Live streaming e-commerce (LSE) represents a novel approach to present products. The communication strategy of streamers within LSE surpasses those of static and impersonal e-commerce websites, offering consumers more engaging and persuasive product presentations, which are pivotal in selling products. Indeed, existing literature identifies content, delivery, and source as three fundamental aspects of effective communication. However, research on the effect of streamer communication on product sales has not been explored. The reason is threefold. First, in terms of communication content, streamers convey dynamic and personal experiences in product presentation with verbal speech structure and dynamic emotional expressions. Nevertheless, there is limited understanding of how such verbal speech structure and dynamic emotional expressions contribute to streamers' experiences and their impact on product sales. Second, in terms of communication delivery, streamers often need to uphold their role as authoritative product recommenders by displaying powerful bodily behaviors in live streaming videos. However, there is a lack of clarity on how these bodily behaviors that deliver an authoritative presentation influence product sales. Third, in terms of communication source, previous research on live streaming typically assumes that consumers process LSE product presentations in detail, overlooking the potential role of mental shortcuts in consumer assessments of these presentations. In reality, people often rely on mental shortcuts, especially through forming facial impressions from others' faces, to make judgments quickly and efficiently. Little is known about how streamers' facial appearance plays a crucial role in consumers' judgment processes, thereby influencing product sales.

To address these research gaps, this study draws on narrative transportation theory, powerful posing literature, stereotype content model, and overgeneralization theory, and proposes the following primary research question: How will the content, delivery, and source of streamer communication influence product sales? To answer this question, the research is divided into three main investigations. First, focusing on communication content, the study explores the sales impact of two narrativity dimensions (temporal embedding and emotional volatility) in LSE, guided by narrative transportation theory. Second, concerning communication delivery, to understand how streamers use bodily behaviors to communicate authoritatively, the research examines the sales impact of two types of powerful bodily displays (posture expansiveness and head upwardness) in LSE. This draws from powerful posing literature and the stereotype content model. Third, regarding the communication source (streamer), the study examines the sales impact of two types of streamer facial appearance: the emotional-face overgeneralization trait (facial dominance) and the baby-face overgeneralization trait (facial babyishness). This investigation is grounded in overgeneralization theory. Out of the 14 hypotheses, 10 are supported by the secondary data consisting of 1,145 product-level observations and video clips gathered from Taobao Live, one of the most prominent LSE platforms in China. By addressing the three research questions outlined earlier, this dissertation aims to make three theoretical contributions.

First, focusing on communication content, this research adopts the narrative transportation lens to enhance the comprehension of streamers' dynamic and personal product experiences. It introduces verbal speech structure (temporal embedding) and dynamic emotional expressions (emotional volatility) as two narrative dimensions and identifies their distinct effects on product sales under conditions of product trial and interaction prompt. Extant e-commerce studies have mainly focused on the static and impersonal product presentation format, with little attention to the streamer's dynamic and personal experience via verbal speech structure and dynamic emotional expressions in LSE. Moreover, prior research on live streaming has provided limited insights into the sales impact of a streamer's verbal speech structure and the dynamic emotional expressions on their faces, suggesting an opportunity for further investigation. By proposing a novel lens of narrative transportation to unveil the inherent narrativity of LSE product presentation, this dissertation argues that streamers can employ narrative elements—specifically, temporal embedding and emotional volatility—in their product presentations, thereby driving significant sales. The distinctive context of LSE also prompts us to contextualize narrative transportation theory under the conditions of product trial and interaction prompt. The results indicate that temporal embedding enhances product sales only when product trial is high. Interestingly, emotional volatility demonstrates a more pronounced positive impact on product sales when streamers don't visually demonstrate products (i.e., low product trial) or don't initiate interactions (i.e., low interaction prompt). As such, this research enriches the understanding of streamers' dynamic and personal experiences and extends narrative research to a broad arena. Furthermore, this research empirically utilizes objective data from a real-world LSE platform, enriching the methodological portfolio of the live streaming literature, which has predominantly relied on the collection of interview or survey data for consumers' subjective evaluations. The study also employs text analysis, computer vision techniques, and manual coding to quantify constructs such as temporal embedding, emotional volatility, product trial, and interaction prompt, which have traditionally been assessed through surveys and experimental works. By doing so, the proposed method enriches the relevant literature by aiding in the operationalization of these constructs, thus providing new insights into the sales impact of streamer communication content.

Second, concerning communication delivery, the research explores how streamers exhibit bodily behaviors to maintain their authoritative role as product recommenders. Drawing from powerful posing literature and the stereotype content model, it investigates the sales impact of powerful bodily displays such as posture expansiveness and head upwardness and examines the moderating effect of interaction response. Although the first study found that the narrativity of streamer communication content is beneficial to improving product sales, there is still limited understanding of how these messages are delivered authoritatively. That is, LSE streamers differ from traditional e-commerce by broadcasting products from multiple brands and sellers. They need to be seen as authoritative figures in their product recommendations. The subtle delivery of authority often occurs through bodily behaviors, specifically the body postures and head movements of LSE streamers during product presentations. Live streaming literature does not address that aspect either. Therefore, this research focuses on powerful bodily display and investigates the effectiveness of LSE streamer posture expansiveness and head upwardness. The findings show a positive relationship between posture expansiveness and product sales. Head upwardness also significantly improves product sales. In addition, interaction response weakens the sales impact of posture expansiveness. Therefore, this research proposes and validates the effectiveness of authoritative communication delivery. Moreover, this study employs empirical analysis of objective data extracted from a real-world LSE platform, enhancing the methodological diversity of the live streaming literature, which has heavily relied on interviews or surveys for subjective consumer evaluations. This research also utilizes machine learning models and manual coding to quantify posture expansiveness, head upwardness, and interaction response. While these concepts have been subjectively evaluated by survey and experimental studies, they have not been quantified using empirical data. By addressing this gap, the proposed approach contributes to the relevant literature by operationalizing powerful bodily displays and interaction response specific to LSE with unstructured video data.

Third, regarding the communication source, this research reveals the significance of facial appearance as a mental shortcut in shaping consumers' perceptions of the communication source (streamer). By integrating literature on trait impressions of faces and overgeneralization theory, it investigates the impact of facial dominance and babyishness on product sales and explores the moderating effect of product trial. While a substantial body of research in live streaming has explored various consumer-, video-, and streamer-related antecedents influencing consumer purchase intention, the central assumption is that consumers deliberately and systematically process the information presented in LSE product presentations. However, this perspective overlooks the potential role of mental shortcuts, especially through forming impressions of streamers' facial appearance, which shape consumer judgments more straightforwardly. Building on overgeneralization theory, this study proposes that facial dominance elicits an emotional-face overgeneralization effect, thereby positively influencing product sales. Facial babyfacedness induces a baby-face overgeneralization effect, leading to enhanced product sales. Furthermore, these relationships are contingent on product trial. These findings complement the extant live streaming literature by highlighting the significance of streamer facial appearance as a mental shortcut in consumer decision-making processes. Additionally, this research relies on empirical analysis of objective data gathered from a real-world LSE platform, thus diversifying the methodological approaches in the live streaming literature. Traditionally, this literature has heavily depended on interviews or surveys for subjective consumer evaluations. This study also utilizes pre-trained machine learning models and manual coding to operationalize facial dominance, facial babyishness, and product trial. These constructs are primarily quantified by current research through subjective assessments, such as survey items. The approach utilized in this study effectively quantifies facial appearance and product trial using real-world data from LSE, providing novel insights into the sales influence of the communication source.

This dissertation also offers practical implications for streamers on product presentation with effective communication strategies, and for LSE platform owners on guiding and selecting streamers in terms of communication content, delivery, and source.