial clustering patterns in urban central districts, peripheral areas, and outer suburban districts have undergone significant changes under the influence of new retail formats, promoting the reconstruction of consumption spaces and the rise of community commerce. These changes compel traditional businesses to respond proactively to market dynamics by innovating their models or transforming their business formats to adapt to evolving demands, thereby maintaining competitiveness and ensuring long-term sustainability.

However, this study presents certain limitations that require further refinement and in-depth exploration in future research. First, since China’s fresh food new retail industry largely originates from the transformation and upgrading of traditional retail enterprises, the sample size of representative new retail enterprises is relatively small. This limitation has prevented a systematic comparative analysis of the spatial-temporal evolution characteristics across different new retail formats. Future research should continue to track the development of new retail formats, with a particular focus on identifying and incorporating a broader range of exemplary cases that have undergone successful transformation and innovation. It should also conduct in-depth comparative analyses of spatiotemporal distribution characteristics, locational logic, and operational strategies across different types of new retail physical stores, and explore heterogeneous impact mechanisms by store type. Second, this study systematically analyzes how new retail physical store locations are influenced by various urban spatial factors from an interdisciplinary perspective, but it does not delve into the micro-level decision-making processes of site selection or consumer behavior patterns. Future research will attempt to introduce more qualitative research methods, such as in-depth interviews and focus group discussions, to complement quantitative analysis. This approach aims to further reveal the underlying driving factors and potential mechanisms of new retail store site selection, providing enhanced theoretical support and practical guidance for effective site selection strategies.

Suggestions

Personalized and diversified retail service positioning strategies are essential in the era of new retail, in which these product development trends are becoming increasingly evident. Freshippo operates a diversified portfolio of formats, including comprehensive flagship stores, dining-experience stores, membership-based warehouse clubs, and value-oriented outlets, thereby delivering differentiated value propositions to a wide range of consumer groups. Notably, value-oriented formats exhibit certain similarities to hard-discount models such as ALDI in pricing stance and siting logic, whereas other formats place greater emphasis on premium product offerings, in-store experience, and rapid fulfillment. To secure sustained consumer support, retailers should undertake comprehensive analyses of consumer behavior and decision-making processes when designing product allocation strategies, with a particular focus on meeting residents’ consumption needs (e.g., fresh food, prepared meals, beverages). Maintaining strong alignment between product offerings and consumer demand enhances competitiveness and the long-term sustainability of store operations. Furthermore, leveraging internet technologies to improve not only the precision of product allocation decisions but also to optimize the spatial mix of store formats, thereby strengthening market resilience within an increasingly competitive retail environment.

To address the mismatch between the spatial distribution of urban retail stores and that of consumer markets, researchers suggest analyzing spatiotemporal patterns and conducting cluster analysis to identify potential market opportunities within regions. Freshippo completed its network rollout across Shanghai’s central, suburban, and outer suburban districts over a 9-year period, achieving market penetration while avoiding undifferentiated competition. This case illustrates that future site selection strategies should focus not only on market potential and accessibility but also on the density and types of competing retail formats and, in highly competitive areas, on adopting functional differentiation or service specialization. However, in practice, it is essential to ensure that store positioning aligns closely with consumers’ distinct shopping behaviors, whether offline, online with real-time delivery, or online with self-pickup. With this approach, retailers can achieve both cost savings and improved operational efficiency. In addition, incorporating competitive landscape indicators into site selection models can guide the deployment of differentiated store formats, ensure complementarity among formats, and help maintain market share in key segments.

Planning optimization solutions for urban commercial spaces in the new retail environment. In this context, urban residents’ consumption patterns have become increasingly flexible and are no longer constrained by traditional temporal and spatial limitations. In response, urban planners and managers should conduct comprehensive, multi-source data analyses to optimize the spatial configuration of urban commercial spaces. Key factors include the market environment, consumer behavior, accessibility, and the broader competitive landscape; these should be evaluated from multiple perspectives to ensure scientific rigor and planning effectiveness. When assessing conditions across different urban areas (central districts, peri-urban zones, and outer suburbs), it is essential to recognize regional heterogeneity and develop targeted strategies. Incorporating spatial mapping of the competitive landscape into the planning process, along with the appropriate siting and allocation of supporting facilities, such as parking facilities and logistics nodes, can reinforce the advantages of different retail formats, enhance operational efficiency, and promote balanced development across regional markets.