Does it matter how I stream? Comparative analysis of livestreaming marketing formats on Amazon Live

The study addresses how different livestreaming marketing product demonstration formats on e-commerce platforms influence consumer engagement. Motivated by the surge in livestreaming commerce during COVID-19 and the practical need for sellers to choose effective formats, the paper examines whether and how the interview, tutorial, and behind-the-scenes formats differentially leverage social capital acquisition and social endorsement to drive engagement. Drawing on social capital and signaling theories, the authors hypothesize positive effects of social capital on consumer engagement (H1a) and on social endorsement (H2a), and a positive effect of social endorsement on consumer engagement (H3a), with each effect differing by format (H1b, H2b, H3b). The work is important because format choice is a controllable decision for sellers that can materially affect real-time engagement behaviors (likes, chats, click-through purchases) and, ultimately, sales.

The theoretical background integrates social capital theory and signaling theory within livestreaming commerce. Social capital theory emphasizes intangible resources embedded in social structures—trust, interactions, reciprocity—facilitating goal attainment. In livestreaming, five determinants of social capital acquisition are adopted: physical proximity, social proximity, familiarity (regular viewers), parasocial interaction (reciprocal, temporal interaction during live sessions), and self-disclosure. These factors enhance ties between sellers and viewers, fostering trust and engagement. Signaling theory underpins social endorsement: observable signals (e.g., followers, viewers, likes) convey product value and seller credibility to others. The study conceptualizes social endorsement as follower endorsement (new followers gained) and viewer endorsement (volume/retention/time). Prior work links endorsements to brand communication and purchase intentions; however, comparisons across livestreaming formats are sparse. The paper also reviews livestreaming marketing formats—interview (with key opinion leaders), tutorial (demonstrating product use), and behind-the-scenes (production/inside stories)—and posits that formats differentially humanize products, accumulate social capital, elicit endorsement, and drive engagement. Hypotheses H1a–H3b formalize expected positive relationships and cross-format differences.

Design and data: A cross-sectional survey of Amazon Live sellers was conducted from January to May 2022. Researchers followed ~1000 streamers, randomly observed sessions, and distributed a hyperlinked questionnaire via email and social media. Respondents provided their Amazon Live screenname for verification. N=766 complete responses (no missing data) were obtained. Sellers self-identified their predominant livestreaming format: interview (27.8%), tutorial (37.7%), behind-the-scenes (34.5%). Product categories included tech (25.6%), beauty (20.5%), baby (9.1%), apparel (34.1%), medical (3.8%), other (6.9%). Most had 2–3 years of livestreaming experience (27.9%) and streamed at least weekly (64.2%).

Measures: All constructs were modeled as first-order reflective constructs, measured on 7-point Likert scales (1=strongly disagree, 7=strongly agree) adapted from validated scales. Social capital acquisition captured physical proximity, social proximity, familiarity, parasocial interaction, and self-disclosure (from Küper & Krämer, Lang). Social endorsement reflected viewer attraction, retention, exposure time, conversion to follower rate, and follower retention rate (Thai & Wang). Consumer engagement measured observable behaviors: likes, frequency of likes, chat participation rate/frequency, and click-through purchase (Vivek et al., Thai & Wang).

Analysis: Partial least squares structural equation modeling (PLS-SEM) was used, with case-wise analyses per format: case 1=interview, case 2=tutorial, case 3=behind-the-scenes. Confirmatory tetrad analysis supported reflective specification. Indicator loadings exceeded 0.7 with significant t-values. Composite reliability (CR>0.7), AVE (>0.5), and rho_A confirmed reliability; Fornell–Larcker criterion supported discriminant validity. Collinearity checks showed VIF<5 (max 1.250). Structural models estimated R² and path coefficients; significance assessed via bootstrapping. Cross-case differences were evaluated with Henseler’s bootstrap-based multigroup analysis (MGA).

• Model fit and variance explained: R² for Consumer Engagement (CE) was 0.881 (case 1: interview), 0.732 (case 2: tutorial), 0.791 (case 3: behind-the-scenes). R² for Social Endorsement (SE) was 0.687, 0.598, and 0.629 for cases 1–3, respectively.
• Structural paths (Table 6):
  • Social Capital (SC) → Consumer Engagement (CE): case 1 c=0.278, t=3.368, p<0.01; case 2 c=0.353, t=3.543, p<0.001; case 3 c=0.202, t=2.027, p<0.05. H1a supported in all cases.
  • Social Capital (SC) → Social Endorsement (SE): case 1 a=0.829, t=11.400, p<0.001; case 2 a=0.580, t=9.482, p<0.001; case 3 a=0.399, t=4.153, p<0.001. H2a supported in all cases.
  • Social Endorsement (SE) → Consumer Engagement (CE): case 1 β=0.695, t=7.177, p<0.001; case 2 β=0.277, t=2.547, p<0.05; case 3 β=0.578, t=5.466, p<0.001. H3a supported in all cases.
• Relative effects by format:
  • SE → CE effects are strongest in interview (β=0.695), then behind-the-scenes (β=0.578), weakest in tutorial (β=0.277).
  • SC → CE effect is strongest in tutorial (c=0.353) versus interview (0.278) and behind-the-scenes (0.202).
  • SC → SE strongest in interview (a=0.829), then tutorial (0.580), then behind-the-scenes (0.399).
• Multigroup analysis (MGA, Table 7): Significant cross-format differences were found for SC→CE between tutorial and behind-the-scenes (diff=0.151, p=0.013); for SC→SE across all pairwise comparisons (p≤0.010); and for SE→CE across all pairwise comparisons (p≤0.021). H1b, H2b, H3b supported.
• Efficiency insight: Despite being the least-used format (27.8%), interview format is more efficient at attracting consumer engagement via stronger social endorsement effects and higher variance explained in CE than tutorial or behind-the-scenes.
• Overall, social endorsement generally exerts a larger impact on engagement than social capital in interview and behind-the-scenes formats, while in tutorial streams, social capital’s direct effect on engagement exceeds that of social endorsement.

Findings confirm that format choice materially shapes how social capital and social endorsement translate into engagement during livestreaming commerce. Interview streams best leverage social endorsement signals (viewers and followers) to drive engagement, suggesting that visible social proof and third-party credibility in interview settings amplify behavioral responses (likes, chats, click-throughs). Behind-the-scenes streams also benefit substantially from endorsement cues, likely due to transparency and authenticity enhancing perceived value. In contrast, tutorial streams rely more heavily on the seller’s ability to acquire social capital—building closeness, familiarity, reciprocity, and self-disclosure—to stimulate engagement, indicating that instructional content may require stronger relational bonds to convert attention into actions. These results address the research question by demonstrating that while SC and SE each positively affect engagement, their magnitudes and interplay vary systematically by format. The evidence underscores the role of signaling (endorsement cues) and relationship-building (social capital) as distinct yet complementary mechanisms, with optimal emphasis depending on the chosen demonstration format. The study advances understanding of livestreaming commerce by shifting focus from general participation motives to format-specific antecedents of observable engagement behaviors on a retail-native platform (Amazon Live).

The study shows that in e-commerce livestreaming, the chosen product demonstration format matters for maximizing engagement. Across interview, tutorial, and behind-the-scenes formats, both social capital acquisition and social endorsement positively influence consumer engagement, but their relative effects differ by format. Interview streams harness endorsement signals most effectively to drive engagement; behind-the-scenes streams also benefit strongly from endorsement; tutorial streams depend more on social capital to elicit engagement. These insights extend social capital and signaling theories to format-specific livestreaming contexts and provide actionable guidance for sellers when designing streams to enhance observable engagement behaviors. Future research should test additional antecedents (e.g., streamer gender, product type), validate the model across other platforms (e.g., Facebook Live, Instagram Live), and explore causal designs to strengthen inference.

• Potential omitted variables: Factors such as streamer gender and product type may influence engagement but were not modeled.
• Single platform: Data were collected solely from Amazon Live, which may limit generalizability to other platforms or cultural contexts.
• Cross-sectional, self-reported measures: Although observable behaviors were included, reliance on self-reported constructs and a single time point constrains causal inference.
• Format self-identification: Sellers chose the format that best described their presence, which may introduce classification subjectivity.