Next Article in Journal
Examining Consumer Impulsive Purchase Intention in Virtual AI Streaming: A S-O-R Perspective
Previous Article in Journal
Dynamic Supply Chain Decision-Making of Live E-Commerce Considering Netflix Marketing Under Different Power Structures
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

The Power of Interaction: Fan Growth in Livestreaming E-Commerce

School of Business Administration, Southwestern University of Finance and Economics, Chengdu 611130, China
*
Author to whom correspondence should be addressed.
J. Theor. Appl. Electron. Commer. Res. 2025, 20(3), 203; https://doi.org/10.3390/jtaer20030203
Submission received: 12 June 2025 / Revised: 21 July 2025 / Accepted: 30 July 2025 / Published: 6 August 2025

Abstract

Fan growth serves as a critical performance indicator for the sustainable development of livestreaming e-commerce (LSE). However, existing research has paid limited attention to this topic. This study investigates the unique interactive advantages of LSE over traditional e-commerce by examining how interactivity drives fan growth through the mediating role of user retention and the moderating role of anchors’ facial attractiveness. To conduct the analysis, real-time data were collected from 1472 livestreaming sessions on Douyin, China’s leading LSE platform, between January and March 2023, using Python-based (3.12.7) web scraping and third-party data sources. This study operationalizes key variables through text sentiment analysis and image recognition techniques. Empirical analyses are performed using ordinary least squares (OLS) regression with robust standard errors, propensity score matching (PSM), and sensitivity analysis to ensure robustness. The results reveal the following: (1) Interactivity has a significant positive effect on fan growth. (2) User retention partially mediates the relationship between interactivity and fan growth. (3) There is a substitution effect between anchors’ facial attractiveness and interactivity in enhancing user retention, highlighting the substitution relationship between anchors’ personal characteristics and livestreaming room attributes. This research advances the understanding of interactivity’s mechanisms in LSE and, notably, is among the first to explore the marketing implications of anchors’ facial attractiveness in this context. The findings offer valuable insights for both academic research and managerial practice in the evolving livestreaming commerce landscape.

1. Introduction

With the rapid advancement of information technology and the widespread adoption of mobile devices, livestreaming e-commerce (LSE) on social networking platforms such as Douyin Live and Sina Live has experienced remarkable growth. According to a public report, the number of enterprises in China’s LSE industry reached 18,700 in 2022, representing a year-on-year increase of 17.61%. During the 2022 Double Eleven shopping festival—China’s equivalent of Black Friday—leading platforms like Douyin and Taobao Live attracted nearly 500 million users and generated sales totaling CNY 181.4 billion. Given the growing popularity of LSE and its key role in the development of e-commerce, many researchers have investigated this phenomenon, with a primary focus on the antecedents of LSE performance, such as customer engagement, sales, conversion rates, gifting behavior, return rates, and impulse buying [1,2,3,4,5,6,7]. Studies have found that anchor characteristics, product types, social contexts, perceived experience, perceived trust, scarcity, technological fluency, and promotional strategies are key factors influencing LSE performance [8,9,10,11,12,13,14,15]. Despite numerous findings, these studies have mostly focused on short-term performance in LSE, neglecting research on long-term performance. In LSE, anchors with larger fan bases wield greater attraction, influence, and premium-generating capabilities, thereby creating substantial business value for brands and livestreaming managers [16,17]. However, despite the growing popularity of LSE, the average fan conversion rate remains below 1.5%, posing challenges for platforms and managers in attracting and retaining fans. Additionally, a concerning trend has emerged where top-tier anchors dominate approximately 80% of livestreaming traffic. This dominance creates a skewed distribution of traffic, increasingly marginalizing smaller anchors and potentially pushing them off the platform [18]. Unfortunately, no comprehensive theory currently explains why fan growth varies significantly across livestreaming rooms.
Interactivity is a crucial feature of LSE. In this context, anchors introduce and promote products to users in real time through the camera. Users can ask personalized questions regarding product quality, after-sales service, logistics, and more, while also engaging with other users through bullet comments and other interactive features [19,20], thereby forming a dynamic, interactive community that transcends physical boundaries. This process greatly enhances users’ sense of presence and product experience, creating favorable conditions for positive behavioral decisions [21,22]. According to the Interactive Ritual Chain Theory (IRCT), as users watch livestreams and interact with both the anchor and other viewers, they form an interactive ritual chain that fosters emotional attachment and sustained engagement in livestreaming activities [23]. However, interactivity may also yield adverse effects, such as user attrition and the disruption of online community relationships [24,25], underscoring the complex impact of interactivity on user behavior and marketing performance. Therefore, it is necessary to empirically investigate the role of interactivity in LSE to design more effective interaction strategies that support its sustainable growth. Meanwhile, with increasing competition across the livestreaming industry, user attention has become highly fragmented. Standing out among a multitude of livestreams and maintaining high user retention have become crucial to LSE’s success. Prior research suggests that when users perceive interactions as informative, entertaining, and practical, they tend to engage more actively, devote more time to in-depth exchanges, and ultimately enter an immersive state that promotes positive behaviors [26,27]. Thus, whether interactivity influences fan growth by enhancing user retention remains an important question worthy of further exploration.
Furthermore, we hypothesize that an anchor’s facial attractiveness may moderate the relationship between interactivity in LSE and user retention. According to Dual Systems Theory (DST), observable user features in online environments function as important heuristic cues that shape consumer decision-making [28]. These cues often complement content-based information, assisting consumers in evaluating the credibility of the source [29]. In the context of LSE, an anchor’s facial attractiveness serves as the first visual cue encountered in a livestreaming room, playing a crucial role in shaping users’ initial impressions of the marketing content. Previous research suggests that individuals with higher facial attractiveness are perceived as having greater social skills compared to their less attractive counterparts and are generally more effective in domains such as product marketing [30]. Therefore, it is plausible that the impact of interactivity on user retention in LSE may vary depending on the anchor’s facial attractiveness. However, existing studies have largely overlooked the potential interaction effects between livestreaming room characteristics (e.g., interactivity) and anchor-specific attributes, thereby limiting our understanding of their combined influence on user behavior. These considerations lead to the following research questions:
RQ1: How does interactivity influence fan growth in LSE?
RQ2: What role does user retention play in mediating the relationship between interactivity and fan growth?
RQ3: Does an anchor’s facial attractiveness moderate the relationship between interactivity and user retention?
The contributions of this study are reflected in the following three aspects: (1) This study reveals the significant impact of interactivity on fan growth and uncovers its underlying mechanisms in the context of LSE, thereby broadening the research scope of LSE performance. Furthermore, it identifies a substitution effect between livestreaming room characteristics and anchor characteristics in influencing user retention. (2) By employing objective data, the empirical analysis validates the applicability of Interactive Ritual Chain Theory (IRCT) and Dual Systems Theory (DST) within the LSE context, enriching the theoretical foundation of the field and offering new directions for future research. (3) This study offers practical implications for the operation and management of LSE, particularly in the areas of anchor selection and training.
The remainder of this paper is structured as follows: Section 2 reviews the relevant literature on LSE, Interactive Ritual Chain Theory (IRCT), and Dual Systems Theory (DST). Section 3 presents the main research hypotheses. Section 4 describes the study design, data collection procedures, and variable measurements. Section 5 provides descriptive statistics and reports the empirical findings. Finally, Section 6 discusses the results, draws conclusions, offers marketing management recommendations, and outlines directions for future research.

2. Literature Review

2.1. Performance Research in LSE

As an innovative online sales model, LSE integrates image transmission technology with social interaction, fundamentally transforming how people shop and engage with one another [6,31]. Unlike traditional e-commerce, LSE enables real-time interactions between anchors and users, as well as among users themselves, offering richer information and more engaging, immersive experiences. This unique feature has attracted a growing number of merchants and consumers to participate in LSE [7,32], thereby prompting increasing interest in its performance metrics. Among these, consumer purchases and reward-giving behaviors are the most frequently studied indicators [2,3,5,17]. Previous studies have identified several critical factors that influence consumer purchases and sales outcomes in LSE, including the variety of products offered [33], the level of social interaction [34], the presence of informational cues [35], and anchor characteristics [36]. From the perspective of dual attachment, Zhang et al. [37] found that both bond-based and identity-based social attachments significantly drive users’ reward-giving behavior. However, due to the real-time nature of LSE, researchers have also turned their attention to its potential downsides, such as compulsive purchasing and order cancellations. Sun and Bao [19] demonstrated that fear of missing out and social influence play a significant role in driving compulsive purchases during livestreaming sessions. Similarly, Wang et al. [38], drawing on emotional contagion theory, identified group emotion as a key factor contributing to consumers’ impulse-driven order cancellations in LSE.
Despite the growing body of research on LSE performance, the critical performance indicator of fan growth has been largely overlooked. In the context of LSE, fan growth typically refers to the process by which ordinary users are converted into loyal followers of a particular brand, product, or anchor through targeted strategies and activities [16]. Zhao et al. [17] noted that anchors with larger fan bases tend to exhibit greater appeal, influence, and monetization potential, thereby enhancing the operational value of both the brand and the livestreaming room. Prior research has also shown that fan growth can serve as a leading indicator of users’ purchase intentions in livestreaming settings [18]. However, few studies have directly investigated fan growth within the LSE context, and its underlying drivers remain insufficiently understood. This study seeks to address this gap by examining the determinants and mechanisms of fan growth through the lens of interactivity, thereby offering theoretical insights to inform the effective operation and strategic management of LSE platforms.

2.2. Interactivity in LSE

Interactivity is a core advantage and defining feature of LSE, a consensus widely recognized in both academic research and industry practice [24]. In LSE, anchors engage with users in real time through product demonstrations, topical discussions, and timely responses, fostering a positive and dynamic environment. Such engagement enhances consumers’ perceived value and significantly influences their behavior [19]. Existing research on interactivity in LSE primarily centers on three key areas: First, scholars have examined how interactivity affects consumers’ purchase intentions and participatory behaviors through various psychological mechanisms and moderating factors. For instance, interactivity has been found to increase perceived usefulness, reduce perceived risk, and strengthen psychological connections with consumers, thereby promoting purchasing behavior [15,34,39,40]. Additionally, interactivity is closely associated with consumers’ impulse-buying tendencies [41]. Second, researchers have investigated the role of interactivity in value co-creation within online communities. Studies indicate that user participation in value co-creation is influenced by interactive behaviors, with the transition from human–computer interaction to human–human interaction largely driven by factors such as community incentives, social enhancement, and mutual co-creation [42]. Third, studies have explored the constituent elements of interactivity across different contexts. Kang et al. [24] conceptualized interactivity as comprising two dimensions: responsiveness and personalization. Responsiveness refers to the speed at which anchors respond to user input, while personalization pertains to the degree to which the information or service aligns with users’ specific needs. Bonner [27] further characterized interactivity through bi-directionality, participation, and joint problem-solving, highlighting the intensity and richness of user engagement. Moreover, the frequency and quality of interaction have been identified as critical indicators of interactive behavior [43].
Although existing studies have examined the value of interactivity in LSE, few have specifically investigated its effect on fan growth. Moreover, prior research has primarily focused on users’ behavioral intentions based on relatively subjective factors, such as psychological perceptions [12,14], while overlooking the mechanisms through which more objective factors—such as user retention—influence actual behavioral outcomes. In particular, limited attention has been given to the complex interplay between livestreaming room interactivity and the personal characteristics of anchors in shaping user behavior. Additionally, most existing studies rely on survey-based data, whereas empirical analyses utilizing real-time behavioral data in the LSE context remain scarce. Addressing these research gaps, this study proposes a conceptual framework of “interactivity–user retention–fan growth” and explores the boundary conditions under which interactivity affects user behavior. This approach not only enhances the current understanding of LSE mechanisms but also contributes to the development of more robust theoretical models and methodological tools in this domain.

2.3. Interactive Ritual Chain Theory

Interactive Ritual Chain Theory (IRCT), first proposed by Collins, posits that interaction is fundamentally ritualistic in nature and comprises four core elements: (1) physical co-presence, (2) boundaries that exclude outsiders, (3) a shared focus of attention, and (4) a common emotional experience and behavioral synchronization. Participants gather within a defined space to engage in emotionally charged symbolic activities, which generate or reinforce collective identity and emotional energy, thereby increasing their motivation to continue interacting with one another [44]. As a context-sensitive theory, IRCT has been widely applied in research on community relations and group dynamics. It elucidates how members of a group or community draw upon emotional energy to foster social interactions and cultivate a sense of belonging and shared identity [45,46,47]. For instance, in virtual environments such as online communities, interaction rituals frequently emerge, eliciting collective enthusiasm, reinforcing group identity, and generating emotional energy that helps sustain social bonds [48]. Recent theoretical advancements have extended IRCT into the realm of LSE. In the LSE context, users interact with anchors and fellow viewers in real time, forming virtual community ties that foster emotional resonance and a sense of belonging, ultimately shaping user behavior. Meng et al. [23] found that as users engage with livestream content and participate in real-time interactions, they form interactive ritual chains that generate emotional energy and foster a sense of cohesion. This emotional engagement, in turn, promotes sustained participation and prolonged viewership in livestreaming sessions.
Nevertheless, most of the existing research on IRCT remains at a theoretical and qualitative exploratory stage, with limited empirical studies validating its explanatory power. Within the context of LSE, users collectively form a virtual community with a sense of physical co-presence, naturally creating barriers that exclude non-participating outsiders. Additionally, anchors, products, and other livestream features serve as shared focal points of attention, establishing the conditions for interaction and fostering shared emotional experiences and synchronized behaviors. Building on these contextual features, this study investigates the effect of interactivity on fan growth through the lens of IRCT, thereby extending the empirical application of IRCT in the LSE domain.

2.4. Dual Systems Theory of Decision-Making

Dual Systems Theory (DST) delineates two cognitive systems engaged in human judgment and decision-making processes. The intuitive–heuristic system operates based on intuition, processes information rapidly, provides automatic responses, and typically considers only partial information during evaluation. This system is particularly vulnerable to various cognitive biases [49]. In contrast, the rational–analytical system functions through deliberate effort, processes information systematically, and employs algorithmic reasoning to solve problems. It operates more slowly, requires greater cognitive resources, and can effectively reduce decision-making biases [50]. While the heuristic system tends to rely on contextual cues and stereotypes—focusing more on outcomes than on the reasoning process—the analytical system emphasizes logical rules and maintains awareness of both reasoning processes and results [51].
As DST continues to evolve, it has been widely applied in consumer behavior research [52]. Recent studies have identified online user characteristics as important heuristic cues influencing consumer decision-making [28]. In the context of LSE, an anchor’s physical appearance—particularly their facial attractiveness—serves as one of the most salient and intuitive online features. It provides an immediate visual impression and acts as a powerful heuristic cue for viewers. The effect of facial attractiveness on consumer decision-making is well documented in fields such as behavioral economics [53,54]. However, in the relatively nascent domain of LSE research, it remains unclear whether anchors’ facial attractiveness, as a heuristic cue, significantly affects user behavior. In particular, the potential interaction between this easily perceived cue and livestreaming room characteristics has yet to be empirically explored. To address this gap, the present study incorporates anchors’ facial attractiveness into a moderation model, thereby enriching the theoretical framework of marketing management within the LSE context.

3. Research Hypothesis

3.1. Interactivity and Fan Growth

In this study, we argue that interactivity in LSE plays a critical role in driving fan growth. Prior research grounded in Interactive Ritual Chain Theory (IRCT) suggests that ritualized interactions frequently emerge in virtual communities, evoking collective euphoria, group identity, and emotional energy, thereby contributing to the stability and maintenance of social relationships [48]. In the context of LSE, the livestreaming room functions as a virtual community in which users engage in real-time interactions with both the anchor and other users. Such interactions foster quasi-social connections that generate emotional energy and cultivate a sense of belonging, which, in turn, influences user behavior. Meng et al. [23] found that when users actively engage with a livestream and interact with the anchor and other participants, they form chains of interactive rituals that produce emotional energy and a shared sense of togetherness. This emotional dynamic encourages users to repeatedly participate and maintain long-term viewing behavior. Moreover, real-time interaction contributes to a perceived sense of “virtual physical presence,” or spatial co-presence, within the livestreaming environment. This perceived presence enhances users’ feelings of transparency, engagement, trust, and enjoyment—factors that have been shown to increase user satisfaction and platform loyalty [15,34,40]. Hu and Chaudhry [55] further demonstrated that interactions between anchors and users often resemble quasi-social relationships, wherein both parties develop a sense of intimacy and view each other as “real friends.” This perceived intimacy helps reduce users’ uncertainty and mitigates psychological barriers [42], ultimately increasing the likelihood that users will become loyal fans.
In LSE, the frequency of interaction reflects the intensity of engagement between anchors and users. Higher interaction frequency can enhance users’ perceived role identity and foster stronger community ties [40]. In addition, frequent interactions allow users to obtain more accurate and detailed information about products and services, thereby elevating their perception of the value provided during the livestream [42]. Prior studies have shown that regular contact between sales representatives and customers enhances customer satisfaction and commercial performance [56]. Therefore, frequent interactions in LSE may contribute to the development of interpersonal emotional bonds, foster a sense of belonging, and thus promote fan growth. Furthermore, the quality of interaction captures the effectiveness of anchor-led engagements within the livestreaming room. High-quality interactions are more likely to induce immersive engagement and deepen the relationship between users and anchors [31,40]. Based on these insights, this study proposes the following hypotheses:
H1a. 
Interaction frequency significantly and positively affects fan growth.
H1b. 
Interaction quality significantly and positively affects fan growth.

3.2. The Mediating Role of User Retention

IRCT posits that an interaction ritual involves individuals jointly engaging in emotionally driven symbolic activities, which, in turn, generate or reinforce collective identity and emotional energy. This emotional energy strengthens individuals’ willingness to continue interacting, thereby sustaining quasi-social relationships and fostering positive emotional experiences characterized by enjoyment and pleasure [23]. Prior research further suggests that when interactions are perceived as informative, entertaining, and compelling, individuals are more inclined to engage actively, devote more time to communication, and ultimately enter an immersive state [27,57]. In support of this, Pelet et al. [58] demonstrated that immersive experiences significantly enhance visit frequency and browsing duration on websites. Building on these findings, this study argues that higher interaction frequency and superior interaction quality within a livestreaming room increase users’ willingness to remain engaged, deepen their interaction with the anchor, and focus more intently on the promoted products or services. Furthermore, prior research has confirmed that user retention plays a significant positive role in shaping consumers’ purchase intentions and behaviors across diverse digital contexts [59]. Based on this theoretical and empirical foundation, this study proposes the following hypotheses:
H2a. 
User retention mediates the relationship between interaction frequency and fan growth.
H2b. 
User retention mediates the relationship between interaction quality and fan growth.

3.3. The Moderating Role of Anchors’ Facial Attractiveness

Beauty influences perceptions and shapes judgments and decisions across a wide range of domains [53]. In traditional marketing, consumers often rely on observable online cues—such as avatars—as indicators of source credibility. These visual cues, alongside textual information, attract attention and help shape online impressions [60]. Attractive individuals are frequently perceived as possessing socially desirable personality traits and higher competence. This “beauty premium” has been widely documented across various fields, including the labor market, advertising, marketing, and online socialization [53,54]. In the context of LSE, an anchor’s facial attractiveness functions as an intuitive and salient heuristic cue that can dominate users’ information processing and significantly influence their viewing behavior. Anchors with higher facial attractiveness generally achieve higher levels of user retention in livestreaming sessions.
However, a substitution effect may exist between an anchor’s facial attractiveness and livestream interactivity. According to Dual Systems Theory (DST), people form judgments about social competence, age, health, and service quality based on facial cues, which contribute to more favorable first impressions [61]. For instance, a visually appealing personal photo can facilitate positive evaluations and increase customer attachment to the service provider [62]. Consequently, higher facial attractiveness may enhance perceived value and extend user retention time. Conversely, research has also acknowledged the “dark side” of attractiveness stereotypes [53], particularly in LSE environments, where interaction and social engagement are vital. When an anchor’s facial attractiveness is lower, users may interpret frequent and high-quality interactions as signals of competence and professionalism. These interactions can then compensate for lower attractiveness, foster immersive experiences, and more strongly influence user retention. Based on these insights, this study hypothesizes the following:
H3a. 
In a livestreaming room where the anchor’s facial attractiveness is higher, the promotion effect of interaction frequency on user retention will be significantly reduced.
H3b. 
In a livestreaming room where the anchor’s facial attractiveness is higher, the promotion effect of interaction quality on user retention will be significantly reduced.
Figure 1 shows the research model of this paper.

4. Research Design

4.1. Data Collection Platform

The data for this study were collected from Douyin, the largest and fastest-growing LSE platform in Mainland China (the domestic counterpart of TikTok) [38]. As of 2024, Douyin hosts over 700 million daily active users and offers a wide array of products, including fresh produce, apparel and accessories, daily necessities, and electronic goods. Due to its widespread popularity, Douyin’s LSE ecosystem has been the focus of extensive research in authoritative academic literature [8,16,34]. The platform’s highly interactive livestreaming rooms facilitate real-time communication between anchors and users, creating a dynamic environment that is well suited for the objectives of this study. Moreover, Douyin’s social features support the random sampling of livestreaming sessions through the “Live Square,” thereby enhancing the external validity of the research. The platform also provides an official data interface, allowing researchers to retrieve data programmatically using a Python-based API.

4.2. Data Collection Process and Processing

Data Collection: Data collection was conducted through a combination of crawler programs and manual efforts by research assistants. A total of twelve research assistants were recruited for this study. Each assistant was instructed to enter a randomly selected livestreaming room via the Douyin Live Square once every hour between 9:00 a.m. and 9:00 p.m. daily, from January to March 2023, excluding lunch breaks and the Chinese New Year period to avoid holiday-related anomalies. During each session, the assistants recorded a video that captured a complete product presentation delivered by the anchor. Simultaneously, they used a designated Python-based program to retrieve livestreaming room links and collect multiple types of data, including bullet comments, product information, anchor attributes, and user statistics. After completing each session, the research assistants downloaded and compiled additional structured data from a third-party analytics platform (www.huitun.com), which provided key performance metrics such as livestream sales, average user engagement time, broadcast duration, and viewer counts.
Data Processing: We processed unstructured video and text data. Specifically, video processing was carried out using OpenCV (4.12.0) (Open Source Computer Vision Library: an open-source computer vision and machine learning software library; it provides a wealth of image processing and computer vision algorithms that can be applied to a variety of fields, such as machine vision, autonomous driving, medical image processing, and so on) to capture keyframes (specific frames extracted from the video at equal time intervals or frame intervals that represent key information or change points in the video content), which were then uploaded to the Face++ (Face++ is an open AI platform that provides developers with AI capabilities such as face recognition, portrait processing, human body recognition, text recognition, and image recognition) image processing interface to analyze facial keypoints and assess the attractiveness of the anchor. Additionally, bullet comments’ text was analyzed for sentiment using the Baidu Intelligent Platform API. The raw dataset initially contained 1,020,5902 bullet comments from 1891 livestreaming rooms. However, after excluding records with missing data on host facial recognition, the final empirical dataset used for this study comprised 1472 livestreaming rooms. The complete data flow is shown in Figure 2.

4.3. Variable Measurement

Explanatory Variables: Following the framework of Hawkins et al. [43], we categorize interactivity into two dimensions: interaction frequency and interaction quality. Interaction frequency refers to the intensity of interactions between the anchor and users, as well as among the users themselves, within a specific timeframe in the livestreaming room. We use the number of bullet comments captured per minute by the crawler program as a proxy measure; a higher frequency of bullet comments per unit of time indicates increased interaction frequency. Interaction quality mainly refers to the quality and depth of user interactions within the livestreaming room. Compared to interaction frequency, interaction quality places more emphasis on the substance and effectiveness of user interactions. We measure interaction quality through sentiment analysis of bullet comments’ text. Specifically, we use the sentiment analysis API from Baidu’s open AI platform, which can assess the emotional tendency of text containing subjective information, effectively identifying the user emotions in each bullet comment and categorizing them as positive, neutral, or negative. We calculate the ratio of positive comments to total comments in the livestreaming room as a proxy measure of interaction quality.
Explained Variable: Fan growth is measured by the number of new fans gained after each livestream, reflecting the livestream’s appeal to new users. In addition, in the robustness test, we measure fan growth through the fan conversion rate in the livestreaming room.
Mediating Variable: User retention refers to the average duration of each user’s stay in the livestreaming room, serving as a key indicator of the content’s and merchandise’s attractiveness. It is calculated by dividing the total duration of user visits by the total number of visitors in the livestreaming room.
Moderating Variable: The anchor’s facial attractiveness is assessed using the Face++ facial attribute detection API. Face++ is a reputable provider of facial recognition and body detection services and has been used in multiple studies for tasks such as facial attractiveness, age classification, and emotion recognition [53]. We use the captured keyframes of the anchor’s face as inputs. Face++ provides two attractiveness scores for each image, ranging from 0 to 100, representing perspectives from both female and male users. We calculate the average of these scores to determine the anchor’s facial attractiveness, following the methodology of Ma et al. [53] and Peng et al. [61]
Control Variables: Based on existing research, other factors that may influence audience behavior are included in the model [63,64,65]. These factors include anchor characteristics (reputation, gender, age, speed), product characteristics (commodity, price), and livestreaming room characteristics (duration, type, recommendation, likes). To mitigate heteroskedasticity and reduce the impact of outliers, all continuous variables are log-transformed. The measurements of the control variables are shown in Table 1.

4.4. Model Setting

To test the research hypotheses, the following models are specified in this study:
F a n   g r o w t h i = β 0 + β 1 i n t e r a c t i o n   f r e q u e n c y i + γ C o n t r o l i + ε i
F a n   g r o w t h i = a 0 + a 1 i n t e r a c t i o n   q u a l i t y i + γ C o n t r o l i + ε i
where F a n   g r o w t h i is the dependent variable, while i n t e r a c t i o n   f r e q u e n c y i   and i n t e r a c t i o n   q u a l i t y i are the independent variables. Considering the cross-sectional nature of the data and potential issues of heteroskedasticity, we adopt the ordinary least squares (OLS) regression model with robust standard errors. OLS is a widely used method for estimating linear relationships between variables, and the use of robust standard errors helps correct potential heteroskedasticity that may bias the standard errors, leading to more reliable inference [8,38]. Given our large sample size and clearly defined independent variables, OLS with robust standard errors provides an appropriate baseline estimation strategy for evaluating the relationships among interactivity, user retention, and fan growth.
To account for potential endogeneity and omitted-variable bias, particularly concerning selection into high- or low-interactivity groups, PSM is employed as a robustness check. A binary treatment indicator is constructed, assigning treatment status to livestreaming rooms with above-mean interactivity (frequency and quality). The propensity scores are estimated using a logit model based on pre-treatment covariates (e.g., duration, price, commodity, type, recommendation, likes). Matching is performed using 1-to-1 nearest-neighbor matching without replacement to ensure high match quality. Meanwhile, sensitivity analysis is employed in this study to address concerns related to potential endogeneity, particularly from omitted-variable bias. In LSE contexts, unobserved factors—such as anchor charisma, real-time marketing efforts, or backstage operational support—might simultaneously influence both interactivity and fan growth, leading to biased estimates. This method allows us to quantify how strong an unobserved confounder would need to be, relative to the observed covariates, in order to overturn our results [66].

5. Empirical Analysis

5.1. Descriptive Statistics and Correlation Analysis

Table 2 and Table 3 present the results of the descriptive statistics and correlation analysis, respectively. The data in Table 2 illustrate significant variability in fan attraction capabilities across different livestreaming rooms, with fan growth ranging from 0 to 12.145 and a standard deviation of 6.82. The maximum values for interaction frequency and interaction quality are 3.508 and 1, respectively, while their minimum values are as low as 0.082 and 0, respectively, highlighting the diverse nature of interactions in livestreaming. The descriptive statistics for the other variables are consistent with findings from existing studies.

5.2. Regression Analysis

5.2.1. Baseline Regression Results

The regression results for the main effects are displayed in Table 4. Models 1 and 2 reveal that the coefficients for interaction frequency and interaction quality are 0.475 and 0.235, respectively, both of which are significantly positive. This indicates that both interaction frequency and interaction quality positively influence fan growth in the livestreaming room. In other words, higher interaction frequency and better interaction quality lead to greater fan growth, thus confirming H1a and H1b. In Models 3 and 4, the coefficients for interaction frequency and quality on user retention are 0.164 and 0.963, respectively, which are also significantly positive. This demonstrates that both higher interaction frequency and better interaction quality positively affect user retention in the livestreaming room, indicating that users tend to stay longer.

5.2.2. Regression Results of Mediating Effect and Moderating Effect

The regression results for the mediating and moderating effects are presented in Table 5. Models 5 and 6 indicate that even after including the mediating variable (user retention), the coefficients of interaction frequency and quality on fan growth remain significantly positive, although they are reduced from 0.475 to 0.443 and from 0.235 to 0.224, respectively. The coefficients of user retention on fan growth are also significantly positive, confirming that user retention partially mediates the relationship between livestream interactions and fan growth, thus verifying H2a and H2b.
Models 7 and 8 explore the moderating effects of the anchor’s facial attractiveness. The interaction terms of interaction frequency and quality with the anchor’s facial attractiveness yield coefficients of −2.668 and −0.778, respectively, both of which are significantly negative. This suggests that the anchor’s facial attractiveness negatively moderates the positive impacts of interaction frequency and quality on fan growth. Specifically, livestreaming rooms with less attractive anchors experience a more pronounced positive effect from interaction frequency and quality on fan growth compared to those with more attractive anchors. This result implies that the effects of the anchor’s facial attractiveness and interactivity are interrelated and compensatory, thus verifying H3a and H3b.

5.3. Robustness Test

We tested the robustness of the results using several methods. First, we replaced the regression model. Since fan growth is truncated count data, we used the Tobit model for the main effect regression. The results are shown in Model 9 and Model 10 in Table 6, and they remain consistent with the previous findings. Second, we replaced the dependent variable with the fan conversion rate, which is calculated as the ratio of the number of new fans acquired during the LSE session to the total number of viewers. The regression results are shown in Model 11 and Model 12 in Table 6, and there were no significant changes in the results. Finally, to eliminate the impact of extreme values, we applied a 1% trimming to all continuous variables. The results are shown in Table 7, and all findings are consistent with the previous ones, indicating that the research conclusions are robust.

5.4. Endogeneity Test

5.4.1. Propensity Score Matching (PSM)

Considering the potential endogeneity issues arising from selection bias, we used propensity score matching (PSM) to simulate a randomized experiment, allowing for a more accurate test of the causal effects of interaction frequency and interaction quality on fan growth. Based on the average interaction frequency and interaction quality in the sample, we divided the livestream rooms into a high-interaction-frequency and high-interaction-quality group (treatment group) and a low-interaction-frequency and low-interaction-quality group (control group). We also used all of the control variables in the model as matching variables and applied 1:1 caliper matching to calculate propensity scores. The distribution of characteristics before and after matching for the treatment and control groups is shown in Figure 3. Model 21 and Model 22 in Table 8 present the regression results for fan growth after matching. The results indicate that the results of this study are not affected by selection bias.

5.4.2. Sensitivity Analysis

We considered potential endogeneity issues that may arise from omitted variables and conducted a sensitivity analysis. This methodology attempts to demonstrate that the direction of the study’s results is not affected by omitted variables when the strength of these omitted variables is less than some multiple of the core control variables [66]. We used the duration of the livestreaming as a comparison variable. As shown in Figure 4, the direction of the coefficients (a) and the T-value for interaction frequency (b) did not change when the omitted variable explained less than or equal to three times the strength of the comparison variable. The results for the interaction quality also remained consistent. This suggests that omitted variables are unlikely to affect this study’s results.

6. Discussion

6.1. Key Findings

This study found that both interaction frequency and interaction quality in LSE significantly affect fan growth. This finding supports previous research, which demonstrated that perceived interactivity stimulates users’ intentions to engage in positive behaviors in online communities [67]. Additionally, these results align with IRCT, which suggests that real-time interactions in livestreaming rooms create a virtual physical presence [23]. Through repeated interaction with anchors and other users, an interactive ritual chain is formed, fostering a sense of belonging and emotional solidarity. This emotional connection enhances users’ perceived transparency, trust, and engagement, further promoting fan conversion [31,56].
Secondly, we discovered that user retention mediates the effect of interactivity on fan growth. This suggests that improving interactivity alone is not sufficient; sustaining viewer engagement over time is the key channel through which interactivity translates into fan acquisition. This aligns with prior studies showing that interactivity enhances emotional energy and collective identity in online communities, leading to longer user retention [68]. In LSE, this extended retention time increases exposure to the content and atmosphere of the livestreaming room, thereby increasing the probability of fan conversion. This mediating effect highlights the importance of designing not only interactive but also engaging and emotionally resonant livestreaming experiences to foster long-term audience attachment.
Thirdly, the results show that the anchor’s facial attractiveness negatively moderates the effect of interactivity on user retention, indicating a substitution effect. While previous studies suggested that facial attractiveness as a heuristic cue could reinforce other persuasive signals (e.g., verbal content), our findings suggest that, in the context of LSE, attractive facial features may reduce the marginal benefits of interactive behavior. One possible explanation for this is that highly attractive anchors may trigger passive consumption patterns, where users are drawn to appearances but less motivated to engage or remain. Conversely, when facial attractiveness is low, interactive behavior becomes a more salient indicator of anchor competence, increasing perceived authenticity and emotional connection. Additionally, our findings may reflect the complexity and saturation of multimodal cues in livestreaming settings. When too many signals (e.g., beauty, interaction) are present simultaneously, cognitive overload or distraction may occur [38], leading to diminishing returns for interaction strategies.

6.2. Theoretical Contributions

This study offers several theoretical contributions: (1) Broadening the scope of LSE performance research: Previous studies have primarily focused on short-term performance metrics such as livestreaming sales, conversion rates, and gift-giving [69,70,71] but have overlooked fan growth as a long-term performance indicator. Our research demonstrates that interactivity has a significant impact on fan growth, thereby broadening the scope of LSE performance research and providing a new perspective on the long-term sustainability of LSE. (2) Revealing the mechanism of interactivity’s impact on fan growth in LSE: Although previous research has explored the impact of interactivity on social performance, these studies have primarily focused on short-term user behaviors such as immediate engagement and interaction frequency [34], and the underlying mechanisms have not been fully clarified. Our findings show that interactivity promotes fan growth by enhancing user retention, providing a new theoretical framework for studying the relationship between interactivity and LSE performance. (3) Incorporating the anchor’s facial attractiveness into the LSE research framework: Existing research on anchors’ facial features mostly focuses on aspects such as emotional fluctuations [72,73], neglecting facial attractiveness as an important digital marketing cue. This study is one of the first to explore the moderating role of anchors’ facial attractiveness in the process of interactivity influencing fan growth, clarifying the substitution effect between interactivity and anchor characteristics (such as facial attractiveness) on user retention, and providing a new theoretical perspective for the study of multiple signal interactions in LSE. (4) Validating the applicability of IRCT and DST in the LSE context: Existing research on IRCT and DST has mostly focused on traditional business or other social fields [74,75]. This study establishes the relationships between interactivity, user retention, anchors’ facial attractiveness, and fan growth, validating the applicability of IRCT and DST in the emerging context of LSE, enriching the theoretical foundation of LSE, and providing new insights for future research.

6.3. Practical Implications

The findings of this study have significant practical implications for livestreaming platforms, marketing managers, and anchors, especially in terms of enhancing fan growth, user retention, and livestreaming room marketing performance. The specific implications are as follows: (1) The role of interactivity in promoting fan growth: The results indicate that both the frequency and quality of livestreaming interactivity have a significant positive impact on fan growth. Therefore, marketing managers and anchors should fully recognize the key role that interactivity plays in driving fan growth. To maximize the effectiveness of interactivity, it is essential to provide anchors with systematic training in interaction skills, helping them improve the quality of their interactions, including how to better engage with the audience, answer questions, and guide discussions. Furthermore, anchors are encouraged to provide professional explanations and thematic guidance during livestreams, creating a positive and harmonious interactive atmosphere that fosters more active bullet comments, enhancing users’ sense of participation and belonging. (2) User retention as a mediating factor: This study found that user retention plays a mediating role in the impact of interactivity on fan growth. Therefore, livestreaming platforms can promote fan growth by improving user retention. To enhance user retention, platforms should innovate interactive methods, such as incorporating gamification elements, reward-based Q&A, interactive challenges, etc., to increase audience engagement and stickiness. Additionally, platforms should optimize the interactive experience by setting up more interactive sessions, such as raffles and Q&A segments, to improve the frequency and quality of audience participation during the livestream. This will help extend user retention time, further promoting fan growth. (3) The substitution effect between the anchor’s personal characteristics and interactivity: This study confirms the potential substitution effect between an anchor’s facial attractiveness and interactivity on user retention. Specifically, anchors with higher facial attractiveness can compensate for lower interactivity by leveraging the “facial attractiveness premium,” which helps mitigate the decline in user retention caused by lower interactivity. Based on this, we recommend that when selecting anchors, especially for newly established livestreaming rooms with fewer bullet comments and lower comment activity, facial attractiveness and other personal characteristics should be included as selection criteria. By choosing attractive anchors, platforms can compensate for retention issues caused by insufficient interactivity. Additionally, for livestreaming rooms with lower interactivity, selecting anchors with higher facial attractiveness can enhance user retention, while supplementing other interactive strategies to improve the room’s marketing performance.

6.4. Limitations and Prospects

This study has several limitations: (1) The empirical data were collected exclusively from Douyin, China’s leading LSE platform. Due to differences in culture, economic systems, and platform functionality, our findings may not be directly generalizable to international platforms such as TikTok, YouTube Live, or Amazon Live. Future studies should consider cross-country or cross-platform comparisons to provide more comprehensive and generalizable insights. (2) Our analysis did not incorporate user-level demographic information (e.g., age, gender, income, occupation) due to data access restrictions and ethical concerns related to user privacy. While this omission is consistent with practices in prior research [61,76], individual user characteristics may influence fan acquisition dynamics. Future research could enrich the model by integrating user-level attributes through ethically approved and privacy-compliant data collection strategies. (3) This study focused primarily on visual (e.g., facial attractiveness) and textual (e.g., bullet comments) signals in the livestreaming environment. However, other important multimodal cues, such as tone of voice, background music, and color schemes, were not examined. Future research could adopt a more holistic approach by incorporating these non-verbal and ambient factors to better capture the full spectrum of user engagement mechanisms. (4) The data collection period excluded the Chinese New Year holiday to avoid confounding results with abnormal seasonal patterns. However, this also limits the scope of the analysis, as holidays often represent peak user activity and unique behavioral patterns. Future studies should consider including holiday data to examine how temporal contexts affect interactivity and fan growth. (5) This study involved the use of large-scale behavioral and visual data, including facial recognition techniques to assess anchors’ attractiveness. While data were collected from publicly accessible sources and no personally identifiable information was used, we acknowledge the growing importance of aligning with data ethics and platform policies. Future research should further address issues of informed consent and data governance in the context of large-scale social media analytics.

7. Conclusions

Increasing the number of fans in LSE through marketing has become a crucial topic in the field of management. This study explored the determinants of fan growth in LSE from the perspectives of interactivity and user retention, while also considering the complex interaction effect between anchor characteristics and interactivity on fan growth. Using a real-time, objective dataset from Douyin, we validated the theoretical model. Empirical analyses showed that interactivity significantly and positively affects fan growth by enhancing user retention. Additionally, we found that interactivity and an anchor’s facial attractiveness have a substitution effect on fan growth, suggesting a substitution relationship between anchor characteristics and livestreaming room characteristics. These findings offer both practical and theoretical insights for LSE operations managers and future researchers.

Author Contributions

Conceptualization, H.Y. and B.W.; methodology, H.Y.; software, H.Y.; formal analysis, H.Y. and B.W.; investigation, H.Y. and B.W.; data curation, H.Y. and B.W.; writing—original draft preparation, H.Y.; writing—review and editing, B.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest

The authors declare no conflict of interest.

Abbreviations

AbbreviationDescription
LSELivestreaming e-commerce.
IRCTInteractive Ritual Chain Theory.
DSTDual Systems Theory.
DouyinOne of China’s largest livestreaming e-commerce and social platforms.
Face++Face++ is an open AI platform that provides developers with AI capabilities such as face recognition, portrait processing, human body recognition, text recognition, and image recognition.
OpenCVOpen Source Computer Vision Library: an open source computer vision and machine learning software library; it provides a wealth of image processing and computer vision algorithms that can be applied to a variety of fields, such as machine vision, autonomous driving, medical image processing, and so on.

References

  1. Feng, N.; Liu, X.; Wu, Y.; Jia, C.; Dong, C. Exploring Message Sidedness on Purchase Intention in Live Streaming E-Commerce: The Moderation of Live Streamer Type. Ind. Manag. Data Syst. 2025. [Google Scholar] [CrossRef]
  2. Zhang, W.; Xie, R.; Quan, P.; Ma, Z. Product Return Prediction in Live Streaming E-Commerce with Cross-Modal Contrastive Transformer. Decis. Support Syst. 2025, 194, 114470. [Google Scholar] [CrossRef]
  3. Hao, S. What Inhibits Live-Streaming Commerce from Winning Consumers? An Investigation from the Stressor-Strain-Outcome Perspective. Electron. Mark. 2025, 35, 38. [Google Scholar] [CrossRef]
  4. Zhao, J.; Zhou, J.; Wu, P.; Liang, K. Boosting E-Commerce Sales with Live Streaming: The Power of Barrages. Electron. Commer. Res. 2024, 1–38. [Google Scholar] [CrossRef]
  5. Guo, Y.; Zhang, Y.; Goh, K.-Y.; Peng, X. Can Social Technologies Drive Purchases in E-Commerce Live Streaming? An Empirical Study of Broadcasters’ Cognitive and Affective Social Call-to-Actions. Prod. Oper. Manag. 2024, 10591478241276131. [Google Scholar] [CrossRef]
  6. Wang, D.; Luo, X.; Hua, Y.; Benitez, J. Big Arena, Small Potatoes: A Mixed-Methods Investigation of Atmospheric Cues in Live-Streaming e-Commerce. Decis. Support Syst. 2022, 158, 113801. [Google Scholar] [CrossRef]
  7. Wongkitrungrueng, A.; Assarut, N. The Role of Live Streaming in Building Consumer Trust and Engagement with Social Commerce Sellers. J. Bus. Res. 2020, 117, 543–556. [Google Scholar] [CrossRef]
  8. Yang, H.; Xu, X.; Wang, B. The Power of Body Language: Understanding Anchor’s Body Language and the Success of Live-Streaming e-Commerce. Nankai Bus. Rev. Int. 2025, 16, 332–358. [Google Scholar] [CrossRef]
  9. Li, W.; Cujilema, S.; Hu, L.; Xie, G. How Social Scene Characteristics Affect Customers’ Purchase Intention: The Role of Trust and Privacy Concerns in Live Streaming Commerce. J. Theor. Appl. Electron. Commer. Res. 2025, 20, 85. [Google Scholar] [CrossRef]
  10. Xu, X.; Zeng, Q.; Na, R.; Fan, W. Unveiling the Influence of Streamer Characteristics on Sales Performance in Live Streaming Commerce. Electron. Commer. Res. Appl. 2025, 72, 101510. [Google Scholar] [CrossRef]
  11. Zhang, X.; Zhang, S. Investigating Impulse Purchases in Live Streaming E-Commerce: A Perspective of Match-Ups. Technol. Forecast. Soc. Change 2024, 205, 123513. [Google Scholar] [CrossRef]
  12. Peng, L.; Zhang, N.; Huang, L. How the Source Dynamism of Streamers Affects Purchase Intention in Live Streaming E-Commerce: Considering the Moderating Effect of Chinese Consumers’ Gender. J. Retail. Consum. Serv. 2024, 81, 103949. [Google Scholar] [CrossRef]
  13. Cuong, D.T. Using Theory of Interpersonal Behavior to Consumers’ Online Impulse Buying Behavior in Social Commerce Live-Streaming: Perceived Price’s Moderating Role. J. Mark. Anal. 2024, 1–14. [Google Scholar] [CrossRef]
  14. Chen, Y.; Li, M.; Chen, A.; Lu, Y. Trust Development in Live Streaming Commerce: Interaction-Based Building Mechanisms and Trust Transfer Perspective. Ind. Manag. Data Syst. 2024, 124, 3218–3239. [Google Scholar] [CrossRef]
  15. Fu, J.-R.; Hsu, C.-W. Live-Streaming Shopping: The Impacts of Para-Social Interaction and Local Presence on Impulse Buying through Shopping Value. Ind. Manag. Data Syst. 2023, 123, 1861–1886. [Google Scholar] [CrossRef]
  16. Lang, X.; Lin, S.; Ma, X.; Li, T. Can Comments and Dialogues Make Sense? The Effect of Two-Way Interactions on Sales and Followers in Live Streaming Commerce. Decis. Support Syst. 2025, 193, 114451. [Google Scholar] [CrossRef]
  17. Zhao, K.; Hu, Y.; Hong, Y.; Westland, J.C. Understanding Characteristics of Popular Streamers on Live Streaming Platforms: Evidence from Twitch.Tv. J. Assoc. Inf. Syst. 2021, 22, 1076–1098. [Google Scholar] [CrossRef]
  18. Guo, Y.; Zhang, K.; Wang, C. Way to Success: Understanding Top Streamer’s Popularity and Influence from the Perspective of Source Characteristics. J. Retail. Consum. Serv. 2022, 64, 102786. [Google Scholar] [CrossRef]
  19. Sun, Y.; Bao, Z. Live Streaming Commerce: A Compulsive Buying Perspective. Manag. Decis. 2023, 61, 3278–3294. [Google Scholar] [CrossRef]
  20. Li, Y.; Li, X.; Cai, J. How Attachment Affects User Stickiness on Live Streaming Platforms: A Socio-Technical Approach Perspective. J. Retail. Consum. Serv. 2021, 60, 102478. [Google Scholar] [CrossRef]
  21. Wu, D.; Wang, X.; Ye, H.J. Transparentizing the “Black Box” of Live Streaming: Impacts of Live Interactivity on Viewers’ Experience and Purchase. IEEE Trans. Eng. Manag. 2024, 71, 3820–3831. [Google Scholar] [CrossRef]
  22. Guan, Z.; Hou, F.; Li, B.; Phang, C.W.; Chong, A.Y. What Influences the Purchase of Virtual Gifts in Live Streaming in China? A Cultural Context-sensitive Model. Inf. Syst. J. 2022, 32, 653–689. [Google Scholar] [CrossRef]
  23. Meng, L.; Zhao, Y.; Jiang, Y.; Bie, Y.; Li, J. Understanding Interaction Rituals: The Impact of Interaction Ritual Chains of the Live Broadcast on People’s Wellbeing. Front. Psychol. 2022, 13, 1041059. [Google Scholar] [CrossRef] [PubMed]
  24. Kang, K.; Lu, J.; Guo, L.; Li, W. The Dynamic Effect of Interactivity on Customer Engagement Behavior through Tie Strength: Evidence from Live Streaming Commerce Platforms. Int. J. Inf. Manag. 2021, 56, 102251. [Google Scholar] [CrossRef]
  25. Nitzan, I.; Libai, B. Social Effects on Customer Retention. J. Mark. 2011, 75, 24–38. [Google Scholar] [CrossRef]
  26. Yang, J.; Zeng, Y.; Liu, X.; Li, Z. Nudging Interactive Cocreation Behaviors in Live-Streaming Travel Commerce: The Visualization of Real-Time Danmaku. J. Hosp. Tour. Manag. 2022, 52, 184–197. [Google Scholar] [CrossRef]
  27. Bonner, J.M. Customer Interactivity and New Product Performance: Moderating Effects of Product Newness and Product Embeddedness. Ind. Mark. Manag. 2010, 39, 485–492. [Google Scholar] [CrossRef]
  28. Son, J.; Lee, J.; Oh, O.; Lee, H.K.; Woo, J. Using a Heuristic-Systematic Model to Assess the Twitter User Profile’s Impact on Disaster Tweet Credibility. Int. J. Inf. Manag. 2020, 54, 102176. [Google Scholar] [CrossRef]
  29. Xu, Q. Should I Trust Him? The Effects of Reviewer Profile Characteristics on eWOM Credibility. Comput. Hum. Behav. 2014, 33, 136–144. [Google Scholar] [CrossRef]
  30. Kanazawa, S.; Still, M.C. Is There Really a Beauty Premium or an Ugliness Penalty on Earnings? J. Bus. Psychol. 2018, 33, 249–262. [Google Scholar] [CrossRef]
  31. Zhang, M.; Liu, Y.; Wang, Y.; Zhao, L. How to Retain Customers: Understanding the Role of Trust in Live Streaming Commerce with a Socio-Technical Perspective. Comput. Hum. Behav. 2022, 127, 107052. [Google Scholar] [CrossRef]
  32. Lu, B.; Chen, Z. Live Streaming Commerce and Consumers’ Purchase Intention: An Uncertainty Reduction Perspective. Inf. Manag. 2021, 58, 103509. [Google Scholar] [CrossRef]
  33. Chen, H.; Dou, Y.; Xiao, Y. Understanding the Role of Live Streamers in Live-Streaming e-Commerce. Electron. Commer. Res. Appl. 2023, 59, 101266. [Google Scholar] [CrossRef]
  34. Yang, Q.; Huo, J.; Li, H.; Xi, Y.; Liu, Y. Can Social Interaction-Oriented Content Trigger Viewers’ Purchasing and Gift-Giving Behaviors? Evidence from Live-Streaming Commerce. Internet Res. 2023, 33, 46–71. [Google Scholar] [CrossRef]
  35. Xiao, L.; Lin, X.; Mi, C.; Akter, S. The Effect of Dynamic Information Cues on Sales Performance in Live Streaming E-Commerce: An IFT and ELM Perspective. Electron. Commer. Res. 2023, 1–30. [Google Scholar] [CrossRef]
  36. Xu, W.; Cao, Y.; Chen, R. A Multimodal Analytics Framework for Product Sales Prediction with the Reputation of Anchors in Live Streaming E-Commerce. Decis. Support Syst. 2024, 177, 114104. [Google Scholar] [CrossRef]
  37. Zhang, Y.; Sun, Y.; Wang, N. Understanding the Determinants of Gifting and Social Sharing in Live Streaming: A Dual-Attachment Perspective. Ind. Manag. Data Syst. 2024, 124, 564–582. [Google Scholar] [CrossRef]
  38. Wang, Z.; Luo, C.; Luo, X.; Xu, X. Understanding the Effect of Group Emotions on Consumer Instant Order Cancellation Behavior in Livestreaming E-Commerce: Empirical Evidence from TikTok. Decis. Support Syst. 2024, 179, 114147. [Google Scholar] [CrossRef]
  39. Tan, S. How to Interact with Consumers to Enhance Their Purchase Intention? Evidence from China’s Agricultural Products Live Streaming Commerce. Br. Food J. 2024, 126, 2500–2521. [Google Scholar] [CrossRef]
  40. Zhang, W.; Wang, Y.; Zhang, T.; Chu, J. Live-Streaming Community Interaction Effects on Travel Intention: The Mediation Role of Sense of Community and Swift-Guanxi. Inf. Technol. Tour. 2022, 24, 485–509. [Google Scholar] [CrossRef]
  41. Vonkeman, C.; Verhagen, T.; Van Dolen, W. Role of Local Presence in Online Impulse Buying. Inf. Manag. 2017, 54, 1038–1048. [Google Scholar] [CrossRef]
  42. Qian, T.Y.; Seifried, C. Virtual Interactions and Sports Viewing on Social Live Streaming Platforms: The Role of Co-Creation Experiences, Platform Involvement, and Follow Status. J. Bus. Res. 2023, 162, 113884. [Google Scholar] [CrossRef]
  43. Hawkins, A.; Graham, C.R.; Sudweeks, R.R.; Barbour, M.K. Academic Performance, Course Completion Rates, and Student Perception of the Quality and Frequency of Interaction in a Virtual High School. Distance Educ. 2013, 34, 64–83. [Google Scholar] [CrossRef]
  44. Münch, R. Interaction Ritual Chains; Collins, R., Ed.; Princeton University Press: Princeton, NJ, USA, 2004; Volume 110, pp. 1531–1533. [Google Scholar] [CrossRef]
  45. Xiang, K.; Zhang, H.; Qiao, G. How to Create Interactive Rituals: Key Factors and Strategic Configurations for Host-Guest Interaction in Boutique Hotels. J. Hosp. Tour. Manag. 2022, 52, 428–438. [Google Scholar] [CrossRef]
  46. Wong, I.A.; Lin, X.; Lin, Z.; Lin, Y. Gaining Resilience through Transformative Services in Cause-Related Events: An Interaction Ritual Chain Perspective. Int. J. Contemp. Hosp. Manag. 2023, 35, 3235–3262. [Google Scholar] [CrossRef]
  47. Ferguson, T.W. Whose Bodies? Bringing Gender Into Interaction Ritual Chain Theory. Sociol. Relig. 2020, 81, 247–271. [Google Scholar] [CrossRef]
  48. Maloney, P. Online Networks and Emotional Energy: How pro-Anorexic Websites Use Interaction Ritual Chains to (Re)Form Identity. Inf. Commun. Soc. 2013, 16, 105–124. [Google Scholar] [CrossRef]
  49. Hofmann, W.; Friese, M.; Strack, F. Impulse and Self-Control From a Dual-Systems Perspective. Perspect. Psychol. Sci. 2009, 4, 162–176. [Google Scholar] [CrossRef]
  50. Evans, J.S.B.T. Dual-Processing Accounts of Reasoning, Judgment, and Social Cognition. Annu. Rev. Psychol. 2008, 59, 255–278. [Google Scholar] [CrossRef]
  51. Chen, Q.; Gong, Y.; Lu, Y.; Chau, P.Y.K. How Mindfulness Decreases Cyberloafing at Work: A Dual-System Theory Perspective. Eur. J. Inf. Syst. 2023, 32, 841–857. [Google Scholar] [CrossRef]
  52. Turel, O.; Qahri-Saremi, H. Problematic Use of Social Networking Sites: Antecedents and Consequence from a Dual-System Theory Perspective. J. Manag. Inf. Syst. 2016, 33, 1087–1116. [Google Scholar] [CrossRef]
  53. Ma, S.; Ling, X.; Zhang, C.; Li, Y. Does Physical Attractiveness Facilitate Customer Citizenship Behaviors? Cross-Cultural Evidence from the Peer-to-Peer Economy. Int. J. Inf. Manag. 2023, 70, 102565. [Google Scholar] [CrossRef]
  54. Ahmed, S.; Ranta, M.; Vähämaa, E.; Vähämaa, S. Facial Attractiveness and CEO Compensation: Evidence from the Banking Industry. J. Econ. Bus. 2023, 123, 106095. [Google Scholar] [CrossRef]
  55. Hu, M.; Chaudhry, S.S. Enhancing Consumer Engagement in E-Commerce Live Streaming via Relational Bonds. Internet Res. 2020, 30, 1019–1041. [Google Scholar] [CrossRef]
  56. Zhou, G.; Liu, Y.; Hu, J.; Cao, X. The Effect of Tourist-to-Tourist Interaction on Tourists’ Behavior: The Mediating Effects of Positive Emotions and Memorable Tourism Experiences. J. Hosp. Tour. Manag. 2023, 55, 161–168. [Google Scholar] [CrossRef]
  57. Hawkins, M.A.; Saleem, F.Z. Influencer Advertising: Facilitating Poor-Fitting Influencer Posts. Manag. Decis. 2024, 62, 200–218. [Google Scholar] [CrossRef]
  58. Pelet, J.-É.; Ettis, S.; Cowart, K. Optimal Experience of Flow Enhanced by Telepresence: Evidence from Social Media Use. Inf. Manage. 2017, 54, 115–128. [Google Scholar] [CrossRef]
  59. Chung, B.D.; Park, J.H.; Koh, Y.J.; Lee, S. User Satisfaction and Retention of Mobile Telecommunications Services in Korea. Int. J. Hum. Comput. Interact. 2016, 32, 532–543. [Google Scholar] [CrossRef]
  60. Pang, H.; Ge, Y.; He, J.; Zhang, L. Eye-Tracking Study of the Celebrity Effect on Microblogging Browsing: An Example from Sina Microblog. Behav. Inf. Technol. 2021, 40, 1564–1578. [Google Scholar] [CrossRef]
  61. Peng, L.; Cui, G.; Chung, Y.; Zheng, W. The Faces of Success: Beauty and Ugliness Premiums in e-Commerce Platforms. J. Mark. 2020, 84, 67–85. [Google Scholar] [CrossRef]
  62. Wan, L.C.; Wyer, R.S. Consumer Reactions to Attractive Service Providers: Approach or Avoid? J. Consum. Res. 2015, 42, 578–595. [Google Scholar] [CrossRef]
  63. Liu, L.; Fang, J.; Yang, L.; Han, L.; Hossin, M.A.; Wen, C. The Power of Talk: Exploring the Effects of Streamers’ Linguistic Styles on Sales Performance in B2B Livestreaming Commerce. Inf. Process. Manag. 2023, 60, 103259. [Google Scholar] [CrossRef]
  64. Zhang, Z.; Zhang, Z.; Chen, P.-Y. Early Bird Versus Late Owl: An Empirical Investigation of Individual Shopping Time Habits and Its Effects. MIS Q. 2021, 45, 117–162. [Google Scholar] [CrossRef]
  65. Gao, W.; Jiang, N.; Gu, F. Understanding the Role of Streamers in Livestreaming Commerce: A Vocal–Visual Perspective. J. Manag. Anal. 2023, 10, 247–269. [Google Scholar] [CrossRef]
  66. Cinelli, C.; Hazlett, C. Making Sense of Sensitivity: Extending Omitted Variable Bias. J. R. Stat. Soc. Ser. B Stat. Methodol. 2020, 82, 39–67. [Google Scholar] [CrossRef]
  67. Chia-An Tsai, J.; Kang, T.-C. Reciprocal Intention in Knowledge Seeking: Examining Social Exchange Theory in an Online Professional Community. Int. J. Inf. Manag. 2019, 48, 161–174. [Google Scholar] [CrossRef]
  68. Zhang, L.; Liu, X. Interactivity and Live-Streaming Commerce Purchase Intention: Social Presence as a Mediator. Soc. Behav. Personal. Int. J. 2023, 51, 1–7. [Google Scholar] [CrossRef]
  69. Lin, S.-C.; Lee, Y.-Y. Explaining the Gift-Giving Intentions of Live-Streaming Audiences through Social Presence: The Perspective of Interactive Marketing. J. Res. Interact. Mark. 2024, 18, 945–969. [Google Scholar] [CrossRef]
  70. Luo, X.; Lim, W.M.; Cheah, J.-H.; Lim, X.-J.; Dwivedi, Y.K. Live Streaming Commerce: A Review and Research Agenda. J. Comput. Inf. Syst. 2025, 65, 376–399. [Google Scholar] [CrossRef]
  71. Lyu, W.; Qi, Y.; Liu, J. Proliferation in Live Streaming Commerce, and Key Opinion Leader Selection. Electron. Commer. Res. 2024, 24, 1153–1186. [Google Scholar] [CrossRef]
  72. Xi, D.; Xu, W.; Tang, L.; Han, B. The Impact of Streamer Emotions on Viewer Gifting Behavior: Evidence from Entertainment Live Streaming. Internet Res. 2024, 34, 748–783. [Google Scholar] [CrossRef]
  73. Du, P.; Huang, Z. Happiness Backfires: Emotion and Sales in Live Streaming. Electron. Commer. Res. 2023, 1–32. [Google Scholar] [CrossRef]
  74. Ming, X.; Bai, X.; Fu, J.; Yang, J. Decreasing Workplace Unethical Behavior Through Mindfulness: A Study Based on the Dual-System Theory of Ethical Decision-Making. J. Bus. Ethics 2025, 196, 189–207. [Google Scholar] [CrossRef]
  75. Yang, L.; Yuan, X.; Yang, X. Study of the Influencing Mechanism of User Interaction Behavior of Short Video E-Commerce Live-Streaming from the Perspective of SOR Theory and Interactive Ritual Chains. Curr. Psychol. 2024, 43, 28403–28415. [Google Scholar] [CrossRef]
  76. Xu, W.; Zhang, X.; Chen, R.; Yang, Z. How Do You Say It Matters? A Multimodal Analytics Framework for Product Return Prediction in Live Streaming e-Commerce. Decis. Support Syst. 2023, 172, 113984. [Google Scholar] [CrossRef]
Figure 1. Research model.
Figure 1. Research model.
Jtaer 20 00203 g001
Figure 2. Complete data processing flow.
Figure 2. Complete data processing flow.
Jtaer 20 00203 g002
Figure 3. Density map of PSM.
Figure 3. Density map of PSM.
Jtaer 20 00203 g003
Figure 4. Results of sensitivity analysis.
Figure 4. Results of sensitivity analysis.
Jtaer 20 00203 g004
Table 1. Variable descriptions.
Table 1. Variable descriptions.
VariableDescriptionReferences
ReputationThe anchor’s electronic word-of-mouth score (0–5)[63,64,65]
GenderThe gender of the anchor (1 for female and 0 for male)
AgeThe age of the anchor
SpeedThe number of Chinese words spoken by the anchor every minute
CommodityThe number of products sold in the LSE room
PriceThe average price of products sold in the LSE room
DurationThe LSE section duration (minutes)
TypeWith rooms hosted by independent anchors marked as 1 and those hosted by brand-affiliated anchors marked as 0
RecommendationPercentage of viewers entering the LSE room through the recommendations of Douyin video (0–1)
LikesThe number of likes in the LSE room
Table 2. Descriptive statistics.
Table 2. Descriptive statistics.
VariableObsMeanStd. Dev.MinMax
Fan growth14726.821.964012.145
Interaction quality14720.6660.15801
Interaction frequency14723.5081.8630.08210.375
Duration14725.3990.733.0868.371
Reputation14724.7730.2333.025
Commodity14723.690.90.6936.648
Price14724.091.0630.2628.993
Likes147210.5752.1992.99618.858
Recommendation14720.0830.08101
Type14720.4540.49801
Gender14720.810.39301
Age147226.1687.135364
Speed14725.6340.40.7426.679
Facial attractiveness147279.8387.93542.83897.399
Retention14724.2940.7542.1977.051
Table 3. Correlation analysis.
Table 3. Correlation analysis.
Variables(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)
(1) Fan growth1.000
(2) Interaction quality0.3081.000
(3) Interaction frequency0.6560.1421.000
(4) Duration0.3260.034−0.1071.000
(5) Reputation0.1460.0000.1540.1021.000
(6) Commodity0.208−0.0340.1920.1430.1661.000
(7) Price0.0370.0270.0500.1530.058−0.0071.000
(8) Likes0.7040.0960.7240.2060.2000.2330.0511.000
(9) Recommendation−0.115−0.041−0.081−0.0690.048−0.0490.016−0.1081.000
(10) Type0.050−0.1380.197−0.1970.0980.078−0.0170.1970.1201.000
(11) Gender−0.0920.013−0.1270.033−0.015−0.046−0.019−0.1850.038−0.0761.000
(12) Age0.046−0.0900.037−0.0490.0100.059−0.0330.124−0.0080.074−0.0721.000
(13) Speed0.1300.0710.0210.101−0.0170.0040.017−0.001−0.024−0.0670.031−0.0491.000
(14) Facial attractiveness−0.163−0.555−0.0880.0160.0260.0740.070−0.0710.0170.1090.059−0.030−0.0401.000
(15) Retention0.3040.9680.1430.035−0.009−0.0550.0170.101−0.037−0.1490.013−0.0940.071−0.5861.000
Table 4. Baseline regression results.
Table 4. Baseline regression results.
(1)(2)(3)(4)
Model 1
(Fan)
Model 2
(Fan)
Model 3
(Retention)
Model 4
(Retention)
Interaction frequency0.475 *** 0.164 ***
(15.4895) (3.9667)
Interaction quality 0.235 *** 0.963 ***
(11.8856) (79.7437)
Duration0.304 ***0.172 ***0.012−0.001
(17.1543)(10.2905)(0.3943)(−0.0819)
Reputation−0.012−0.002−0.014−0.008
(−0.7109)(−0.1145)(−0.4822)(−1.1502)
Commodity0.0050.042 **−0.078 ***−0.024 ***
(0.2905)(2.3109)(−2.8600)(−3.8869)
Price−0.050 ***−0.029 *−0.001−0.010
(−3.3124)(−1.7431)(−0.0430)(−1.5986)
Likes0.307 ***0.640 ***0.0490.022 ***
(9.7489)(38.0040)(1.0733)(3.0192)
Recommendation−0.015−0.020−0.0000.006
(−0.7334)(−1.0423)(−0.0175)(0.7909)
Type−0.036 **−0.003−0.170 ***−0.018 **
(−2.1159)(−0.1661)(−6.3281)(−2.5725)
Gender0.0100.0170.0180.001
(0.6848)(1.0718)(0.6876)(0.1076)
Age0.012−0.001−0.084 ***−0.007
(0.6981)(−0.0490)(−3.3940)(−0.9845)
Speed0.087 ***0.095 ***0.050 **0.002
(3.3307)(3.5690)(2.1166)(0.2785)
N1472147214721472
Adj. R20.63250.59600.06300.9381
Standardized beta coefficients; t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.
Table 5. Results of mediating effect and moderating effect.
Table 5. Results of mediating effect and moderating effect.
(1)(2)(3)(4)
Model 5Model 6Model 7Model 8
Interaction frequency0.443 *** 2.737 ***
(14.7774) (22.1521)
Interaction quality 0.224 ** 1.854 ***
(2.0118) (15.2527)
Retention0.198 ***0.087 *
(11.1876)(1.7951)
Facial attractiveness 0.208 ***0.366 ***
(8.3003)(6.4285)
Interaction frequency * Facial attractiveness −2.668 ***
(−22.1925)
Interaction quality * Facial attractiveness −0.778 ***
(−7.5363)
Duration0.302 ***0.172 ***−0.0030.001
(17.7651)(10.2874)(−0.1186)(0.2178)
Reputation−0.010−0.002−0.006−0.006
(−0.5563)(−0.1152)(−0.2580)(−0.8981)
Commodity0.0200.042 **−0.054 **−0.020 ***
(1.2569)(2.3129)(−2.3555)(−3.4140)
Price−0.049 ***−0.029 *0.013−0.004
(−3.3851)(−1.7473)(0.5583)(−0.6947)
Likes0.297 ***0.640 ***0.0500.017 **
(9.7369)(37.8592)(1.3169)(2.4243)
Recommendation−0.015−0.020−0.0200.004
(−0.7993)(−1.0401)(−0.8076)(0.5450)
Type−0.003−0.003−0.108 ***−0.013 *
(−0.1740)(−0.1672)(−4.6977)(−1.9165)
Gender0.0070.0170.0170.004
(0.4608)(1.0718)(0.7949)(0.6972)
Age0.028 *−0.001−0.047 **−0.014 **
(1.7508)(−0.0497)(−2.2622)(−2.0129)
Speed0.077 ***0.095 ***0.031 *0.003
(3.1795)(3.5680)(1.6527)(0.4896)
N1472147214721472
Adj. R20.66890.59570.32810.9441
Standardized beta coefficients; t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.
Table 6. Robustness test using model replacement and dependent variable substitution.
Table 6. Robustness test using model replacement and dependent variable substitution.
(1)(2)(3)(4)
Model 9Model 10Model 11Model 12
Interaction frequency0.475 *** 0.112 **
(18.8965) (2.2884)
Interaction
quality
0.235 *** 0.521 ***
(13.8617) (26.3221)
Duration0.304 ***0.172 ***0.007−0.007
(16.3457)(9.6139)(0.2363)(−0.2799)
Reputation−0.012−0.002−0.073 **−0.069 **
(−0.7492)(−0.1368)(−2.1952)(−2.4280)
Commodity0.0050.042 **−0.135 ***−0.104 **
(0.3054)(2.4283)(−2.6767)(−2.0923)
Price−0.050 ***−0.029 *0.0050.001
(−3.1000)(−1.7388)(0.1837)(0.0531)
Likes0.307 ***0.640 ***−0.028−0.024
(11.6585)(34.3125)(−0.5611)(−1.0565)
Recommendation−0.015−0.020−0.0000.003
(−0.9204)(−1.1785)(−0.0016)(0.1650)
Type−0.036 **−0.003−0.155 ***−0.073 ***
(−2.1646)(−0.1647)(−6.7551)(−3.8477)
Gender0.0100.0170.043 **0.034 **
(0.6387)(0.9825)(2.0774)(2.2521)
Age0.012−0.001−0.058 ***−0.017
(0.7226)(−0.0523)(−2.6829)(−0.9799)
Speed0.087 ***0.095 ***0.0360.011
(5.4779)(5.7190)(1.6298)(0.5606)
N1472147214721472
Adj. R2 0.05840.3131
Standardized beta coefficients; t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.
Table 7. Robustness test using winsorized continuous variables.
Table 7. Robustness test using winsorized continuous variables.
(1)(2)(3)(4)(5)(6)(7)(8)
Model 13Model 14Model 15Model 16Model 17Model 18Model 19Model 20
Interaction frequency0.478 *** 0.163 *** 0.445 *** 2.772 ***
(15.3898) (3.9509) (14.6690) (23.0799)
Interaction quality 0.233 *** 0.966 *** 0.204 *** 1.729 ***
(12.5223) (93.7488) (2.7964) (16.1169)
Retention 0.198 ***0.030
(11.6543)(0.4120)
Facial attractiveness 0.206 ***0.313 ***
(8.2904)(6.1992)
Interaction frequency * Facial attractiveness −2.701 ***
(−23.0974)
Interaction quality * Facial attractiveness −0.663 ***
(−7.4236)
Control YesYesYesYesYesYesYesYes
N14721472147214721472147214721472
Adj. R20.63050.59150.06320.94380.66700.59130.33330.9481
Standardized beta coefficients; t statistics in parentheses; *** p < 0.01.
Table 8. PSM.
Table 8. PSM.
(1)(2)
Model 21Model 22
Interaction
frequency
0.378 ***
(9.0666)
Interaction
quality
0.232 ***
(8.4640)
Duration0.361 ***0.188 ***
(8.5946)(7.8809)
Reputation−0.004−0.012
(−0.1250)(−0.3947)
Commodity0.0250.069 ***
(0.7656)(2.8228)
Price−0.036−0.027
(−1.1739)(−1.1372)
Likes0.225 ***0.622 ***
(4.2699)(26.5425)
Recommendation−0.065−0.040
(−1.5820)(−1.4341)
Type−0.0300.037
(−0.7975)(1.5435)
Gender0.0050.011
(0.1589)(0.5437)
Age−0.0230.004
(−0.5902)(0.1639)
Speed0.091 *0.140 ***
(1.8076)(3.3272)
N455757
Adj. R20.44360.6064
Standardized beta coefficients; t statistics in parentheses; * p < 0.1, *** p < 0.01.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Yang, H.; Wang, B. The Power of Interaction: Fan Growth in Livestreaming E-Commerce. J. Theor. Appl. Electron. Commer. Res. 2025, 20, 203. https://doi.org/10.3390/jtaer20030203

AMA Style

Yang H, Wang B. The Power of Interaction: Fan Growth in Livestreaming E-Commerce. Journal of Theoretical and Applied Electronic Commerce Research. 2025; 20(3):203. https://doi.org/10.3390/jtaer20030203

Chicago/Turabian Style

Yang, Hangsheng, and Bin Wang. 2025. "The Power of Interaction: Fan Growth in Livestreaming E-Commerce" Journal of Theoretical and Applied Electronic Commerce Research 20, no. 3: 203. https://doi.org/10.3390/jtaer20030203

APA Style

Yang, H., & Wang, B. (2025). The Power of Interaction: Fan Growth in Livestreaming E-Commerce. Journal of Theoretical and Applied Electronic Commerce Research, 20(3), 203. https://doi.org/10.3390/jtaer20030203

Article Metrics

Back to TopTop