How Does Short Video Advertisement Congruence Drive Sales? The Underlying Mechanism of Sociability

Abstract

With the rapid growth of social media platforms, short video advertisements (SVAs) have been a dominant channel for product sales. However, how to design SVAs that effectively drive product sales, especially in relation to previous SVAs and the related product information, remains underexplored. This study investigates how SVA title congruence influences sales performance through the mediating role of sociability. Specifically, we conceptualized video-video title congruence and video-product title congruence as two forms of content congruence and investigated their effects using data collected from Douyin, the leading short video platform. The empirical results with two-way fixed effects show that high video-video title congruence and low video-product title congruence are both associated with higher product sales. Sociability mediates the relationship between title congruence and sales performance. This study also finds that the creation frequency and product brand significantly moderate these relationships. Furthermore, this study develops several checks to ensure the robustness of the research model and findings, including the Heckman two-stage test. These findings provide theoretical insights into content creation strategies and offer practical implications for both creators and platform managers.

1. Introduction

Short video advertisements (SVAs) have become a dominant force in social commerce, transforming how consumers discover and purchase products. These SVAs, lasting from a few seconds to several minutes, leverage the interactive and emotionally engaging nature of social media to promote products [1]. They often contain embedded purchase links that enable consumers to complete instant transactions while watching the video [2]. Due to their informativeness, rapid dissemination, and low production cost [3], SVAs significantly enhance consumer engagement [4] and influence online purchasing decisions [5,6]. The proliferation of short videos has reshaped online consumption behavior, yet it has also raised concerns such as content saturation, user fatigue, and declining trust in created content. This situation highlights a pressing social issue: How to balance commercial persuasion with authentic social interaction in digital environments.

Unlike traditional e-commerce platforms (i.e., Taobao and JD, Douyin (the Chinese version of TikTok) has become China’s leading short video platform, integrating entertainment, social interaction, and e-commerce into a unified digital ecosystem [7,8]. Compared with competing platforms such as Kwai in China and its international counterpart TikTok, Douyin distinguishes itself through its sophisticated recommendation algorithms, highly personalized content feeds, and seamless in-app purchasing experience that effectively transforms user engagement into sales [9]. This integration enables both businesses and individuals to promote and sell products through short video advertisements (SVAs) [5,10]. However, it also faces challenges such as content homogenization, algorithmic dependency, and intensified competition for visibility [11,12]. Within this environment, creators can obtain valuable insights into consumer preferences and engagement intensity through sociability metrics, such as likes, comments, and shares [13]. Although some viewers may purchase products directly without engaging in social interactions, these social performance indicators not only reflect a proportion of audience acceptance and content popularity but also influence the platforms’ recommendation algorithms, which, in turn, shape visibility, conversion, and sales performance. Therefore, determining which SVA features can enhance sociability and sales performance is crucial for creators. During the content creation process, creators should balance the congruence between their current and previous SVAs with the congruence between the video and the promoted product. Identifying these key influencing factors allows creators to design more persuasive and market-effective SVAs that optimize both user engagement and commercial outcomes.

Among various factors that shape the effectiveness of SVAs, such as video content, visual design, music, and storytelling, the video title plays a distinctive yet often underappreciated role. As the first textual cue users view when scrolling through videos, the title serves as a cognitive entry point, summarizing the core message, signaling thematic relevance, and helping users decide whether to engage further [14,15]. The title-related characteristics do not replace other multimodal features, such as narrative, music, or content design. It complements them by enhancing recognizability and trust, particularly under conditions of brief exposure and high content overload.

Existing research has concentrated on various consistent cues in short video contexts, such as creator-product congruence [3,16], creator-video congruence [17], and content-product congruence [18]. However, these studies focus on consumer perspective and rely on questionnaires or manual coding, with limited use of large-scale secondary data. To address this gap, this study utilizes real-world data collected from Douyin to explore how the SVA congruence characteristics (congruence with previous SVAs and congruence with the related product) affect product sales performance. By analyzing secondary data from platforms, this study enhances the objectivity and authenticity of research findings, thereby addressing the shortcomings of existing research in terms of data authenticity [3].

In addition, the sociability of SVAs, an essential indicator of content popularity, has been examined from multiple factors, including creator characteristics (number of followers, number of videos) and SVA characteristics (duration, posting days, title features, information disclosure, etc.) [19,20,21]. However, limited attention has been paid to how SVA congruence characteristics influence sociability. And how does sociability affect the relationship between SVA congruence and product sales? Prior research has acknowledged the positive effect of sociability on product sales [13], but they have overlooked the mediating role of sociability between SVA congruence and sales performance.

To fill these gaps, this study focuses on SVA creators as key stakeholders and explores how their consistent creation strategies—particularly title design—can enhance sociability and, in turn, sales performance. Drawing on the Signaling theory, we conceptualize the SVA title congruence as a video signal that influences sociability, which then drives product sales. By analyzing a large dataset from Douyin, we find that video-video title congruence (congruence with previous SVAs) positively affects sociability, while video-product title congruence (congruence with the related product) has a negative effect. In addition, SVA sociability mediates the relationship between title congruence and sales performance. These relationships are further moderated by creator factors (creation frequency) and product characteristics (brand type). Additional analysis reveals that textual features of SVA titles (sentiment, first-person pronouns, and vividness) also significantly affect sales performance.

Our study makes the following contributions. First, this study concentrates on the SVAs sales performance from the perspective of creators, and we uncover the mediating role of sociability and enrich the literature on social media marketing. Second, our findings extend the relevant research on congruence, revealing the differential effects of congruence characteristics in different contexts through empirical analysis. Our findings also provide practical implications for both creators and platforms. This study provides suggestions for creators to optimize SVA creation strategies and boundary conditions to improve their performance. We also suggest platforms to encourage creators to post high-quality and attractive video content. Based on this strategy, platforms not only enhance consumers’ preference and retention but also improve their profitability, ultimately realizing a win-win situation for both creators and platforms.

The remainder of this paper is organized as follows. Section 2 reviews the related literature on SVAs, congruence, and Signaling theory. Section 3 introduces research hypotheses. Section 4 describes the research methodology, including the data source and variable measurements. Section 5 exhibits empirical results, and Section 6 discusses results with theoretical and practical implications, as well as limitations and future directions of this study.

2. Literature Review

2.1. Short Video Advertisements (SVAs)

Digital advertising has been defined as the “Digital advertising involves paid media placements where advertisers bid for space to display promotional messages on digital platforms (e.g., search engines, social media, websites)” [22]. In digital contexts, this function extends to short video advertisements (SVAs), which integrate entertainment and commerce to engage consumers in interactive consumption journeys. SVAs, with durations ranging from a few seconds to minutes, contain embedded purchase links that enable consumers to complete instant transactions while watching the video [1,2,13]. Due to their cost-efficiency, rapid dissemination, and effective aggregation of viewership traffic, SVAs have been increasingly prevalent for product promotion and online sales [23].

With the popularity of this advertising model, scholars have explored how various SVA’s characteristics—such as posting days, duration, thumbnail, and title—affect advertising performance [3,6,15,19,24] (See

Table 1. Overview of literature on short video advertisements.

Research	Research Context	Theory	DV	IV	Moderator (mo) /Mediator (me)	Key Findings
Addo, Akpatsa [6]	JD, Tmall and Taobao	Signaling theory	Product sale	Short video ads (binary variable)	Price (mo)/product quality signal and consumer satisfaction (me)	The use of short video advertisements will promote product sales; the positive relationship is mediated by product quality signal and consumer satisfaction.
Dong, Liu [3]	Weibo short video	-	Consumer social engagement behavior (likes, comment, and share).	Video content features (content matching, information relevance, storytelling and emotionality)	Video release time (mo)	Short video content features significantly affect consumer engagement, compared to videos released in the afternoon, videos released in the morning increase the positive effects of warmth, excitement, and joy on consumer engagement.
Xiao, Li [24]	TikTok	Uses and gratifications theory, signaling theory	Consumer engagement	Performance expectancy; entertainment; tie strength; sales approach	Product type (mo)	Four key features of short videos (performance expectancy, entertainment, tie strength, and sales approach) significantly affect consumer engagement behavior, and product type moderates these relationships.
Meng, Kou [1]	TikTok	Grounded theory	Consumer purchase behavior	Trustworthiness, expertise, attractiveness, authenticity, and brand heritage	-	Trustworthiness, expertise, and attractiveness of video content can promote consumer purchase behavior, while authenticity and brand heritage have a U-shape effect on consumer purchase.
Yang, Wang [10]	Douyin	Heuristic-systematic model theory	Enterprise performance	Voice feature; music feature	-	When short video advertisements have a faster speech rate, high voice quality, faster music tempo, lower pitch, and lower loudness, enterprises achieve a more positive performance in social e-commerce.

). In the context of short video consumption, the purchase process is as follows: Viewers are first exposed to SVAs in the platform, triggering attention and interest through multiple cues. Then, evaluation and trust form as viewers perceive creator credibility or sociability (likes, comments, and shares), ultimately leading to direct purchase via the embedded purchase link [2,13]. Unlike traditional advertising that involves prolonged consideration, SVA consumption involves instant decision-making, where video-related cues may play roles in converting attention into sales.

Among these factors, the title of the SVA is a primary indicator of its content description and a critical determinant of viewership. The video title functions as a key cognitive signal that helps viewers rapidly assess content relevance, trustworthiness, and whether to browse or engage. Previous scholars have explored titles with positive sentiments and high arousal are more likely to attract viewers [14,15]. Meanwhile, concise and rhetorically rich titles enhance information transmission and viewer interests [21,25], while longer titles provide more information that enhances memorability and sociability [26]. These findings suggest that even when viewers do not watch full videos or are only briefly exposed, the title itself shapes viewers’ perceptions, trust, and engagement, underscoring its importance as a research variable in the SVA context. However, existing studies have mainly focused on individual title characteristics, while relatively little attention has been paid to title congruence (congruence with previous SVAs titles and congruence with the related product title) and how these congruence characteristics may affect sociability and sales performance. To address this gap, this study regards SVA title congruence as the primary research variable and investigates its effects on sociability and product sales performance.

Previous studies on SVAs have been grounded in multiple theoretical perspectives that offer complementary insights into users’ cognitive, affective, and behavioral responses. Signaling theory explains how observable cues, such as the title, thumbnail, and hashtag, serve as signals that attract attention and reduce information asymmetry between content creators and viewers, thereby influencing perceptions of credibility and product quality [6]. The Uses and Gratification Theory emphasizes that users engage with short video content to satisfy intrinsic needs such as entertainment, information seeking, and social interaction, which sheds light on the role of sociability as a key outcome in the SVA context [24,27]. Moreover, the Grounded Theory and Heuristics-Systematic Model Theory are used to explain how consumers process and interpret persuasive advertising cues, such as titles, to understand and evaluate the video content [1,10]. Integrating these theoretical perspectives enables a comprehensive understanding of how the effect and mechanism of SVA title congruence influences sociability and sales performance. The present study adopts the Signaling Theory as the primary theoretical framework, given its strong explanatory power in elucidating how title-based cues shape consumer trust and behavioral responses in SVA-driven commerce.

2.2. Congruence

Congruence refers to the degree of similarity between two objects or activities [28]. Consistent information can strengthen associative impressions, further promoting cognitive and forming favorable attitudes [29,30]. In the traditional e-commerce context, congruence has been investigated across multiple dimensions, including congruence between product descriptions, pictures, and comments, which has been found to influence product evaluations and sales [31,32,33,34] (See

Table 2. Overview of congruence literature in the context of non-short video advertisements.

Research	Research Context	Theory	DV	IV	Moderator (mo) /Mediator (me)	Key Findings
Eslami and Ghasemaghaei [31]	Amazon.com	Heuristic-systematic model	Product sale	Online review positiveness, review score inconsistency	Review score, review sentiment and review title sentiment (mo)	For a high-involvement product, review score incongruence negatively influences its sales, review textual sentiment positively influences its sales, and review title sentiment and review score influence low-involvement product sales.
Maier [34]	Online retail website	-	Attitudes towards the product assortment	In congruence of image characteristics on product overview	Disfluency perception of the product assortment	Product image incongruence in product overview negatively affects the assortment evaluations and choice satisfaction of the product.
Park and Lin [33]	Taobao and Weibo	Match-up hypothesis	Purchase intention	Self-product fit, source-product fit, live content-product fit	Perceived trustworthiness (me), attractiveness (me), attitude toward content (me)	Source-product fit positively affects the source trustworthiness and attractiveness, live content-product fit positively affects attitude towards content, and source trustworthiness, hedonic attitude, and self-product fit enhance the purchase intention.
Rungruangjit [32]	Taobao	Source credibility, match-up hypothesis and parasocial relationship theory	Purchase intention	Celebrity-product congruence	Source credibility (me) and parasocial relationship (me)	A good match-up between celebrity and product can enhance perceived source credibility, and perceived expertise positively mediates the positive effect of celebrity-product congruence on purchase intention.

).

In the context of SVAs, scholars have investigated the congruence relationship, such as creators-products [16,18,30], creators-videos [17], videos-product information [3], and creators-consumers [35]. However, few studies have explored how congruence between different SVAs of the same creator influences sociability and sales performance. Additionally, when producing a new SVA, creators also consider and decide which title is consistent with the related product title. These two forms of congruence: video-vide title congruence and video-product title congruence, reflect distinct strategic choices that may signal the creator’s image and product information to audiences.

To fill the above gaps, this study adopts the creator perspective, focusing on two congruence dimensions: (1) title congruence between the current and previous SVAs; (2) title congruence between the SVA and product. Building on the Signaling theory, this study investigates the effects of the above two title congruence on SVA sales performance, further explores the mediating role of sociability, as well as the heterogeneous effects of different characteristics of creators and products.

2.3. Signaling Theory

Signaling theory, originally proposed by Spence [36], explains how the sender conveys credible information to the receiver to reduce information asymmetry. In digital environments, where consumers cannot directly verify product quality or creator reliability, signals help shape perceptions and guide decisions [37,38]. In the short video context, signaling theory has been widely used to explain how various video-related characteristics, including video design, audio, quantity, or engagement metrics, influence consumer behaviors [6,24,39]. Among these signals, the title of the SVA, serving as a concise description of its content [14], also functions as a key cognitive signal. Prior studies have confirmed that an attractive or emotionally arousing title increases viewership and engagement [14,15,25].

Building on this perspective, our paper sets out to explore the effect of title congruence signals from video creators in SVAs. Specifically, it includes two aspects: (1) the title congruence between the creator’s SVAs, and (2) the title congruence between the SVA and the related product. These two features may help creators establish their own image and attract viewership on the platform. The former one signals reliability and stable content positioning, helping users recognize and recall a creator’s brand style [40]. The latter one, while informative, may signal excessive commercial intent and potentially reduce user engagement [41].

Thus, from a signaling theory perspective, video creators should balance these cues to achieve both social resonance (sociability) and commercial effectiveness (sales performance). Drawing on signaling theory, we examine how the title congruence of SVAs influences sales performance and propose our research model and hypotheses in the following sections.

3. Hypotheses Development

3.1. Video-Video (V-V) Title Congruence and Sociability

Creators concentrate on the congruence between their current and previous SVAs to maintain a coherent identity and viewer expectations. Since the title of the SVA serves as a key descriptor of its content [14], research has confirmed that an attractive title increases viewership [25]. Video-video (V-V) title congruence refers to the consistent degree of similarity between the creator’s current and previous SVAs. From a signaling perspective, repeated use of congruent V-V titles communicates reliability and stable content positioning, reducing cognitive uncertainty and reinforcing creator identity [40,42], as well as weakening competitor interference [40,43]. Prior research on advertising congruence shows that perceived coherence across messages enhances credibility, cognitive fluency, and use liking [44,45].

Moreover, processing fluency theory suggests that congruent stimuli are easier to process and tend to elicit more positive affective responses [46,47]. On short-video platforms where viewers make split-second judgments, such as fluency fosters recognition and emotional comfort, increasing social engagement behaviors (reflected as the SVA sociability, i.e., likes, comments, and shares). Even if viewers do not consciously read or compare titles, V-V title congruence may form implicit recognition and perceived familiarity as a peripheral signal [48,49]. High V-V title congruence enables viewers to quickly associate new SVA with prior positive experiences from the same creator, thereby fostering recognition, emotional comfort, and interaction tendency. Hence, we propose that:

Hypothesis 1:

V-V title congruence has a positive effect on SVA sociability.

3.2. Video-Product (V-P) Title Congruence and Sociability

Video-product (V-P) title congruence refers to the degree of similarity between the title of the SVA and the promoted product’s title. High V-P title congruence indicates that the SVA title explicitly conveys product-related information, potentially increasing informativeness and perceived credibility, thereby enhancing product sales [32,41]. However, the core of the SVA platform, such as Douyin, is entertainment rather than commerce [50]. Research shows that high information-product congruence can induce advertising fatigue and psychological resistance [1,51]. Hence, overtly commercial or “ad-like” titles (high V-P title congruence) can disrupt entertainment experience and lower social engagement, from a signaling perspective.

In contrast, low V-P title congruence, where the title maintains a storytelling or the overall style on the SVA platform, may reduce perceived intrusiveness and enhance entertainment value [52]. Such titles can subtly enhance familiarity and a positive effect over time, encouraging social responses without viewers’ explicit awareness of the promotional intent. These indirect signals stimulate curiosity and pleasure [34,53,54], increasing the sociability of the SVA through more viewer social engagement. Based on the above discussion, we propose the following hypothesis:

Hypothesis 2:

V-P title congruence has a negative effect on SVA sociability.

3.3. Sociability and Sale Performance

Sociability refers to the degree to which users engage in social engagement, such as likes, comments, and shares, when watching SVAs. On short video platforms, sociability not only reflects the content’s popularity but also indicates the perceived authenticity and resonance of both the creator and the promoted product [15,55,56]. From a signaling theory perspective, when the SVA receives numerous likes or comments, potential viewers interpret these cues as a signal of quality, credibility, and relevance [57]. Such bandwagon effect encourages subsequent viewers to engage with or purchase the advertising product, thereby amplifying its sales potential [15,55,56].

Creators who maintain high V-V title congruence strengthen stable and recognizable identity signals, helping viewers quickly identify familiar content and strengthen their parasocial relationships with the creator [13,40,42]. These consistent signals increase the likelihood of positive affective reactions, such as enjoyment and trust, which manifest as sociability behaviors.

Conversely, SVAs with low V-P title, which integrate product information subtly, tend to elicit curiosity and emotional resonance rather than resistance [50,52]. Such emotional engagement fosters deeper social interaction (i.e., likes, comments, shares), reinforcing viewer-creator connections and indirectly enhancing purchase potential.

Empirical evidence supports that social engagement indicators, likes, shares, and comments, serve as leading indicators of purchase conversion [13]. These metrics increase exposure and boost viewers’ confidence in the product. Therefore, sociability bridges the relationship between SVA title congruence and sales performance. According to the above discussion, we propose that:

Hypothesis 3:

SVA sociability has a positive effect on sales performance.

Hypothesis 4:

SVA sociability mediates the effects of title congruence on sales performance.

3.4. Moderation Effect of Creation Frequency

Creation frequency, defined as the number of SVAs posted per day by the creator. It serves as a crucial behavioral signal that conveys the creator’s activity level and shapes the degree of exposure and audience familiarity [58]. Higher creation frequency strengthens the creator’s visibility and memorability in platform [59,60]. When combined with high V-V title congruence, frequent posting further reinforces creators’ creation image and credibility image on the platforms [40], improving the visibility of the related product and amplifying sales performance. Hence, we propose that:

Hypothesis 5:

Creation frequency positively moderates the positive effect of V-V title congruence on SVA sales performance.

When V-P title congruence is high, high creation frequency may lead to information overload and cognitive fatigue [61,62]. Repeated exposure to product-focused content heightens viewers’ advertising resistance and reduces attention [1,51]. In contrast, frequent posting of low V-P title congruence (more entertaining) SVA enhances engagement and purchase likelihood by sustaining curiosity and pleasure [53]. Hence, we propose the following hypothesis:

Hypothesis 6:

Creation frequency negatively moderates the negative effect of V-P title congruence on SVA sales performance.

3.5. Moderation Effect of Product Brand

Product brand is a critical determinant of product sales [63,64] and serves as a key attribute in product promotion [3]. In online environments lacking direct product experience, brand information signal functions as a key quality signal [65], reducing perceived risk and strengthening the SVA content credibility [64]. Branded products enhance the persuasion and reinforce the positive effect of V-V title congruence on SVA sales performance. Hence, we propose the following hypothesis:

Hypothesis 7:

Product brand positively moderates the positive effect of V-V title congruence on SVA sales performance.

Previous studies have shown that branded products have higher cognitive familiarity and trust [66], which may mitigate skepticism toward overt advertising [67]. In such cases, when a branded product is involved, even high congruence V-P titles are perceived as product-centered and recognize commercial intention; the negative effect of the SVA on sales performance will be weakened due to the product brand signal [64], thereby weakening its negative effect on SVA sales performance. Based on the above discussion, we propose that:

Hypothesis 8:

Product brand positively moderates the negative effect of V-P title congruence on SVA sales performance.

Figure 1 illustrates the research model and the relationships between variables based on the hypotheses discussed.

4. Research Methodology

4.1. Data Source

To test these hypotheses, we chose Douyin as our data source to explore the effects of SVAs title congruence on sales performance, for the following reasons. First, Douyin has over 600 million daily active users in China, nearly 300 million creators, and produces 20–30 million pieces of content daily [68], which is one of the largest and most successful SVA platforms globally [23,69]. Second, Douyin offers a diverse range of products in SVAs [24], providing us with comprehensive and accurate data, making it an appropriate platform for our research.

Douyin provides an innovative business model by integrating short videos with e-commerce transactions [70]. As shown in Figure 2a, creators can embed product purchase links into their videos, allowing consumers to purchase products when watching the videos [69]. When they click the link, the interface will change into Figure 2b, where consumers can view product information, including price, title, and sale, as well as make purchase decisions by clicking the purchase button. Producers, platforms, and creators share the profits from commodity sales in a certain proportion [69,70]. This business model is the current dominant marketing model on the Douyin platform.

We randomly collected short video data from 400 creators with more than 100,000 fans who posted from January to September 2023 at three-month intervals. Consistent with prior literature [71,72], we adopted 100,000 followers as the threshold to define influencers, as this value is widely used to distinguish influencers from other creators. These creators are responsible for the majority of commercial content and account for a disproportionate share of platform-driven sales revenue [73,74,75]. While smaller creators contribute to platform activity, their commercial influence and advertising engagement rate are lower, which makes them less suitable for studying sales-driving mechanisms.

We randomly collected data using a stratified sampling approach based on content category distribution to ensure representativeness across different sectors. Considering the effect of the active period of short video on data, we removed video data that were posted more than 30 days before the data collection date to minimize data collection bias and keep variable validity, according to prior research [1,21,24]. In addition, since our study focuses on short video ads that provide product purchasing links, we removed those videos without the product purchase link. After data cleaning, 80 creators were excluded due to incomplete or inconsistent data records, or missing product information. The final data on a total of 57,192 short videos with product purchase links posted by 320 creators covering 19 sectors were collected. Each data contains characteristics of the short video ad (i.e., title, posting time, duration, likes, and comments number), related product characteristics (i.e., title, price, brand, and sales), and creator characteristics (i.e., fans number, sector, and experience).

4.2. Variable Measurements

Sales performance. We operationalized SVA success through direct sales performance, representing the ultimate commercial effectiveness. Specifically, we used product sales on the data collection day as the measurement of SVA sales performance, which is recorded by the platform.

Video-video (V-V) title congruence. We measured V-V title congruence as the level of congruence between the creator’s current short video advertisement and previous advertisements posted within the preceding 14-day period. SVAs posted from the 14 days are chosen because the active period of SVAs is relatively short [21]; and to observe whether creators can maintain a consistent SVA creation style, we selected 14 days as the observation period (To test the robustness of research results, we presented the results of three alternative time windows (7 days, 21 days, and 28 days) in the robustness test. These results for different time windows are generally consistent with the main results.). We employed word embedding techniques to measure title congruence among multiple video titles based on semantic vectors. Specifically, we first used Jieba, a widely adopted Chinese text segmentation tool, to segment each video title into word phrases while removing common stop words during this process. Subsequently, we applied word2vec technology to generate 64-dimensional word vectors for the segmented phrases from video titles, thereby capturing semantic information. Considering that each phrase carries different importance across various contexts, we calculated TF-IDF (Term Frequency-Inverse Document Frequency) values for each phrase within video titles as weights representing their significance. Each video title was then represented using the weighted sum of word vectors from its constituent phrases, to simultaneously consider the semantic information of each word and its importance in the current video title [76,77]. Finally, we evaluated the congruence between paired titles using cosine similarity (ranging from −1 to 1), a metric that quantifies the directional similarity between two vectors [78]. The value of V-V title congruence approaching 1 indicates higher similarity between the current and previous SVAs, thus reflecting greater congruence and higher creation image. We calculated the congruence between the current video title and each video title published within the past 14 days, then computed the average value as the final V-V title congruence measurement.

Video-product (V-P) title congruence. Using the same congruence calculation method, we represented the product title as semantic vectors and measured V-P title congruence by calculating the cosine similarity between the SVA title and the product title. The closer the value is to 1, the more product-promoted information is directly conveyed in the SVA.

Sociability. The mediator variable in this study refers to SVA sociability. Following prior research [13,79], we used the total sum of likes, comments, and shares of SVA to reflect its sociability. Using the sum of these three dimensions as the measurement can comprehensively assess the effectiveness of SVA social performance [57]. This variable not only reflects the viewer engagement but also serves as a behavioral process linking through how SVA title congruence translates into commercial outcomes, as suggested by the research conceptual model.

Creation frequency. We calculated the number of SVAs posted by the creator in a day to capture the creation frequency. The higher the value, the higher the creation frequency of the creator.

Product brand. We used the dummy variable to indicate whether the product is a branded product. If the product contains brand information, the value of the variable is 1; otherwise, it is 0.

Control variables. According to prior studies, we controlled potential influences on product sales in terms of SVAs, products, and creators. In terms of SVAs, we included video day of week (whether the video is posted on the working day); video time of day (whether the video is posted in the evening); video length (minutes based); video age (the days intervals between video posting date and data collection date); title length and tag length [3,21,25]. We also added creator-related characteristics to control their effects, such as fans, expertise (the day intervals of creators have been on the platform), and sex [18,20]. Product price is controlled for its potential effect on product sales [1]. In addition, we also controlled the time-fixed effect of short video ads posting month, the creator-level fixed effect, and the creator’s sector-level fixed effect to improve the robustness of our research [3,25].

Table 3. Descriptive statistics of research variables.

Variable	Mean	Std. Dev.	Min	Max
Sales	315,185.900	1,156,790	0	51,100,000
V-V title congruence	0.605	0.295	−0.277	1
V-P title congruence	0.660	0.223	−0.408	1
Sociability	695.997	7332.278	0	1,186,097
Creation frequency	6.136	4.278	1	50
Brand	0.568	0.495	0	1
Video day of week	0.728	0.445	0	1
Video time of day	0.292	0.455	0	1
Video length	0.559	0.588	0	14.280
Video age	46.983	27.163	0	97
Video title length	42.639	20.119	0	462
Video tag number	2.556	1.574	0	10
Product price	80.952	363.585	0.060	22,888
Creator fans	1,001,553	1,476,317	100,000	52,200,000
Creator expertise	112.941	85.377	2	316
Creator sex	0.597	0.490	0	1

states the descriptive statistics of research variables. Taking high skewness, kurtosis, and non-normal distribution into consideration, we applied the logarithmic transformation of these variables (except for title congruence, brand, video day of week, video time of day, video length, and creator sex) [80]. Before conducting the regression model, we analyzed the covariance and correlation among all variables. As

Table 4. VIF values of research variables.

Variable	VIF	1/VIF	Variable	VIF	1/VIF
V-V title congruence	1.230	0.810	Video age	1.010	0.994
V-P title congruence	1.440	0.693	Video title length	1.310	0.764
Sociability	1.860	0.537	Video tag number	1.360	0.737
Creation frequency	2.160	0.464	Product price	1.090	0.918
Brand	1.050	0.952	Creator fans	1.230	0.814
Video day of week	1.000	0.996	Creator expertise	2.020	0.494
Video time of day	1.010	0.987	Creator sex	1.170	0.855
Video length	1.330	0.752	Mean VIF	1.350

shows, the variance inflation factors (VIF) value between research variables is less than 10, and the correlations of all variables were all less than the threshold value (See

Table 5. Correlations among research variables (N = 57,192).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)	(16)
(1) Sales	1
(2) V-V title congruence	0.049 ***	1
(3) V-P title congruence	0.076 ***	0.113 ***	1
(4) Sociability	0.030 ***	−0.267 ***	−0.235 ***	1
(5) Creation frequency	−0.062 ***	0.049 ***	−0.041 ***	−0.417 ***	1
(6) Brand	0.326 ***	0.057 ***	0.050 ***	0.058 ***	−0.120 ***	1
(7) Video day of week	0.010 **	0.012 ***	0.004	−0.014 ***	0.028 ***	0.004	1
(8) Video time of day	0.057 ***	0.016 ***	0.005	0.024 ***	−0.030 ***	0.038 ***	−0.001	1
(9) Video length	0.006	−0.282 ***	−0.080 ***	0.427 ***	−0.198 ***	0.023 ***	0	−0.050 ***	1
(10) Video age	0.069 ***	−0.007 *	−0.021 ***	0.029 ***	−0.008 **	0.008 **	0.043 ***	0.001	−0.016 ***	1
(11) Video title length	0.092 ***	0.316 ***	0.118 ***	−0.005	−0.074 ***	0.061 ***	0.004	−0.032 ***	−0.018 ***	−0.020 ***	1
(12) Video tag number	0.031 ***	0.358 ***	0.131 ***	−0.077 ***	−0.032 ***	0.069 ***	−0.001	−0.003	−0.050 ***	0.011 ***	0.428 ***	1
(13) Product price	−0.417 ***	−0.147 ***	−0.116 ***	0.102 ***	0.093 ***	−0.146 ***	−0.005	−0.032 ***	0.129 ***	−0.024 ***	−0.064 ***	−0.075 ***	1
(14) Creator fans	−0.049 ***	0.073 ***	0.047 ***	0.298 ***	−0.037 ***	0.034 ***	0.001	0.041 ***	0.026 ***	−0.001	−0.034 ***	−0.074 ***	0.076 ***	1
(15) Creator expertise	0.015 ***	0.171 ***	0.045 ***	−0.390 ***	0.682 ***	−0.083 ***	0.005	−0.008 *	−0.251 ***	−0.003	0.035 ***	0.075 ***	−0.001	−0.071 ***	1
(16) Creator sex	−0.101 ***	0.139 ***	−0.246 ***	−0.007	0.014 ***	−0.073 ***	0.002	−0.043 ***	−0.134 ***	0.001	0.033 ***	−0.079 ***	−0.029 ***	0.110 ***	−0.039 ***	1

Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01.

), providing evidence that multicollinearity was not a concern in our research model.

5. Model Specific and Research Results

5.1. Empirical Results

5.1.1. Main Result

To test these relationships in our research model, we applied the three steps proposed by Baron and Kenny [81] to test our hypotheses. The empirical formulations are as follows:

$$\begin{array}{ll}Sales=& \alpha +{\beta }_{1}V-V title congruence+controls\\ & +\eta +v+\lambda +\epsilon \end{array}$$

(1)

$$\begin{array}{ll}Sociability=& {\alpha }_1+{\beta }_{2}V-V title congruence+controls\\ & +{\eta }_1+v_1+{\lambda }_1+{\epsilon }_1\end{array}$$

(2)

$$\begin{array}{ll}Sales=& {\alpha }_2+{\beta }_{3}V-V title congruence+{\beta }_{4}Sociability+controls\\ & +{\eta }_2+v_2+{\lambda }_2+{\epsilon }_2\end{array}$$

(3)

where $controls$ captures variables influencing SVA sale performance, such as video day of week, video time of day, video length, video age, video title length, video tag number, product price, creator fans, creator expertise, and creator sex. In addition, $\eta$ captures the time-fixed effect, $v$ represents the creator-level fixed effect, $\lambda$ represents the creator’s sector-level fixed effect, $\epsilon$ denotes the idiosyncratic error, and $\alpha$ refers to the constant term.

Table 6. OLS regression model results.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	Sale	Sociability	Sale	Sale	Sociability	Sale
V-V congruence	0.356 ***	0.381 ***	0.248 ***
	(0.056)	(0.022)	(0.056)
V-P congruence				−0.201 ***	−0.206 ***	−0.141 **
				(0.069)	(0.028)	(0.069)
Sociability			0.281 ***			0.284 ***
			(0.011)			(0.011)
Video day of week	0.035	−0.022 **	0.042 *	0.035	−0.022 **	0.042 *
	(0.023)	(0.009)	(0.022)	(0.023)	(0.009)	(0.022)
Video time of day	−0.030	0.028 ***	−0.038	−0.031	0.026 ***	−0.039
	(0.024)	(0.010)	(0.024)	(0.024)	(0.010)	(0.024)
Video length	0.706 ***	0.804 ***	0.476 ***	0.709 ***	0.807 ***	0.475 ***
	(0.064)	(0.030)	(0.064)	(0.064)	(0.030)	(0.064)
Video age	0.228 ***	0.086 ***	0.204 ***	0.229 ***	0.087 ***	0.204 ***
	(0.017)	(0.006)	(0.017)	(0.017)	(0.006)	(0.017)
Video title length	0.090 ***	0.026 **	0.082 ***	0.123 ***	0.061 ***	0.105 ***
	(0.027)	(0.012)	(0.027)	(0.027)	(0.012)	(0.027)
Video tag number	0.061 *	0.064 ***	0.044	0.073 **	0.076 ***	0.052
	(0.036)	(0.015)	(0.036)	(0.036)	(0.015)	(0.036)
Product price	−0.986 ***		−0.970 ***	−0.988 ***		−0.971 ***
	(0.020)		(0.020)	(0.020)		(0.020)
Creator fans	0.282	−0.091	0.302	0.304	−0.070	0.317
	(0.195)	(0.090)	(0.194)	(0.196)	(0.091)	(0.195)
Creator expertise	−0.171 ***	−0.199 ***	−0.114 ***	−0.183 ***	−0.212 ***	−0.122 ***
	(0.026)	(0.011)	(0.026)	(0.026)	(0.011)	(0.026)
Creator sex	1.995 ***	5.211 ***	0.569	1.710 ***	4.913 ***	0.355
	(0.442)	(0.171)	(0.443)	(0.442)	(0.169)	(0.443)
Month fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Creator fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Sector fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Constant	6.635 ***	3.486 ***	5.644 **	6.612 ***	3.491 ***	5.614 **
	(2.522)	(1.162)	(2.503)	(2.530)	(1.164)	(2.510)
N	57,192	57,192	57,192	57,192	57,192	57,192
R2	0.549	0.745	0.555	0.549	0.744	0.555
adj. R2	0.546	0.744	0.552	0.546	0.742	0.552

Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01.

shows empirical results, Model 2 verifies that V-V title congruence positively and significantly affects SVA sociability (${\beta }_2$ = 0.381, p < 0.01), which supports H1.

We then applied the three steps proposed by Baron and Kenny [81] to test hypothesis 2, the empirical formulations are as follows:

$$\begin{array}{ll}Sales=& {\alpha }_3+{\beta }_{5}V-P title congruence+controls\\ & +{\eta }_3+v_3+{\lambda }_3+{\epsilon }_3\end{array}$$

(4)

$$\begin{array}{ll}Sociability=& {\alpha }_4+{\beta }_{6}V-P title congruence+controls\\ & +{\eta }_4+v_4+{\lambda }_4+{\epsilon }_4\end{array}$$

(5)

$$\begin{array}{ll}Sales=& {\alpha }_5+{\beta }_{7}V-P title congruence+{\beta }_{8}Sociability+controls\\ & +{\eta }_5+v_5+{\lambda }_5+{\epsilon }_5\end{array}$$

(6)

Model 5 displays that V-P title congruence has a negative and significant effect on social engagement behavior (${\beta }_6$ = −0.206, p < 0.01), thus supporting H2.

Model 3 (${\beta }_4$ = 0.281, p < 0.01) and Model 6 (${\beta }_8$ = 0.284, p < 0.01) show that social engagement behavior positively and significantly affects purchase behavior, which supports H3.

In addition, V-V title congruence positively and significantly affects SVA sociability (${\beta }_2$ = 0.381, p < 0.01) and sociability positively and significantly affects sale performance (${\beta }_4$ = 0.281, p < 0.01), which indicates that sociability plays a mediating role in the relationship between V-V title congruence and sale performance (indirect effect (${\beta }_2\times {\beta }_4$) = 0.107, p < 0.01). Meanwhile, V-P title congruence negatively and significantly affects sociability (${\beta }_6$ = −0.206, p < 0.01), while sociability positively and significantly affects sale performance (${\beta }_8$ = 0.284, p < 0.01), which means that sociability plays a mediating role in the relationship between V-P title congruence and sale performance (indirect effect ${\beta }_6\times {\beta }_8$ = −0.059, p < 0.01). Thus, H4 is supported.

Although the large sample size increases statistical power, the standardized coefficients confirm that SVA title congruence has substantial real-world effects. Specifically, a one standard deviation increase in V-V title congruence is associated with a 0.381 SD increase in sociability and a 0.356 SD increase in sales. Meanwhile, a one standard deviation increase in V-P title congruence is associated with a −0.206 SD increase in sociability and a −0.201 increase in sales. In addition, sociability strongly predicts sales (${\beta }_4$ = 0.281; ${\beta }_8$ = 0.284, p < 0.01).

These results indicate that the V-V title demonstrates the strongest and most consistent influence on sales performance. It not only directly increases purchase potential but also indirectly drives sales through enhanced sociability. Conversely, V-P tile congruence exerts a negative effect on sales, indicating that over-commercialized titles may reduce viewer engagement. While all hypotheses were supported, the standardized coefficients reveal that maintaining consistent title creation patterns is the most critical factor for creators maximizing the commercial effectiveness of SVAs.

5.1.2. Moderation Effect

Based on the above analysis, we further developed the heterogeneity analysis to verify the moderating effect of creation frequency and product brand on the relationship between title congruence and sales performance. The specific equations are as follows:

$$\begin{array}{ll}\hfill Sales={\alpha }_6+{\beta }_{9}V-& V title congruence\\ & +{\beta }_{10}V-P title congruence\\ & +{\beta }_{11}Creation frequency+{\beta }_{12}Product brand\\ & +{\beta }_{13}Creation frequency\times V-V title congruence\\ & +{\beta }_{14}Creation frequency\times V-P title congruence\\ & +{\beta }_{15}Product brand\times V-V title congruence\\ & +{\beta }_{16}Product brand\times V-P title congruence\\ & +controls+{\eta }_6+v_6+{\lambda }_6+{\epsilon }_6\end{array}$$

(7)

where $Creation frequency\times V-V title congruence$ represents the interaction term between creation frequency and V-V title congruence, and ${\beta }_{13}$ represents the moderating effect of creation frequency on the effect of V-V title congruence on sale performance; $Creation frequency\times V-P title congruence$ is the interaction term between creation frequency and V-P title congruence, and ${\beta }_{14}$ represents the moderating effect of creation frequency on the effect of V-P title congruence on sale performance.

In addition, $Product brand\times V-V title congruence$ is the interaction term between product brand and V-V title congruence, $Product brand\times V-P title congruence$ is the interaction term between product brand and V-P title congruence. ${\beta }_{15}$ and ${\beta }_{16}$ denotes the moderating effect of product brand on the effect of V-V title congruence and V-P title congruence on sales performance, respectively.

As the results of Model 7 show in

Table 7. Research results of moderation effects.

Sale	Model 7	Model 8	Model 9	Model 10	Model 11
V-V title congruence	0.354 ***		−0.082		−0.064
	(0.057)		(0.071)		(0.071)
V-P title congruence		−0.250 ***		−0.821 ***	−0.738 ***
		(0.069)		(0.094)	(0.094)
Creation frequency	−0.281 ***	0.139 ***			0.035
	(0.046)	(0.054)			(0.065)
Creation frequency × V-V title congruence	0.148 **				0.153 ***
	(0.060)				(0.058)
Creation frequency × V-P title congruence		−0.501 ***			−0.458 ***
		(0.072)			(0.070)
Product brand			1.330 ***	0.982 ***	0.740 ***
			(0.055)	(0.083)	(0.092)
Product brand × V-V title congruence			0.702 ***		0.653 ***
			(0.082)		(0.083)
Product brand × V-P title congruence				1.092 ***	0.908 ***
				(0.115)	(0.115)
Month fixed effect	Yes	Yes	Yes	Yes	Yes
Creator fixed effect	Yes	Yes	Yes	Yes	Yes
Sector fixed effect	Yes	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes	Yes
Constant	5.139 **	3.503	11.536 ***	11.365 ***	8.848 ***
	(2.540)	(2.556)	(2.443)	(2.440)	(2.455)
N	57,192	57,192	57,192	57,192	57,192
R2	0.550	0.550	0.578	0.578	0.580
adj. R2	0.547	0.547	0.576	0.576	0.577

Standard errors in parentheses: ** p < 0.05, *** p < 0.01.

, creation frequency positively moderates the effect of V-V title congruence on product sales (${\beta }_{13}$ = 0.148, p < 0.05). As shown in Figure 3a, the positive effect of V-V title congruence on SVA sales becomes stronger as the creation frequency increases, thus, H5 is supported. In addition, creation frequency negatively moderates the effect of V-P title congruence on product sales (${\beta }_{14}$ = −0.501, p < 0.01), Figure 3b shows that the negative effect of V-P title congruence on sale performance is stronger as the creation frequency increase, which support H6.

Model 9 shows that product brand positively moderates the effect of V-V title congruence on product sales (${\beta }_{15}$ = 0.702, p < 0.01). In Figure 4a, compared to unbranded products, V-V title congruence has a stronger positive effect on sales performance when promoting branded products. Model 10 verifies that product brand positively moderates the negative effect of V-P title congruence on product sales (${\beta }_{16}$ = 1.092, p < 0.01). As shown in Figure 4b, when promoting branded products in SVAs, the negative effect of V-P title congruence on sales performance is weaker compared to unbranded products. Thus, H7 and H8 are supported.

In summary, this study tested eight hypotheses regarding the effects of SVA title congruence on sales performance through the mediating role of sociability, as well as the moderating role of creation frequency and product brand.

Table 8. Results of hypotheses testing.

Hypothesis		Result
Hypothesis 1	V-V title congruence has a positive effect on SVA sociability.	Supported
Hypothesis 2	V-P title congruence has a negative effect on SVA sociability.	Supported
Hypothesis 3	SVA sociability has a positive effect on sales performance.	Supported
Hypothesis 4	SVA sociability mediates the effects of title congruence on sales performance.	Supported
Hypothesis 5	Creation frequency positively moderates the positive effect of V-V title congruence on SVA sales performance.	Supported
Hypothesis 6	Creation frequency negatively moderates the negative effect of V-P title congruence on SVA sales performance.	Supported
Hypothesis 7	Product brand positively moderates the positive effect of V-V title congruence on SVA sales performance.	Supported
Hypothesis 8	Product brand positively moderates the negative effect of V-P title congruence on SVA sales performance.	Supported

presents the summary of all hypothesis test results.

5.2. Robust Tests

In this section, we performed three robust checks of our research model. First, we applied the Heckman two-stage test to control for the potential self-selection bias based on observed variables. Second, we used the alternative measurement of V-V title congruence to test our research model. Finally, we used the alternative measurement of sociability. These test results that support our findings are robust.

5.2.1. Heckman Two-Stage Test

Self-bias occurs if creators differ systematically from content creation strategies. In our study, the decision of the creator whether to post the SVA or not is endogenous. Because the creation decision simultaneously affects the creator’s income and future creation strategies. In this case, the Heckman two-stage test is the effective method to correct the bias according to prior research [82,83].

The first step is to estimate the probability of an anchor posting the short video ad by using the probit model. We choose the creator’s characteristics (fan number, expertise) and video ad characteristics (viewership, likes, comments, share, duration, posting days, day of week, time of day, title length, and tag length) as the variables (See

Table A1. Results of Selection model.

Variables	Selection Model
	Propensity to Post Short Video Advertise
Creator fans	−0.729 ***
Creator expertise	0.258 ***
Video viewership	−0.194 ***
Video likes	−0.175 ***
Video comments	−0.200 ***
Video shares	0.157 ***
Video length	0.125 ***
Video age	0.068 ***
Video day of week	−0.015
Video time of day	0.001
Video title length	0.097 ***
Video tag number	0.414 ***
Month fixed effect	Yes
Creator fixed effect	Yes
Sector fixed effect	Yes

Standard errors in parentheses: *** p < 0.01.

in Appendix A) that may affect a creator’s propensity to post SVAs.

The second step is to add the inverse Mills ratio (IMR) calculated by the probit model in the observation model to obtain unbiased results.

Table A2. Results of observation model.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
	Sale	Sociability	Sale	Sale	Sale	Sociability	Sale	Sale
V-V title congruence	0.325 ***	0.245 ***	0.234 ***	−0.091
	(0.056)	(0.014)	(0.056)	(0.071)
V-P title congruence					−0.166 **	−0.059 ***	−0.144 **	−0.835 ***
					(0.069)	(0.016)	(0.069)	(0.093)
Sociability			0.372 ***				0.377 ***
			(0.018)				(0.018)
Creation frequency				−0.306 ***				0.265 ***
				(0.058)				(0.068)
Creation frequency × V-V title congruence				0.252 ***
				(0.076)
Creation frequency × V-P title congruence								−0.624 ***
								(0.090)
Product brand				1.327 ***				1.018 ***
				(0.055)				(0.083)
Product brand × V-V title congruence				0.707 ***
				(0.083)
Product brand × V-P title congruence								1.039 ***
								(0.115)
Video day of week	0.037 *	−0.014 **	0.042 *	0.031	0.037 *	−0.013**	0.042 *	0.035
	(0.023)	(0.005)	(0.022)	(0.022)	(0.023)	(0.005)	(0.022)	(0.022)
Video time of day	−0.018	0.079 ***	−0.047 **	−0.024	−0.020	0.078 ***	−0.049 **	−0.034
	(0.024)	(0.006)	(0.024)	(0.023)	(0.024)	(0.006)	(0.024)	(0.023)
Video length	0.659 ***	0.614 ***	0.431 ***	0.483 ***	0.662 ***	0.618 ***	0.428 ***	0.492 ***
	(0.064)	(0.017)	(0.065)	(0.061)	(0.064)	(0.017)	(0.065)	(0.061)
Video age	0.239 ***	0.135 ***	0.189 ***	0.245 ***	0.240 ***	0.136 ***	0.189 ***	0.246 ***
	(0.017)	(0.004)	(0.017)	(0.016)	(0.017)	(0.004)	(0.017)	(0.016)
Video title length	0.130 ***	0.205 ***	0.054 **	0.124 ***	0.160 ***	0.224 ***	0.075 ***	0.171 ***
	(0.027)	(0.008)	(0.027)	(0.027)	(0.027)	(0.008)	(0.027)	(0.026)
Video tag number	0.237 ***	0.837 ***	−0.074 *	0.224 ***	0.248 ***	0.844 ***	−0.070 *	0.236 ***
	(0.038)	(0.010)	(0.040)	(0.036)	(0.038)	(0.011)	(0.040)	(0.036)
Product price	−0.979 ***	−0.025 ***	−0.970 ***	−0.993 ***	−0.980 ***	−0.025 ***	−0.971 ***	−0.990 ***
	(0.020)	(0.004)	(0.020)	(0.019)	(0.020)	(0.004)	(0.020)	(0.019)
Creator fans	−0.039	−1.464 ***	0.506 ***	−0.535 ***	−0.026	−1.474 ***	0.530 ***	−0.390 **
	(0.195)	(0.050)	(0.196)	(0.190)	(0.196)	(0.051)	(0.196)	(0.190)
Creator expertise	0.001	0.528 ***	−0.195 ***	0.030	−0.008	0.522 ***	−0.205 ***	0.025
	(0.027)	(0.008)	(0.028)	(0.027)	(0.027)	(0.008)	(0.028)	(0.027)
Creator sex	2.901 ***	9.008 ***	−0.449	4.966 ***	2.657 ***	8.858 ***	−0.684	4.619 ***
	(0.444)	(0.106)	(0.467)	(0.448)	(0.445)	(0.106)	(0.466)	(0.443)
IMR	0.977 ***	4.303 ***	−0.623 ***	0.951 ***	0.985 ***	4.310 ***	−0.641 ***	0.971 ***
	(0.059)	(0.023)	(0.094)	(0.057)	(0.059)	(0.023)	(0.093)	(0.057)
Month fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Creator fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Sector fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	8.635 ***	12.267 ***	4.073	12.609 ***	8.685 ***	12.508 ***	3.967	10.946 ***
	(2.515)	(0.638)	(2.507)	(2.458)	(2.522)	(0.641)	(2.515)	(2.457)
N	57,192	57,192	57,192	57,192	57,192	57,192	57,192	57,192
R2	0.551	0.905	0.555	0.581	0.551	0.905	0.555	0.581
adj. R2	0.549	0.905	0.552	0.578	0.549	0.904	0.552	0.579

Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01.

shows that the observation model results (shown in Appendix A) are consistent with our results and support the robustness of our research model.

5.2.2. Alternative Measurement of V-V Title Congruence

We also used the alternative measurement of V-V title congruence to check the robustness of the main results. We expanded the day intervals of V-V title congruence between the current and previous SVAs from the original 14 days posted by the creators to the last 7, 21, and 28 days. The results are shown in

Table A3. Results of alternative measurement of V-V congruence.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8	Model 9	Model 10	Model 11	Model 12
	Sale	Sociability	Sale	Sale	Sale	Sociability	Sale	Sale	Sale	Sociability	Sale	Sale
V-V title congruence (7)	0.208 ***	0.329 ***	0.115 **	−0.155 **
	(0.049)	(0.020)	(0.049)	(0.063)
V-V title congruence (21)					0.449 ***	0.432 ***	0.329 ***	−0.075
					(0.061)	(0.024)	(0.061)	(0.075)
V-V title congruence (28)									0.536 ***	0.490 ***	0.399 ***	0.006
									(0.065)	(0.026)	(0.065)	(0.079)
Sociability			0.283 ***				0.280 ***				0.279 ***
			(0.011)				(0.011)				(0.011)
Creation frequency				−0.277 ***				−0.318 ***				−0.390 ***
				(0.044)				(0.046)				(0.046)
Creation frequency × V-V title congruence (7)				0.160 ***
				(0.054)
Creation frequency × V-V title congruence (21)								0.238 ***
								(0.061)
Creation frequency × V-V title congruence (28)												0.361 ***
												(0.063)
Product brand				1.362 ***				1.269 ***				1.266 ***
				(0.055)				(0.056)				(0.056)
Product brand × V-V title congruence (7)				0.596 ***
				(0.077)
Product brand × V-V title congruence (21)								0.835 ***
								(0.087)
Product brand × V-V title congruence (28)												0.860 ***
												(0.090)
Video day of week	0.035	−0.022 **	0.042 *	0.031	0.034	−0.023 **	0.041 *	0.031	0.036	−0.022 **	0.042 *	0.032
	(0.023)	(0.009)	(0.022)	(0.022)	(0.023)	(0.009)	(0.022)	(0.022)	(0.023)	(0.009)	(0.022)	(0.022)
Video time of day	−0.030	0.028 ***	−0.038	−0.034	−0.029	0.028 ***	−0.037	−0.034	−0.030	0.027 ***	−0.038	−0.034
	(0.024)	(0.010)	(0.024)	(0.023)	(0.024)	(0.010)	(0.024)	(0.023)	(0.024)	(0.010)	(0.024)	(0.023)
Video length	0.707 ***	0.801 ***	0.476 ***	0.521 ***	0.704 ***	0.804 ***	0.475 ***	0.523 ***	0.702 ***	0.803 ***	0.475 ***	0.522 ***
	(0.064)	(0.030)	(0.064)	(0.062)	(0.064)	(0.030)	(0.064)	(0.061)	(0.064)	(0.030)	(0.064)	(0.061)
Video age	0.227 ***	0.084 ***	0.203 ***	0.232 ***	0.230 ***	0.088 ***	0.205 ***	0.235 ***	0.228 ***	0.086 ***	0.204 ***	0.234 ***
	(0.017)	(0.006)	(0.017)	(0.016)	(0.017)	(0.006)	(0.017)	(0.016)	(0.017)	(0.006)	(0.017)	(0.016)
Video title length	0.100 ***	0.031 ***	0.091 ***	0.095 ***	0.084 ***	0.023 **	0.077 ***	0.084 ***	0.077 ***	0.017	0.071 ***	0.074 ***
	(0.027)	(0.012)	(0.027)	(0.026)	(0.027)	(0.012)	(0.027)	(0.027)	(0.027)	(0.012)	(0.027)	(0.027)
Video tag number	0.063 *	0.065 ***	0.045	0.056	0.060 *	0.063 ***	0.043	0.051	0.059	0.062 ***	0.042	0.052
	(0.036)	(0.015)	(0.036)	(0.035)	(0.036)	(0.015)	(0.036)	(0.035)	(0.036)	(0.015)	(0.036)	(0.035)
Product price	−0.986 ***		−0.970 ***	−0.999 ***	−0.985 ***		−0.970 ***	−1.000 ***	−0.986 ***		−0.971 ***	−1.000 ***
	(0.020)		(0.020)	(0.019)	(0.020)		(0.020)	(0.019)	(0.020)		(0.020)	(0.019)
Creator fans	0.274	−0.085	0.293	−0.197	0.275	−0.101	0.297	−0.207	0.280	−0.098	0.301	−0.214
	(0.195)	(0.090)	(0.194)	(0.189)	(0.195)	(0.090)	(0.194)	(0.189)	(0.195)	(0.090)	(0.194)	(0.189)
Creator expertise	−0.175 ***	−0.200 ***	−0.118 ***	−0.125 ***	−0.169 ***	−0.199 ***	−0.113 ***	−0.121 ***	−0.169 ***	−0.199 ***	−0.113 ***	−0.122 ***
	(0.026)	(0.011)	(0.026)	(0.026)	(0.026)	(0.011)	(0.026)	(0.026)	(0.026)	(0.011)	(0.026)	(0.026)
Creator sex	1.909 ***	5.168 ***	0.485	3.996 ***	2.050 ***	5.241 ***	0.620	4.138 ***	2.070 ***	5.249 ***	0.644	4.091 ***
	(0.442)	(0.171)	(0.443)	(0.444)	(0.442)	(0.171)	(0.444)	(0.446)	(0.443)	(0.171)	(0.444)	(0.445)
Month fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Creator fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Sector fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	6.789 ***	3.431 ***	5.807 **	10.259 ***	6.666 ***	3.582 ***	5.652 **	10.299 ***	6.602 ***	3.550 ***	5.601 **	10.398 ***
	(2.522)	(1.163)	(2.502)	(2.454)	(2.521)	(1.159)	(2.502)	(2.453)	(2.519)	(1.159)	(2.501)	(2.453)
N	57,192	57,192	57,192	57,192	57,192	57,192	57,192	57,192	57,192	57,192	57,192	57,192
R2	0.549	0.745	0.555	0.578	0.549	0.745	0.555	0.579	0.549	0.746	0.555	0.579
adj. R2	0.546	0.743	0.552	0.576	0.547	0.744	0.552	0.576	0.547	0.744	0.552	0.576

Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01.

(Appendix B), which are consistent with the main findings.

5.2.3. Alternative Measurement of Sociability

Referring to prior research [20,84], we applied the alternative measurement of sociability, using the total sum of likes, comments and shares divided by the number of the creator followers, we obtained a relatively standardized metric that eliminates the influence of followers’ number [20]. We used the methods proposed by Baron and Kenny [81] to re-estimate the research model with the sociability ratio. The results of

Table A4. Results of alternative measurement of mediator variable.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	Sale	Sociability Ratio	Sale	Sale	Sociability Ratio	Sale
V-V title congruence	0.356 ***	0.001 ***	0.346 ***
	(0.056)	(0.000)	(0.057)
V-V title congruence				−0.201 ***	−0.000 ***	−0.199 ***
				(0.069)	(0.000)	(0.069)
Sociability ratio			16.831 ***			17.480 ***
			(4.520)			(4.637)
Video day of week	0.035	−0.000 *	0.036	0.035	−0.000	0.037
	(0.023)	(0.000)	(0.023)	(0.023)	(0.000)	(0.023)
Video time of day	−0.030	−0.000 ***	−0.028	−0.031	−0.000 **	−0.030
	(0.024)	(0.000)	(0.024)	(0.024)	(0.000)	(0.024)
Video length	0.706 ***	0.001 ***	0.682 ***	0.709 ***	0.003 ***	0.683 ***
	(0.064)	(0.000)	(0.064)	(0.064)	(0.000)	(0.064)
Video age	0.228 ***	0.000 ***	0.227 ***	0.229 ***	0.000 ***	0.227 ***
	(0.017)	(0.000)	(0.017)	(0.017)	(0.000)	(0.017)
Video title length	0.090 ***	0.000	0.090 ***	0.123 ***	0.000	0.122 ***
	(0.027)	(0.000)	(0.027)	(0.027)	(0.000)	(0.027)
Video tag number	0.061 *	−0.000	0.062 *	0.073 **	0.000	0.073 **
	(0.036)	(0.000)	(0.036)	(0.036)	(0.000)	(0.036)
Product price	−0.986 ***		−0.984 ***	−0.988 ***		−0.986 ***
	(0.020)		(0.020)	(0.020)		(0.020)
Creator fans	0.282	−0.004 ***	0.342 *	0.304	−0.000 ***	0.367 *
	(0.195)	(0.001)	(0.197)	(0.196)	(0.000)	(0.198)
Creator expertise	−0.171 ***	−0.000 **	−0.169 ***	−0.183 ***	−0.000 ***	−0.180 ***
	(0.026)	(0.000)	(0.026)	(0.026)	(0.000)	(0.026)
Creator sex	1.995 ***	0.006 ***	1.893 ***	1.710 ***	0.000	1.610 ***
	(0.442)	(0.001)	(0.443)	(0.442)	(0.000)	(0.444)
Month fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Creator fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Sector fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Constant	6.635 ***	0.046 ***	5.857 **	6.612 ***	0.004 ***	5.795 **
	(2.522)	(0.007)	(2.536)	(2.530)	(0.000)	(2.544)
N	57,192	57,192	57,192	57,192	57,192	57,192
R2	0.549	0.309	0.549	0.549	0.078	0.549
adj. R2	0.546	0.305	0.547	0.546	0.077	0.546

Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01.

are consistent with those of the original model (shown in Appendix C). Thus, our findings are robust across different measurements of sociability.

5.3. Additional Analysis

To extend our primary findings, we conducted a supplementary analysis to investigate how textual features of video titles affect product sales, aiming to develop strategic recommendations for content creators. We categorized video titles into two distinct types: product-independent and product-related.

We initially randomly selected 500 titles for manual classification. Using Unsloth [85], we fine-tuned the Qwen2.5-7B-Instruct model (one of the most renowned open-source LLMs) based on these manually classified results. The fine-tuned model was then employed to classify all video titles. To evaluate the model’s performance, we randomly selected another 500 titles for manual classification (coded by two human coders) as a test set, and the model achieved an accuracy of 92.85%. In addition, we also calculated the inter-rater reliability (IRR) metrics to evaluate the agreement between human coders and LLM according to prior research [86]. Following Chew et al.’s [87] equations, we calculated the IRR between manual classification and LLM and the result is 93.08%, which meets our experimental requirements.

Table 9. Examples of two short video types.

Short Video Type	Example
Type 1 product-independent	The days don’t panic; we grow old together. Feel free to come to Sister Zhong’s live streaming tonight for peanuts.
Type 2 product-related	9.9 RMB to make a photo frame for yourself with just a cell phone photo ~ love it so much! The days are getting cooler, and wearing shoes is good for sweaty feet. Foot odors dare not take off shoes, must try this shoe and sock freshness spray. Spray it to freshen up all day long #shoe and sock deodorizing spray.

presents examples of the two types of short video titles. Product-independent short video titles do not contain information about the product-related features; while the product-related short video titles directly contain information about the product price, features, functions, and so on. These differences in textual features may lead to different marketing effects. Therefore, we extracted three textual features of the title, including sentiment, first-person pronoun, and vividness, to explore their effects on product sales (The measurement process of sentiment, first-person pronoun, and vividness is stated in Appendix D). We applied the t-test method to compare the differences between the above three textual features of the two short video types.

Table 10. T-test results of video title characteristics.

Variable	Type 1 (N = 4947)	Type 2 (N = 52,245)	Mean Difference
Title sentiment	0.029	0.022	0.007 ***
Title first-person	0.007	0.003	0.003 ***
Title vividness	0.522	0.847	−0.326 ***

Standard errors in parentheses: *** p < 0.01.

indicates the significant differences between the two types of short video titles on the above features.

On this basis, we further discussed the effects of these textual features on product sales. As shown in the result of Model 12 in

Table 11. Results of effects of video title textual characteristics.

Sales	Model 12	Model 13	Model 14
Title sentiment	−1.067 ***
	(0.328)
Type	−0.110 **	−0.128 ***	−0.141 ***
	(0.046)	(0.045)	(0.052)
Type × Title sentiment	0.678 *
	(0.395)
Title first-person		−4.335 ***
		(1.587)
Type × Title first-person		8.515 ***
		(1.931)
Title vividness			−0.034
			(0.046)
Type × Title vividness			0.111 **
			(0.046)
Month fixed effect	Yes	Yes	Yes
Creator fixed effect	Yes	Yes	Yes
Sector fixed effect	Yes	Yes	Yes
Controls	Yes	Yes	Yes
Cons	7.328 ***	7.156 ***	6.781 ***
	(2.520)	(2.518)	(2.517)
N	57,192	57,192	57,192
R2	0.549	0.549	0.549
adj. R2	0.546	0.546	0.546

Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01.

, title sentiment negatively affects product sales (${\beta }_{sentiment}$ = −1.067, p < 0.01). Compared to positive titles, negative titles indicate SVAs contain interesting and curious content and are more likely to obtain more attention [88]. And this relationship is positively moderated by video type (${\beta }_{type\times sentiment}$ = 0.678, p < 0.1). This result indicates that the negative effect of title sentiment is more significant in product-independent SVAs.

The results of Model 13 indicate that first-person pronouns negatively affect product sales (${\beta }_{first-person}$ = −4.355, p < 0.01). While the first-person pronouns in short video ad titles contribute to emotional connection and empathy senses, this narrative perspective is self-centered [89], and the usage of first-person narrative perspectives tends to be perceived as subjective, which is less credible, and likely to exaggerate and distorts the facts for their gain [90,91]. Video type positively moderates the negative relationship (${\beta }_{type\times first-person}$ = 8.515, p < 0.01), which indicates that the negative effect of first-person pronouns is weakened in those product-dependent SVAs.

Moreover, the results of Model 14 show that vividness negatively influences product sales (${\beta }_{vividness}$ = −0.034, p > 0.1), and the negative relationship is moderated by video type (${\beta }_{type\times vividness}$ = 0.111, p < 0.05). These findings suggest that titles containing excessive suspense or storyline may increase the information uncertainty and cognitive cost of consumers, and they may consider these as inappropriate marketing strategies [21].

6. Discussion

6.1. Key Findings

How to effectively design short video advertisements to promote product sales is the main question for creators. Based on the Signaling theoretical framework, this study explores the effects of title congruence characteristics of SVA on sociability and sales performance, as well as the heterogeneous effects from the perspectives of creators and product characteristics. Our findings provide insights into content creation strategies in the short video context.

First, this study found that V-V title congruence positively affects SVA sociability [1]. This suggests that consistent SVA content helps creators develop their creation image [92]. Meanwhile, it helps creators differentiate themselves from others on the platform, reducing interference from others [40,43], fostering familiarity and trust, and thereby increasing interaction willingness. For creators, maintaining a certain level of V-V title congruence can help them establish a personal creation image on the platform, enhance their distinctiveness, and ultimately improve SVA sales performance.

Second, unlike the former, this study found that V-P title congruence negatively affects SVA sociability. This conclusion differs from previous studies on product consistency, which may be due to differences in the attributes of the research platforms. Previous studies were primarily conducted on online e-commerce platforms such as Taobao and Amazon, which are e-commerce platforms designed to facilitate product purchases for consumers. These studies found that cues related to product congruence help foster content trust, which in turn influences product sales [41,93]. However, on SVA platforms centered on social entertainment [50], SVAs with high V-P title congruence highlight product attributes in their contents, making it easier to identify direct product promotion, triggering cognitive resistance, and leading to advertising avoidance [52]. Therefore, for creators on SVA platforms, it is advisable to avoid presenting product information in the SVA titles and instead adopt indirect or more entertaining title strategies to reduce resistance and avoidance of SVAs.

Third, this study found that sociability serves as a partial mediator between title congruence and sales performance, indicating that part of the effect of title congruence on sales remains direct. The sociability of SVAs positively influences SVA sales performance, which is consistent with previous research findings [13]. Sociability reflects trust and popularity of content, which further enhances the promoted product’s visibility and purchase likelihood [94]. This finding emphasizes the mediating role of SVA sociability in sales performance. Moreover, it suggests that some viewers may directly purchase products without performing any social interactions, enhancing sociability through interaction, trust, and ultimately convert into buyers. Therefore, creators could design SVA titles that capture audience attention and encourage interaction, as strengthening sociability remains a practically effective way to achieve the goal of product sales on SVA platforms.

Fourth, our results show that creation frequency positively moderates the relationship between V-V title congruence and product sales. When creators consistently create and publish SVAs with consistent styles, it enhances personal image perception and improves SVA sales performance. Creation frequency negatively moderates the effect of V-P title congruence on sales performance. That is, when creators frequently publish product-oriented high-congruence SVAs (high V-P title congruence), this may exacerbate resistance and avoidance caused by direct advertisements, thereby strengthening the negative effect of V-P title congruence. Additionally, this study found that the product brand significantly moderates the effects of V-V and V-P title congruence on sales performance. Product brands can enhance content trust [66], thereby weakening the advertisement avoidance triggered by high V-P title congruence and strengthening the creator’s personal image formed by V-V title congruence.

Finally, through additional analysis, this study explored and found the significant effects of title textual characteristics (sentiment, first-person pronouns, and vividness) on product sales, and this relationship is moderated by whether the SVA highlights product information [14]. This result indicates that creators should consider the degree of matching between the SVA content type and title expression when creating content to avoid unnecessary avoidance due to the title text of the SVA.

6.2. Theoretical Contribution

Our study contributes to existing literature in the following aspects. First, previous research has mostly focused on the effects of short video characteristics on consumer engagement or purchasing behavior, with little attention paid to the congruence characteristics of the creator’s SVAs. From the perspective of creators, this study proposes and verifies the influence of title congruence of SVAs as a creation strategy, revealing its impact on advertising effectiveness and mechanism.

Second, this paper deepens our understanding of congruence characteristics. Previous studies have shown that congruence can enhance content trust and product purchase preference [32,33], while this study found that different congruence features may have varying effects on short video platforms: video-video title congruence of the same creator positively influences SVA performance, while video-product title congruence may lead to advertisement avoidance and resistance [1,51]. Dang, Saravade [95] finds that product type moderates the effect between trending content elements in SVAs and viewer engagement. Consistent with their findings, our conclusion reveals that the presentation of SVAs needs to consider content preferences, especially on short video platforms primarily aimed at social entertainment.

Third, this study reveals that the mechanism between title congruence and SVA sales performance is sociability, enriching the application of Signaling theory [96]. This finding expands the theoretical boundaries of Signaling theory in the context of short video research, indicating that sociability is not only an interaction indicator for social performance but may also become a linking mechanism driving factor for SVA sales performance.

Fourth, this study proposes creator characteristics and product characteristics as moderating variables and identifies their boundary conditions [39,69]. By introducing creation frequency and product brand as moderating variables, this study finds that the influence of title congruence on sales performance across different levels of creation frequency and product brand further emphasizes the adaptive boundaries of congruence-based creation strategies in short video platforms. This provides a foundation for future research to explore other boundary conditions, such as platform or consumer features.

Finally, this study identifies the impact of title textual features on SVA sales performance. The study finds that sentiment, first-person pronouns, and vividness in titles have a negative effect on sales performance, indicating that SVA textual features have different effects on its sales performance. This finding provides a basis for future research exploring the interaction effects of SVA text and other features.

6.3. Practical Implications

With the development of short video platforms, short video ads have increasingly become the main product-selling channel for enterprises and creators. From the perspective of creators, this study investigates and demonstrates that SVA title congruence significantly influences sales performance through the mediating role of sociability, as well as the moderating effects of creation frequency and product brand based on signaling theory. Our findings reinforce the validity of our proposed research model and offer the following practical implications for SVAs platforms:

First, creators need to maintain a reasonable level of congruence in their SVA content to establish a recognizable personal style and avoid content fatigue. This study found that creators who maintain a certain level of congruence in video-video titles can help them establish a stable creation image on the platforms and gain more recognition. However, they should also avoid content fatigue caused by excessive repetition, especially at high creation frequency, which may lead to content avoidance. Therefore, creators can tailor their congruence creation strategies, such as reusing keywords and maintaining a consistent expression style, while innovating content themes and presentation formats to enhance consumer stickiness. Recent research has shown that content congruence significantly enhances viewer engagement in SVAs, reinforcing the importance of strategic [1].

Second, SVAs should avoid direct sale promotion, ensuring that product information is conveyed while enhancing the entertainment value and indirect nature of the SVA content. The findings of this study reveal that when video titles are highly consistent with product titles, the advertising intent is more likely to be recognized, which in turn triggers resistance, particularly on short video platforms that focus on social entertainment. Therefore, creators can adopt indirect product placement methods in the SVA creation, presenting and promoting products through storytelling, emotional appeal, or contextualization to enhance the acceptability and appeal of SVAs.

Third, introducing brand information can be an effective strategy to enhance content trust. For branded products that have already established consumer recognition and emotional attachment, their presence in SVAs can weaken advertising resistance caused by product-oriented content. Therefore, creators can promote branded products when selecting suitable products for advertising and moderately highlight brand elements in SVA contents to enhance content trust in product quality [97].

Fourth, creators should also tailor their title creation strategies to different SVA types to avoid mismatches between type and content. This study finds that sentiment, first-person pronouns, and vividness of SVA titles may weaken sales performance. Hence, when creating SVAs that do not include or indirectly mention the product, creators can try using titles with suspense or exploratory; in SVAs that include products and are aimed at product promotion, titles may focus more on a clear expression style.

Finally, although this study focuses on the Douyin platform, many insights are also transferable to its international version, TikTok, which shares comparable short video formats, integrated e-commerce functions, and algorithmic recommendations [23]. Given these structural and operational similarities, the observed relationships and mechanisms between the SVA title congruence and SVA effectiveness are expected to be suitable across TikTok and other platforms.

6.4. Limitations and Future Work

Although this study is based on real data collected from the Douyin platform, it still has the following research limitations:

First, the sample in this study focuses on creators with follower numbers larger than 100,000; this group constitutes the main content and commercial revenue on Douyin. Therefore, while the findings may not be fully generalized to micro-creators, they still provide insights into the mechanisms that drive commercial effectiveness on SVAs platforms. Future research could extend this research model by collecting data from creators with smaller follower bases to examine whether the observed relationships across different creator segments and to compare potential variations with different audience scales.

Second, the study focuses on the title of SVAs and lacks a comprehensive exploration of video content features. This study primarily focuses on the effect of title congruence (V-V and V-P title congruence) on SVA sales performance, without further analyzing how video content (such as visual style, characteristics, etc.) influences SVA performance. Future research could integrate multi-modal content analysis techniques, including video, image, and audio, to examine the overall congruence of SVA contents from a holistic perspective, thereby delving deeper into its influence and mechanism in stimulating and enhancing product sales in the context of SVAs.

Third, the selection of moderator variables is limited and fails to cover multiple heterogeneous characteristics that influence product sales. This study only selected creation frequency and product brand as moderating variables, but product sales may also be influenced by creator features (i.e., gender, verification, and followers); product category (i.e., search-based and experiential); and individual consumer features (i.e., advertising sensitivity and purchase motivation). Future research could expand the scope of the moderating variables to reveal the conditions under which SVAs are effective.

Finally, the research data of this study is sourced from the Douyin platform in China. Considering that short video platforms in different countries or regions may vary in terms of consumer usage habits, content preferences, and cultural values, the applicability of the research conclusions across different cultural and national contexts requires further examination. Cultural and regulatory distinctions may lead to variations in viewer engagement and response to title congruence signals. Future research could conduct comparative research on other mainstream platforms, such as TikTok, YouTube, and Instagram, to examine whether the observed relationships exist across diverse cultural and platform environments. Such investigations could enhance the external validity and generalizability of the findings and provide a deeper understanding of how platform and cultural attributes shape SVA effectiveness.