Empirical Research on the Influencing Factors and Causal Relationships of Enterprise Positive Topic Heat on Online Social Platforms

Abstract

Purpose: As online social platforms have increasingly become a key arena for enterprises to disseminate information, enhancing the heat of enterprise positive topic on such platforms plays a critical role in improving brand value and achieving high-quality development. Against this backdrop, it is critically important for enterprises to precisely identify the key factors influencing positive topic heat and to reveal the causal relationships between these factors and topic heat. Method: This paper focuses on the influencing factors and causal pathways of enterprise positive topic heat on online social platforms. To achieve this, this paper collects data on enterprise positive topic publicity from Sina Weibo, taking topic heat as the dependent variable, and uses regression analysis to examine its causal relationships with independent variables such as topic host attribute and activity level. Subsequently, a cost performance indicator was constructed and quantitatively evaluated to assess the effectiveness of different host attributes based on both performance and cost. Result: This research shows topic hosts play an active role in the topic spreading process, but some of them an exhibit asymmetric ability to increase topic heat compared to their level. The relationship between the topic heat contribution and host’s activity level changes with the change of the host attribute. In addition, the host’s cost performance for each attribute is significant different. By empirically analyzing the influencing factors of positive topic heat and revealing their underlying mechanisms, enterprises can identify the key factors for enhancing positive topic heat on online social platforms and optimize decisions such as the selection of hosts for related topic promotion.

1. Introduction

1.1. Research Background

With continuous advancements in communication technology, the number of China’s Internet users in the new media era has reached 1.108 billion, with an Internet penetration rate of 78.6% (CNNIC, 2025) [1]. By the end of 2021, the cumulative total of social communication applications had reached 209,000 (MIIT, 2022). The rapid proliferation of online social platforms has significantly transformed the dynamics of information dissemination, and the communication model between enterprises and the public has also undergone profound changes. Online social platforms have gradually become a vital arena for enterprises to disseminate information, build brand image, and engage with the public.

Positive-oriented topics released by enterprises are more likely to resonate with users and stimulate interactive behaviors. In the context of online social platforms, enterprises frequently adopt strategies such as official accounts and designated topic hosts to disseminate positive topic and enhance their visibility. Among various indicators, topic heat—reflecting both the breadth of content dissemination and the depth of user engagement, has emerged as a core metric for evaluating the effectiveness of positive topic publicity. Enhancing positive topic heat not only broadens an enterprise’s social media presence but also strengthens public trust, contributes to the construction of a favorable corporate image, enhances opinion leadership, and elevates market reputation. Against this backdrop, it is especially critical for enterprises to improve the visibility and heat of positive topics, ensuring that such content enters the public’s view across the digital space. Given the necessity of enhancing enterprises’ positive topic heat on online social platforms, accurately identifying positive topic heat influencing factors and revealing the causal relationships between these factors and topic heat have become urgent in the context of new media era.

1.2. Literature Review

Focusing on the influencing factors of enterprise topic heat on online social platforms, a review of the literature reveals that existing studies primarily explore this issue from two perspectives: information content and information subject. The specific findings are as follows.

From the perspective of information content, based on the background of the new media era, academics have studied the influencing factors of enterprise topic heat on online social platforms from the aspects of positive topics and negative topics. For positive topics, research has focused on how emotional positivity, sentiment orientation, and language style affect topic heat. For example, Oliveira et al. (2022) [2] emphasized the impact of emotional valence on topic heat and found that topics with higher positive sentiment tend to attract more user attention—indicating that emotionally positive content can effectively enhance topic heat. Similarly, Ma et al. (2022) [3], based on enterprise topics on the Weibo platform, found that enterprise topics with stronger positive emotional tones are more likely to generate user engagement, such as likes and shares, thereby promoting wider dissemination. Deng et al. (2021) [4] further examined the relationship between language style and topic heat using interaction data and concluded that more positively framed language significantly boosts topic heat. For negative topics, researchers have primarily examined how topic content type, narrative strategy, and emotional tone shape topic heat. Kleer & Kunz (2023) [5] found that, during crisis periods, posting topics that directly address the negative events themselves can actually enhance topic heat. Their study further indicated that different content types and narrative formats play a moderating role in shaping communication effectiveness, suggesting that enterprises can improve crisis communication outcomes by optimizing the structure of their social media content. Yu et al. (2022) [6], through analysis of tweets related to public controversies, found that negatively toned topics tend to generate higher topic heat than their positive counterparts. In scenarios involving crisis response or customer complaints, enterprise posts that contain a moderate level of negative emotion may facilitate broader dissemination. Overall, these studies underscore the central role of information content in driving topic heat and suggest that enterprises should carefully balance emotional tone and communication effectiveness when planning social media content strategies. Content design should be tailored to the communication objectives and context to maximize dissemination outcomes (Alalwan, 2018) [7].

Another perspective is information subject. With the rise of online social platforms, official platform accounts have become a direct and influential medium for enterprise publicity, enabling the delivery of promotional content directly to target audiences. Existing studies have examined how topic heat is influenced by two types of publicity approaches: using enterprise-owned accounts, and leveraging external accounts. In the context of enterprise-owned accounts, scholars have investigated various social media platforms and identified several key influencing factors, including interactive elements, visual components, timing of post release, and narrative style. For instance, focusing on platforms such as Twitter and Instagram, Yoo (2024) [8] found that excessive use of hashtags may reduce user participation in topic discussions, while a balanced combination of hashtags and textual descriptions can significantly increase topic heat. This suggests that enterprises should use hashtags judiciously and align them appropriately with narrative content to optimize engagement. Andariesta & Wasesa (2023) [9], based on a predictive model using Weibo data, demonstrated that specific topic features—such as the inclusion of hyperlinks, hashtags, and images—have a significant positive effect on topic heat. Regarding Sina Weibo, Li (2023) [10] optimized enterprise account publicity strategies by examining posting time and content type. The study suggests that enterprises should post promotional content in the morning or evening and increase the proportion of incentive-based posts, such as giveaways and repost campaigns, to enhance topic heat. For short-video platforms like TikTok, studies primarily include case analyses (Pieter et al., 2021) [11], empirical investigations (Han, 2020) [12], and strategic approaches (Guo, 2023) [13]. These studies identify key strategies for maximizing topic heat in official account operations, such as fragmented storytelling, scene-oriented narratives, and emotionally expressive content, complemented by innovative and eye-catching promotional techniques.

On the WeChat platform, public accounts serve as the primary medium for official publicity. Enterprises are advised to encourage professional teams to engage in real-time interactions with users to enhance user stickiness, thereby expanding the influence of their public accounts through increased topic heat (Xu et al., 2023) [14]. Regarding publicity through external accounts, existing research emphasizes leveraging the influence of opinion leaders (such as topic hosts) on social platforms. For example, Da & Alturas (2018) [15] confirmed that enterprises should leverage opinion leaders as spokespersons and suggest selecting those with strong audience rapport and a deep understanding of the promotional content to maximize their positive impact on topic heat. Similarly, Van et al. (2011) [16] highlighted that opinion leaders tend to be more innovative in designing publicity content, and enterprises should utilize their celebrity effect to amplify topic heat. Additionally, governments can strategically cultivate opinion leaders on platforms like Weibo to promote authoritative and positive narratives, thereby enhancing topic heat and guiding public discourse (Chen & Gao, 2023) [17].

In summary, the existing literature has explored the factors influencing enterprise topic heat on online social platforms from multiple perspectives, providing a solid theoretical foundation and practical guidance for enhancing positive topic heat. However, there remain several gaps in this field that warrant further investigation. (1) Most studies examine the impact of information content or information subject—such as emotional expression, visual form, and posting time—on topic heat [2,3,4,9,10]. Yet, few adopt an empirical analytical approach. The majority of prior research focuses on strategic or content-based analysis, lacking empirical validation of the causal relationships between influencing factors and topic heat [7,8,10,14,15]. (2) Existing studies lack detailed analysis of the impact paths and mechanisms of various influencing factors. For example, why positive topic content, well-designed hashtags, and topic hosts are more likely to enhance topic heat remains underexplored, with little systematic theoretical framework or mechanism validation [2,3,4,8,9,15,16]. (3) Although some studies have acknowledged the positive role of topic hosts in enhancing topic heat [15,16,17], few have quantitatively evaluated the relationship between hosting costs and topic heat contribution, and rarely has the concept of “cost performance” been incorporated into the analytical framework for positive topic heat enhancement. This paper innovatively constructs a cost performance index for topic hosts by integrating both performance and cost dimensions. By comprehensively considering resource input and communication outcomes, this paper accurately identifies the most cost-effective host types and provides a quantitative basis from a cost performance perspective to support the enhancement of positive topic heat.

Focusing on the key issue of enhancing enterprise positive topic heat on online social platforms, this paper takes enterprise positive topic publicity on Sina Weibo as an example, using topic heat as the core dependent variable. By constructing empirical analysis models, it examines the contributions of topic hosts, host attributes, and host activity levels to topic heat, and introduces a cost-performance indicator for hosts with quantitative characterization. This research aims to assist enterprises in identifying the key factors for boosting positive topic heat on online social platforms and in revealing the causal pathways between these factors and topic heat.

2. Materials and Methods

2.1. Research Hypotheses

The topic mechanism on online social platforms has become an important channel for enterprises to conduct positive publicity and build brand image. As information dissemination becomes increasingly dependent on algorithmic recommendation and user interactions, digital influence has emerged as a critical mechanism shaping topic visibility and propagation effectiveness. Digital influence refers to an individual’s capacity to affect the opinions, behaviors, or decisions of others within digital environments (Rachmad, 2023) [18]. Topic hosts are not only content publishers but also central nodes in the chain of information dissemination. Their digital influence—manifested through follower networks, engagement activity, and network position—can significantly affect how widely and rapidly a topic spreads. Topic hosts, as designated content managers, play a central role in organizing, publishing, and maintaining topic-related content. On one hand, users with digital influence can guide content delivery and encourage user interaction, thereby expanding both the reach and engagement of information (Zak & Hasprova, 2020) [19]. On the other hand, the identity attributes and influence level of the hosts often attract the attention and participation of platform users, which in turn increases the discussion and stickiness of users on topics (De et al., 2017) [20]. In addition, topic hosts have the authority to edit topic descriptions, highlight selected posts, and pin content. These functions help improve the structure and clarity of the topic page, supporting the accumulation of topic heat.

Based on this, this paper sets topic host as one of the core explanatory variables and proposes Hypothesis 1:

Hypothesis 1.

The participation of topic host contributes to the enhancement of enterprise positive topic heat.

As key nodes in the dissemination of information on online social platforms, topic hosts possess distinct attribute characteristics that significantly shape both the formation and diffusion of topic heat. These attributes—such as account identity, institutional affiliation, and communication style—determine the host’s influence and positioning in the topic propagation process. Existing studies point out that hosts with strong institutional backgrounds (e.g., government agencies or media accounts) tend to exhibit high levels of perceived authority. Such accounts are more likely to gain platform endorsement and public attention, especially in political or public service-related topics (Zhang et al., 2022) [21]. In addition, non-official hosts—including industry experts and social influencers—often display greater resonance in lifestyle, entertainment, and cultural topics, thus generating higher user interaction and enhancing topic heat (Hudders et al., 2021) [22].

In addition, taking Sina Weibo as an example, platform mechanisms treat hosts differently depending on their verified status. Verified accounts often enjoy enhanced privileges—such as prioritized recommendations or pinned content. These privileges increase their content’s visibility and reach. In contrast, unverified users may lack formal recognition but still gain traction through authentic interaction and high frequency participation, especially within niche communities or specific topics.

Therefore, this paper sets host attribute as one of the core explanatory variables and proposes Hypothesis 2:

Hypothesis 2.

The attribute of topic host affects enterprise positive topic heat.

In the opinion dissemination system of online social platforms, a topic host’s activity level on the platform may reflect their motivation and capacity to participate in information propagation (Weeks et al., 2017) [23]. And user activeness and influence play significant roles in topic diffusion on social media.

Activity level can be reflected by two dimensions: posting volume, and the number of followers. Posting volume reflects the host’s sustained engagement with the topic and their capacity to generate content, while the number of followers indicates the potential audience base and the breadth of content dissemination. A higher posting volume helps maintain topic visibility and platform presence, thereby increasing the likelihood of algorithmic recommendation (Pricopoaia & Susanu, 2022) [24]. Additionally, hosts with larger follower bases are more likely to have their posts reshared and commented on, which promotes wider diffusion of the topic across the platform (Kim, 2020) [25].

Beyond basic dissemination capabilities, the host activity level may also show differential effects on topic heat in different contexts. In environments of information overload and attention scarcity, users are more likely to be fatigued by high frequency posting, reducing their willingness to engage (Zheng & Ling, 2021) [26]. Conversely, less active hosts who provide targeted content and personalized expression may elicit stronger emotional resonance, encouraging deeper user discussions and resharing (Chandra et al., 2022) [27]. Moreover, platform algorithms tend to evaluate both activity level and audience interaction preferences in determining content distribution, further complicating the relationship between activity level and topic heat.

Therefore, this paper sets host activity level as one of the core explanatory variables and proposes Hypothesis 3:

Hypothesis 3.

The activity level of topic host affects enterprise positive topic heat.

2.2. Data and Samples

As one of China’s leading online social platforms, Sina Weibo has grown rapidly since its launch in 2010, becoming a major channel for users to access information, express opinions, and engage in public discourse. Its massive user base, diverse topic ecosystem, and high content dissemination efficiency make it the preferred platform for enterprises to seek brand exposure and public engagement. In this context, this paper selects Sina Weibo as the research site. Taking all the microblogging hot search topics on the Sina Weibo platform from 1 June 2023 to 31 May 2024 as initial samples, machine learning methods were employed to crawl various metrics for each topic, including topic heat, time of day on the list, reading volume, discussion volume, interaction volume, originality volume, and accurate posting volume. A total of 140,564 topics were included in the dataset. Additionally, topics were classified based on the presence of a topic host. If a host was present, further data on host attributes—such as posting volume, follower count, and following count—were also collected. The metrics were recorded up until 31 May 2024 with topic heat, host posting volume, and follower count all reported in tens of thousands.

Based on the initial sample, the following three criteria were applied to refine the dataset for the study:

(1) The topic should be related to enterprise positive topic publicity, e.g., “Xiaomi Conference,” “What to watch for Huawei Nova 11.” (2) In moderated topics, the host should not be the company itself. (3) Topics that had appeared on hot search multiple times were de-emphasized, retaining only the latest instance of the topic in the trending data to avoid potential errors due to data duplication. After eliminating samples related to non-existent topics and those missing key variables, a final dataset of 4686 observation samples was obtained.

2.3. Variables and Measurements

2.3.1. Explained Variable

Topic heat ($Heat$) is taken as the explained variable. According to Weibo’s internal algorithm (Weibo Administrator, 2021), topic heat is the product of the sum of search heat, discussion heat, dissemination heat, and the interaction rate. Specifically, search heat refers to the search volume, including both manually input searches and click-through searches, reflecting users’ attention to and exploration of the topic. Discussion heat refers to the volume of discussions, including original posts and retweets, reflecting users’ enthusiasm for engaging in heated discussions and participating. Dissemination heat refers to the reading volume, reflecting the topic’s spread within the microblogging system. Finally, the interaction rate refers to the engagement rate on the hot search results page, reflecting users’ willingness to interact with the content. From this algorithm, topic heat holistically reflects user activity in terms of search, discussion, and dissemination, providing a more comprehensive measure of a topic’s publicity impact than other individual indicators, such as reading volume or discussion volume.

2.3.2. Explanatory Variables

The first explanatory variable is the topic host ($Host$). If the topic has a host, it is assigned a value of 1; otherwise, it is assigned a value of 0. A topic host refers to a Sina Weibo user who is appointed either through self-application or by others to manage, publish, and guide content related to a specific topic. Topic hosts are typically granted permissions such as editing the topic description, curating featured posts, and pinning selected content. They play a central role as organizers and facilitators in the formation and dissemination of topic heat.

The second explanatory variable is the host attribute ($Attribute\text{\_}n)$. Within Sina Weibo’s user hierarchy classification system (Weibo Administrator, 2025), topic hosts are categorized into two groups: ordinary users and verified users (Microblogging big V). Verified users are further delineated into three tiers distinguished by platform-issued verification badges: Gold V, Blue V, and Yellow V. In this study, the classification of host attributes follows this method, with ordinary users categorized as “Non-V.” The classification of host attributes is shown in

Table 1. Classification method for host attributes.

Host Attributes	Account Characteristics
Gold V	High-impact creators, the most influential and valuable V; the number of account followers, reading volume, interaction rate and other indicators have high requirements.
Blue V	Enterprise official accounts, traditional media accounts and government microblogs, representing a certain degree of authority.
Yellow V	Certified authors; users can apply if they meet the certification threshold, and their influence is small compared to gold V.
Non-V	Ordinary users

. In the empirical analysis, the host attribute is treated as a dummy variable, with a dummy variable $Attribute\text{\_}n$ created for each host’s attribute. n ranges from 1 to 4, representing Non-V, Blue V, Gold V, and Yellow V, respectively. If the attribute of the topic host matches a given dummy variable’s attribute, the value is set to 1; otherwise, it is set to 0.

The third explanatory variable is the host activity level ($Rank\text{\_}n$). On various online social platforms, user activity often serves as the basis for platform grading. Building on existing research [28,29], the number of posts and followers of hosts are innovatively selected as the criteria for determining their activity level within each attribute. The specific grading method is shown in

Table 2. Specific classification method of host activity levels.

Activity Level	Basis for Classification
Rank I (high-high)	Both the number of posts and followers are higher than the average of the corresponding variables for all hosts of this attribute
Rank II (high-low)	Posts above the average number of posts for all hosts of this attribute, and followers below the average number of followers for all hosts of that attribute
Rank III (low-high)	Posts below the average number of posts for all hosts of this attribute, and followers above the average number of followers for all hosts of that attribute
Rank IV (low-low)	Both the number of posts and followers are lower than the average of the corresponding variables for all hosts of this attribute

. After classification, dummy variables are applied for empirical analysis of the host rank ($Rank$), generating the corresponding dummy variable $Rank\text{\_}n$ for each host’s rank, with n ranging from 1 to 4 to represent ranks I to IV, respectively. If the rank of the topic host matches the corresponding dummy variable rank, it is assigned a value of 1; otherwise, it is assigned a value of 0.

2.3.3. Control Variables

The influence of opinion leaders can affect the heat of the topics they moderate to a certain extent. Based on existing studies related to the identification of opinion leaders, this paper selects the following control variables that may be correlated with the explanatory variable, topic heat ($Heat$): time of day on the list ($Time$), original volume ($Original$), accurate posting volume ($Accurate$), host posting volume ($Posting$), host’s follower number ($Follower$), and host’s following number ($Following$). Among them, time of day on the list ($Time$) reflects the persistence of user attention to the topic over time. Original volume ($Original$) and accurate posting volume ($Accurate$) reflect users’ enthusiasm for discussing and sharing the topic. Host posting volume ($Posting$), follower number ($Follower$), and following number ($Following$) reflect the influence and activity level of opinion leaders.

To summarize, the definitions and measurement indices of the main variables selected in this study are shown in

Table 3. Definition of main variables.

Typology	Variable Name (Abbreviation)	Variable Definition
Explained Variable	topic heat ($Heat$)	The product of the sum of the search heat, discussion heat and spread heat and the interaction rate; a larger value indicates that the topic publicity effect is better.
Explanatory Variables	topic host ($Host$)	If the topic has a host, it is assigned a value of 1; otherwise, it is assigned a value of 0.
	host attribute ($Attribute\text{\_}n$)	n ranges from 1 to 4, representing Non-V, Blue V, Gold V, and Yellow V, respectively. If the attribute of the topic host matches a given dummy variable’s attribute, the value is set to 1; otherwise, it is set to 0.
	host activity level ($Rank\text{\_}n$)	n ranging from 1 to 4 to represent ranks I to IV, respectively. If the rank of the topic host matches the corresponding dummy variable rank, it is assigned a value of 1; otherwise, it is assigned a value of 0.
Control Variables	time of day on the list ($Time$)	The amount of time a topic stays on the Weibo Hotlist.
	original volume ($Original$)	Number of original tweets posted carrying hot topics, not including related tweet retweets.
	accurate posting volume ($Accurate$)	Total number of tweets that can be searched by omitting some of the similar results under the topic.
	host posting volume ($Posting$)	Total number of tweets posted by topic hosts since account creation.
	host’s follower number ($Follower$)	Total number of followers a topic host has.
	host’s following number ($Following$)	Number of other Weibo users followed by the topic host.

.

2.4. Research Models

To realize the research purpose, the following three sample regression models are designed sequentially.

Regression Model 1 examines the effect of topic host ($Host$) on topic heat ($Heat$). The specific model is as follows:

Heat_i = a₀ + a₁ × Host_i + a₂ × Time_i + a₃ × Original_i + a₄ × Accurate_i + ε_i

(1)

Regression Model 2 examines the effect of host attributes ($Attribute\text{\_}n$) on topic heat ($Heat$). The specific model is as follows:

Heat_i = b₀ + b₁ × Attribute_n_i + b₂ × Time_i + b₃ × Original_i + b₄ × Accurate_i

+b₅ × Posting_i + b₆ × Follower_i + b₇ × Following_i + ε_i

(2)

Regression Model 3 examines the effect of host activity level ($Rank\text{\_}n$) on topic heat ($Heat$). The specific model is as follows:

Heat_i = c₀ + c₁ × Rank_n_i + c₂ × Time_i + ε_i

(3)

In these models, subscript i represents the $i$th topic and $\epsilon$ represents the random error term.

2.5. Descriptive Statistics

The descriptive statistics of each variable, based on the grouping of topic host attributes, are shown in

Table 4. Descriptive statistics of main variables.

Host Attributes	Variables	Sample Size	Mean	Median	Standard Deviation	Minimum	Maximum
No host	Heat	483	58.919	30.948	120.310	0.932	1505.691
	Time		157.723	93	173.216	1	1023
	Original		62,052.317	3765	267,532.391	8	4,288,000
	Accurate		1416.605	302	10,482.745	0	208,198
Non-V	Heat	783	46.004	29.820	47.393	1.040	451.717
	Time		144.783	114	130.700	1	757
	Original		12,406.987	1220	129,075.510	9	3,023,900
	Accurate		237.628	224	372.510	0	10,044
	Posting		1.021	0.644	1.070	0	5.534
	Follower		188.700	107.595	440.581	0.001	8400
	Following		696.807	394	777.218	0	4548
Blue V	Heat	1152	46.935	27.200	64.404	0.248	1143.487
	Time		172.396	123.500	170.511	1	1225
	Original		8489.743	637	154,885.350	9	4,999,000
	Accurate		200.605	198	158.541	0	1437
	Posting		10.235	9.575	7.326	0.002	33.618
	Follower		2312.941	578.600	3942.240	0.385	15,300
	Following		1517.947	1264	1508.607	9	32,525
Gold V	Heat	1177	74.729	58.580	110.602	0.901	1218.941
	Time		227.760	187	182.942	1	1325
	Original		6911.025	1409	49,175.089	12	1,196,000
	Accurate		230.425	245	105.848	0	968
	Posting		3.541	2.375	3.862	0.034	54.062
	Follower		662.033	617.668	482.721	3.400	6329.351
	Following		1076.427	969	822.529	2	5652
Yellow V	Heat	1091	67.989	53.619	55.173	0.924	543.417
	Time		229.630	205	175.304	1	878
	Original		5958.312	2373	12,193.318	33	137,000
	Accurate		234.388	249	114.821	0	1697
	Posting		2.228	1.417	2.522	0.008	29.733
	Follower		429.443	329.359	438.902	0.010	7060
	Following		1028.798	850	826.710	7	4753

. Upon comparison, it is found that in topics with no host, Non-V, Blue V, Gold V, and Yellow V as hosts, the average topic heat values are 58.919, 46.004, 46.935, 74.729, and 67.989, respectively. The average heat of the topics follows the order: Gold V > Yellow V > no host > Blue V > Non-V. The topics with Gold V as the host have the highest average heat, while those with Non-V as the host have the lowest. In terms of original volume and precise posting volume, the average values for topics with no host are 62,052.317 and 1416.605, respectively, and for topics with Non-V as the host, the averages are 12,406.987 and 237.628, respectively. These values are ranked at the top of all topics, indicating that netizens are more enthusiastic about making original posts in enterprise positive topics without hosts or with ordinary users as hosts.

3. Results

Before conducting the empirical analysis, the following data preprocessing steps were performed to ensure the model’s performance: the main continuous variables were minorized at the 5th and 95th percentiles; a Variance Inflation Factor (VIF) diagnosis was conducted for all explanatory and control variables, with the results indicating that the VIF values of all variables were below 10, suggesting the absence of multicollinearity. Furthermore, the White test was employed to check for heteroscedasticity, and robust standard errors were used in the regression analysis to account for potential heteroscedasticity.

3.1. Testing the Relationship Between Topic Host and Topic Heat

The regression results examining the relationship between topic host and topic heat are shown in

Table 5. Regression results of topic host and topic heat.

Variables	Heat
Host	13.028 ***
	(7.345)
Time	0.086 ***
	(22.075)
Original	0.001 ***
	(6.560)
Accurate	0.051 ***
	(10.819)
cons	9.923 ***
	(5.229)
N	4686
R2	0.221

Note: *** p < 0.01, values in parentheses below are t-statistics.

. The results reveal that the regression coefficient for the variable topic host ($Host$) is 13.028, with $p$ < 0.01, indicating that the presence of a host significantly impacts the topic’s heat. Specifically, when other variables remain constant, the average heat of a topic with a host increases by 130,280 compared to a topic without a host. Higher heat contributes to the topic occupying a higher position on the Sina Weibo hot search list, ultimately enhancing the publicity effect for the enterprise. This result validates Hypothesis 1. Among the other control variables, the coefficients of $Time$, $Original$, and $Accurate$ are all significantly positive, suggesting that the longer the topic remains on the list each day, the more original and accurate posts it generates and the higher the topic’s heat, which aligns with expectations.

3.2. Testing the Relationship Between Host Attributes and Topic Heat

A regression analysis was conducted on all topics with hosts to obtain the regression results for the relationship between host attributes and topic heat, as shown in

Table 6. Regression results of host attributes and topic heat.

Variables	Heat
	Non-V.	Blue V	Gold V	Yellow V
Attribute_1	−8.843 ***
	(−6.422)
Attribute_2		−9.532 ***
		(−7.125)
Attribute_3			3.825 ***
			(3.682)
Attribute_4				8.659 ***
				(6.435)
Time	0.081 ***	0.084 ***	0.084 ***	0.082 ***
	(19.285)	(20.264)	(20.157)	(19.844)
Original	0.001 ***	0.001 ***	0.001 ***	0.001 ***
	(6.477)	(5.913)	(6.535)	(5.796)
Accurate	0.067 ***	0.063 ***	0.065 ***	0.065 ***
	(12.140)	(11.399)	(11.813)	(11.848)
Posting	−0.729 ***	−0.030	−0.519 ***	−0.400 ***
	(−4.811)	(−0.184)	(−3.381)	(−2.682)
Follower	0.001	0.001	0.001	0.001
	(0.904)	(0.871)	(0.981)	(1.374)
Following	0.002 ***	0.002 ***	0.003 ***	0.002 ***
	(3.192)	(3.509)	(3.675)	(3.202)
_cons	22.200 ***	20.371 ***	17.894 ***	16.818 ***
	(14.710)	(14.434)	(13.293)	(12.659)
N	4203	4203	4203	4203
R2	0.242	0.242	0.236	0.244

Note: *** p < 0.01.

. According to the experimental results, the coefficients of the dummy variables corresponding to each host attribute are all significant at the 1% level. This result validates Hypothesis 2. The order of contribution of each host attribute to topic heat is as follows: Yellow V $(\beta =8.659)>$ Gold V $(\beta =3.825)>$ Non-V $(\beta =-8.843)$ > Blue V $(\beta =-9.532).$ Among these, the coefficients of the dummy variables for Gold V and Yellow V are positive, indicating that these hosts enhance the topic heat. The regression coefficient for the dummy variable corresponding to Yellow V is 8.659, meaning that the average heat of topics with a Yellow V host increases by 86,590 compared to those with hosts of other attributes, which is greater than the increase associated with Gold V by 38,250. On the other hand, the coefficients for the dummy variables of Non-V and Blue V are negative, suggesting that these hosts reduce the topic heat. Specifically, the contribution of Blue V to topic heat is −95,320, which is lower than the value of −88,430 associated with Non-V.

From the ranking of the contribution to topic heat by each host attribute, it can be observed that, in addition to the expected ranking (Yellow V, Gold V > Non-V, Gold V > Blue V), Yellow V not only ranks higher than Blue V but also surpasses Gold V, which is superior to Yellow V in terms of account readership, influence, fan interaction rates, and other factors. Moreover, Non-V, representing general users with the least influence, ranks higher than the official Blue V accounts. There is a noticeable discrepancy in the overall ranking compared to the average heat rank of topics categorized by host attributes, indicating some variation in the topic heat levels.

3.3. Testing the Relationship Between Host Activity Level and Topic Heat

Before conducting the empirical analysis, an empirical model was developed to investigate the relationship between the number of posts ($Posting)$ and followers ($follower$) of hosts and topic heat ($Heat$). The results indicate that the coefficients for both variables are significantly positive, thus supporting the rationale for using the number of posts and followers as the basis for grading hosts. All samples of topics with hosts were categorized into four groups according to the attributes of the hosts to examine the relationship between host activity rank and topic heat. The regression results for each group are presented in

Table 7. Regression results of Non-V host activity level and topic heat.

Variables	Heat
	Rank I (High-High)	Rank II (High-Low)	Rank III (Low-High)	Rank IV (Low-Low)
rank_1	−4.397
	(−1.617)
rank_2		3.644
		(1.196)
rank_3			−7.443 ***
			(−2.766)
rank_4				3.932 *
				(1.848)
Time	0.072 ***	0.072 ***	0.071 ***	0.072 ***
	(9.746)	(9.659)	(9.482)	(9.648)
_cons	31.082 ***	29.719 ***	31.109 ***	28.062 ***
	(21.449)	(21.202)	(21.289)	(16.016)
N	783	783	783	783
R2	0.132	0.131	0.133	0.132

Note: *** p < 0.01, * p < 0.1.

,

Table 8. Regression results of Yellow V host activity level and topic heat.

Variables	Heat
	Rank I (High-High)	Rank II (High-Low)	Rank III (Low-High)	Rank IV (Low-Low)
rank_1	−0.417
	(−0.406)
rank_2		−2.156 *
		(−1.854)
rank_3			−0.636
			(−0.622)
rank_4				1.589 *
				(1.728)
Time	0.092 ***	0.092 ***	0.092 ***	0.092 ***
	(25.219)	(25.227)	(25.230)	(25.239)
_cons	26.587 ***	26.790 ***	26.604 ***	25.718 ***
	(42.307)	(45.902)	(44.712)	(41.374)
N	1091	1091	1091	1091
R2	0.077	0.077	0.077	0.077

Note: *** p < 0.01, * p < 0.1.

,

Table 9. Regression results of Blue V host activity level and topic heat.

Variables	Heat
	Rank I (High-High)	Rank II (High-Low)	Rank III (Low-High)	Rank IV (Low-Low)
rank_1	17.637 ***
	(12.718)
rank_2		2.445 **
		(2.333)
rank_3			−9.786 ***
			(−6.995)
rank_4				−14.476 ***
				(−16.497)
Time	0.151 ***	0.153 ***	0.153 ***	0.150 ***
	(44.908)	(43.950)	(44.505)	(44.383)
_cons	24.186 ***	27.800 ***	28.359 ***	36.746 ***
	(41.533)	(55.995)	(56.921)	(48.004)
N	1152	1152	1152	1152
R2	0.053	0.049	0.049	0.052

Note: *** p < 0.01, ** p < 0.05.

and

Table 10. Regression results of Gold V host activity level and topic heat.

Variables	Heat
	Rank I (High-High)	Rank II (High-Low)	Rank III (Low-High)	Rank IV (Low-Low)
rank_1	−0.474
	(−0.507)
rank_2		1.954 *
		(1.902)
rank_3			−3.030 ***
			(−3.436)
rank_4				1.554 **
				(2.238)
Time	0.078 ***	0.078 ***	0.079 ***	0.078 ***
	(27.147)	(27.087)	(27.210)	(27.184)
_cons	26.203 ***	25.806 ***	26.738 ***	25.446 ***
	(55.614)	(53.428)	(57.188)	(48.758)
N	1177	1177	1177	1177
R2	0.076	0.077	0.077	0.077

Note: *** p < 0.01, ** p < 0.05, * p < 0.1.

.

From the experimental findings presented in Table 7 and Table 8, it is evident that, when the host’s attribute is Non-V, the regression coefficients for Rank III (low-high) and IV (low-low) are statistically significant. Rank IV (low-low) has a positive effect on topic heat ($\beta =$ 3.932), while Rank III (low-high) exhibits a negative effect on topic heat ($\beta =-$7.443). The coefficients for Rank I (high-high) and Rank II (high-low) do not meet the significance threshold. When the host’s attribute is Yellow V, the regression coefficients for Ranks II (high-low) and IV (low-low) are significant. Rank IV (low-low) continues to show a positive contribution to topic heat ($\beta =$ 1.589), while Rank II (high-low) negatively impacts topic heat ($\beta =-$2.156). The coefficients for Rank I (high-high) and Rank III (low-high) fail to pass the significance test. This result validates Hypothesis 3. Contrary to the initial expectation that Rank I (high-high) hosts would possess the strongest capability to boost topic heat, Rank IV (low-low) emerges as the optimal Rank for Non-V and Yellow V hosts in terms of enhancing topic heat. Furthermore, the coefficients for Rank I (high-high) do not achieve statistical significance for either host attribute group.

Table 9 presents the regression results for the relationship between the activity level of Blue V hosts and topic heat. As shown in Table 9, when the host attribute is Blue V, the order of contribution to topic heat is as follows: Rank I ($\beta =$ 17.637) > Rank II ($\beta =$ 2.445) > Rank III $(\beta =-$9.786) > Rank IV ($\beta =-$14.476), with all regression coefficients passing the significance test. Specifically, Rank I (high-high) hosts capitalize on their high posting volume and large follower base, leading to the greatest positive impact on topic heat. In contrast, Rank IV (low-low) hosts, with lower posting volume and fewer followers, are unable to leverage the high account activity associated with a greater posting frequency and larger audience to boost the topic heat. As a result, their effect on topic heat is the most detrimental. These findings are consistent with expectations. Notably, Rank II (high-low) demonstrates a stronger contribution to topic heat than Rank III (low-high).

As shown in Table 10, when the host attribute is Gold V, the coefficients for Rank II (high-low), Rank III (low-high), and Rank IV (low-low) pass the significance test, with Rank II (high-low) making the greatest contribution to topic heat ($\beta =$ 1.954). Rank IV (low-low) ranks second ($\beta =$ 1.554), while Rank III (low-high) has the lowest contribution ($\beta =-$3.030). The coefficient for Rank I (high-high) does not pass the significance test. Combined with the descriptive statistical analysis, it is observed that as posting volume increases (from Rank IV to Rank II), the corresponding coefficient of activity level increases (from 1.554 to 1.954). However, when the number of followers increases (from Rank IV to Rank III), the corresponding coefficient decreases (from 1.554 to −3.030), suggesting that an increase in the number of followers for Gold V hosts diminishes their contribution to topic heat.

3.4. Robustness Test

The robustness test is conducted in two parts: alternative measurements, and endogenous discussion. In terms of alternative measurements, the explanatory variable topic heat is replaced by other indicators that can reflect the publicity effect of a topic, including reading volume, discussion volume, and interaction volume. Reading volume refers to the total number of times all related posts under the topic have been viewed by users. Discussion volume refers to the total number of user interactions centered on the topic. Interaction volume measures the total number of likes, comments, and reposts on posts related to the topic. Regression analysis is then re-conducted using these alternative measures, and the results remain consistent with the initial findings.

Regarding endogenous discussion, the empirical results may be influenced by omitted variables or self-selection bias. The higher heat of topics hosted by hosts, or by hosts with certain attributes or activity levels, could be attributed to other factors. For example, the topic may involve a company with high visibility, or the host’s posts may coincide with peak usage periods on Sina Weibo. To address this, the propensity score matching (PSM) model is employed to match samples for each explanatory variable. The differences in topic heat ($Heat$) between the two groups of matched samples are then analyzed. To better control for differences in the characteristics of the sample topics beyond the explanatory variables, the control variables from each empirical model are used as matching variables. In selecting the matching method, the 1:1 nearest neighbor matching method is chosen to construct the treatment and control group samples. The results of this test are consistent with the previous findings, further validating the conclusions.

3.5. Optimization of Enterprise Positive Topic Heat Enhancement Strategies for Cost Performance

3.5.1. Defining Cost Performance of Hosts

The empirical results in the previous section suggest that enterprises can leverage hosts with different attributes and activity levels to enhance topic heat. However, cost constraints play a significant role for enterprise publicizing on online social platforms. Considering limited budgets, how to select the most cost-effective strategies to enhance positive topic heat under resource constraints has become a key issue in enterprise operation and management. In this regard, this section proposes to construct a host/cost performance ratio index to calculate the cost performance of hosts with the optimal heat contribution level for each attribute. This index will help identify the host attribute with the highest cost performance ratio, thus optimizing the enterprise positive topic heat enhancement strategies from the perspective of cost performance.

The cost performance of a host is defined as the increase in topic heat per unit of hiring cost. In terms of performance, the regression coefficient corresponding to each attribute’s optimal contribution level is selected, representing the best contribution of that host attribute to the topic heat. In terms of cost, a host’s activity is directly proportional to the hiring cost, with activity primarily reflected in the number of posts and the number of followers. Given that different enterprises place varying degrees of emphasis on these activity indices when selecting hosts, a preference coefficient ($\alpha$) is introduced to capture this variability: a higher value of $\alpha$ indicates a greater emphasis on the host’s posting volume, while a lower value of $\alpha$ suggests a focus on the host’s number of followers. Based on the analysis above, the $CostPerformance$ ($CP$) formula is defined as follows:

$$\mathrm{CP}=\frac{\mathrm{\beta }}{\mathrm{Posting}\times \mathrm{\alpha }+\mathrm{Follower}\times \left(1-\mathrm{\alpha }\right)}$$

(4)

3.5.2. Host Cost Performance Calculation

Due to the significant disparity between the values of hosts’ posting volume and number of followers, both posting volume and follower count were standardized using the maximization method within a linear proportional standardization framework. This allowed for the derivation of standardized values for posting volume and follower count. The cost performance calculation was then conducted by taking the average values of posting volume and follower count for each host attribute as representative indicators. Furthermore, by integrating the results of regressions where posting volume and follower count served as explanatory variables for topic heat, it was found that posting volume contributes more significantly to topic heat than follower count. Thus, $\alpha$ should be greater than 0.5. Using $\alpha$ values of 0.6 and 0.7, the cost performance ratios for hosts of each attribute were calculated, as presented in

Table 11. Calculation results for cost performance of each host attribute.

Host Attribute	Topic Heat Contribution	Average Posting Volume	Average Number of Followers	$\mathit{C}\mathit{P}$ (α = 0.6)	$\mathit{C}\mathit{P}$ (α = 0.7)
Non-V	3.932	0.01086	0.00643	432.65845	412.54853
Blue V	17.637	0.10979	0.12945	149.90566	152.45315
Yellow V	1.589	0.02292	0.01654	78.014533	75.645054
Gold V	1.954	0.04155	0.02886	53.572408	51.771189

.

The results indicate that the cost performance rankings of host attributes are consistent when $\alpha$ is set to 0.6 or 0.7. Non-V hosts exhibit the highest cost performance when serving as topic host, followed by Blue V, Yellow V and Gold V, which ranks the lowest in cost performance. This finding is notably different from the previous rankings of host contribution to topic heat by attribute, where cost performance was not considered. The cost performance ranking of Non-V, Yellow V, and Gold V is in direct contrast to the ranking of Sina Weibo big V levels. It is important to note that when calculating hiring costs, this study only accounts for the posting volume and number of followers. Additionally, the engagement cost includes other unquantifiable factors, such as those based on the host’s attributes. Nevertheless, regardless of the cost nature, Gold V, as the highest-level V, incurs the highest cost, while Non-V, as an ordinary user, incurs the lowest cost. Therefore, the absence of these additional cost factors in the calculation does not impact the overall cost performance calculation.

4. Discussion

According to the experimental results in Table 5, once a topic generates enough heat to attract attention, interactions and retweets by netizens amplify the visibility and influence of the topic host. As a topic manager, the host will typically craft the topic introduction carefully, aligning it with current trends and promoting the topic content from multiple perspectives to encourage user participation. In addition to editing the basic elements of the topic introduction, the host may also recommend related users and microblogs. Some of these recommended users or hosts may be high-profile experts, authorities, or celebrities in a given field, and their celebrity effect encourages netizens to engage in discussions, further boosting the heat of topics with a host compared to those without one.

This paper attributes the Table 6 experimental results to the unique grassroots identity of Non-V and Yellow V. As certified authors, most Yellow Vs are ordinary users with verified accounts, sharing characteristics similar to Non-V accounts. This grassroots status allows Yellow V and Non-V hosts to conceptualize topics, express opinions, and serve as public opinion promoters from the perspective of the general public. Topics closely related to people’s everyday lives resonate deeply with netizens, sparking their enthusiasm for discussion, and consequently drawing more attention to the enterprise-related content. For instance, a topic such as “Which is more practical, Huawei’s small folding phone or the iPhone 14?” encourages users to share personal experiences, making the topic highly engaging and increasing its heat. In contrast, the “serious” nature of Blue V accounts gives the topics they host a more formal tone, and netizens mostly regard them as news, leading to a more factual and less interactive discussion. Gold V topics, on the other hand, tend to be more marketing-oriented and entertaining, such as “Giant Popping Cookies,” which attracts many users but often results in shallow engagement and limited in-depth discussion. Thus, the unexpected positive contribution of Non-V and Yellow V to the topic heat is not consistent with their influence level. Compared to ordinary users, Yellow V hosts have already built a certain fan base and hold a more prominent voice on social platforms, which aids in the spread of the topics they manage. As a result, Yellow V surpasses Gold V in terms of topic heat contribution, emerging as the most effective host attribute in boosting topic heat.

According to the experimental results in Table 7 and Table 8, building on the results from the previous test examining the relationship between host attributes and topic heat, this paper attributes the observed phenomenon to the approachable and relatable nature of the accounts managed by Non-V and Yellow V hosts. For these hosts who are closer to the masses, the level of posting and followers reflects the degree of relatability they offer. Hosts with lower posting volumes and fewer followers typically produce content that resonates more with daily life, which encourages users to interact and participate in topic discussions. These Non-V and Yellow V hosts who maintain this relatable communication style are better positioned to preserve authenticity when promoting enterprise-related topics. As a result, they are more effective at drawing public interest and achieving higher topic heat. In contrast, hosts with higher posting volumes and larger numbers of followers often show reduced relatability. Netizens show less enthusiasm for discussing positive enterprise-related topics issued by such accounts. Many Non-V and Yellow V hosts gradually adopt commercial strategies similar to those used by Gold V hosts. They begin to post marketing or entertainment-oriented content, aiming to elevate their influence level. In this phase, their posting volume and follower counts increase. However, their content becomes more distant from everyday users’ concerns. As a result, they lose the relatability advantage that previously made their enterprise topics more engaging. In addition, this phenomenon can be further interpreted through the perspective of content perception mechanisms [30]. On online social platforms, users tend to evaluate and engage with topic content based not only on factual information, but also on perceived relevance, emotional resonance, and social alignment. When topic hosts curate content that aligns with users’ values, lifestyles, or daily experiences, users are more likely to perceive the content as authentic and relatable, which fosters higher willingness to interact and disseminate the topic. Conversely, content perceived as overly commercial, institutional, or detached from user realities may be interpreted as less trustworthy or engaging, reducing its potential to generate sustained heat.

According to the experimental results in Table 9, as an officially certified account, Blue V is primarily characterized by formal and authoritative content. Beneath the surface of highly active accounts, the high posting volume often indicates that Blue V hosts possess unique information channels or exceptional analytical abilities, enabling them to consistently produce forward-looking content to meet users’ informational needs. Furthermore, unique and authoritative content tends to attract other big V accounts to engage with the topic, stimulating more users to participate in the discussion. For example, the topic “Good Price List” posted by the Blue V account “Finance and Economics” on 29 May 2023 with a posting volume of 264,600, analyzed the product list from a forward-looking perspective and highlighted affordable products for users. This topic became a useful reference for shoppers and naturally attracted many other big V accounts to recommend and forward the topic, ultimately propelling it to the top of the hot search list due to its high heat. In contrast, while the number of followers reflects Blue V’s recognition by netizens, it does not necessarily signify the uniqueness of the content they post. Therefore, for Blue V hosts, the contribution of posting volume to topic heat is more substantial than the number of followers.

According to the experimental results in Table 10, this paper attributes this phenomenon to the misleading nature of the number of followers indicator. As the highest tier of big V, one of the criteria for certification is having a substantial follower count. Among the Gold V hosts included in this empirical analysis, the minimum follower count reached 34,000. However, the commercial value and associated economic interests of large follower bases, along with the marketing-oriented operation models of accounts, have incentivized certain Gold V hosts to engage in artificial follower inflation. These practices may include the use of informal methods to boost follower numbers, resulting in a portion of the follower base consisting of inauthentic or automated accounts. The presence of such accounts undermines the credibility of follower count as a reliable measure of digital influence. Additionally, hosts who benefit from inflated follower numbers may no longer prioritize producing high-quality or engaging content. A lack of originality and relevance in the topics they promote reduces audience interest and limits participation. This ultimately leads to a disconnect between Gold V’s high follower count and the actual heat level of the topics they host.

According to the experimental results in Table 11, this phenomenon is explained by the varying degrees of alignment between host performance and cost. Specifically, Non-V hosts, with the lowest posting volume and number of followers, have the lowest employment costs. Their grassroots status enables them to craft enterprise positive topics with a livelihood-focused core from the perspective of the general public. This relatability advantage fosters greater trust from netizens in the content they post, motivating more original discussions based on personal experiences. This leads to an exceptional contribution of Non-V hosts to the topic heat, and the significant disparity between this contribution and their low hiring cost results in the highest cost performance. In contrast, Gold V hosts’ high hiring costs are driven by their second higher posting volume and follower count, as well as their position as the highest-ranking big V. However, unlike Non-V hosts, Gold V hosts’ commercialized account nature prevents their topics from resonating with netizens. Their content is often perceived as “marketing.” Moreover, with social platforms becoming a major venue for enterprise publicity, high-influence accounts like Gold V are frequently chosen for advertising. The overt and aggressive nature of advertising implantation has generated a resistance among netizens to high-ranking accounts like Gold V, resulting in a lack of trust in their positive topic content. Consequently, the performance of Gold V hosts when serving as topic host does not align with their high hiring cost, leading to the lowest cost performance. The higher alignment between performance and cost of both Blue V and Yellow V hosts, along with Blue V hosts’ greater contribution to topic heat, places them second and third in the cost performance ranking, respectively.

5. Conclusions

5.1. Research Findings

As online social platforms have become new channels for enterprise publicity, accurately identifying the key factors influencing topic heat is essential for enterprises aiming to enhance the heat of their positive topics. This paper investigates enterprise positive topic publicity on Sina Weibo, proposes relevant hypotheses, and constructs several empirical analysis models to examine the relationship between topic host, host attributes, host activity levels and topic heat. In addition, a cost performance index is further developed by integrating performance and cost dimensions, enabling the quantification and evaluation of host cost performance. All the proposed hypotheses are empirically supported. This study not only identifies the key factors that effectively enhance the topic heat of positive topics, but also fills existing research gaps by introducing causal explanations and a cost performance perspective. The main conclusions are as follows:

(1) Topic hosts play an active role in positive topic publicity. Topics with hosts exhibit higher levels of topic heat compared to those without hosts. There are differences in the contribution of various host attributes to topic heat, with Yellow V showing the most significant impact, followed by Gold V, Non-V and Blue V, in that order. Ordinary users and Yellow V hosts show a topic heat enhancement ability that is not aligned with their influence levels. (2) The contribution of hosts’ attributes to topic heat varies according to their activity levels. There is not always a direct match between high activity levels and high topic heat contribution. Among the various host attributes, Non-V and Yellow V (both with lower posting volumes and fewer followers), Blue V (with higher posting volume and more followers), and Gold V (with higher posting volume but lower follower count) produce the best topic heat contribution. (3) The cost performance ranking of host attributes shows significant differences from the ranking based on their topic heat contribution and Sina Weibo big V levels. The cost performance ranking is as follows: Non-V > Blue V > Yellow V > Gold V. Based on these findings, this study provides targeted suggestions for enterprises in selecting online social platforms publicity strategies.

5.2. Practical Implications

The empirical analysis results above provide significant guidance for enhancing the positive topic heat of enterprises on online social platforms. The first recommendation is based on the experimental results shown in Table 5. The “mutually beneficial symbiotic” relationship between the host and the positive topic encourages hosts to increase topic heat as much as possible by employing strategies such as recommending users. Therefore, when conducting positive topic publicity, enterprises should prioritize hiring users as hosts to achieve higher topic heat.

The second recommendation is based on the experimental results shown in Table 6. Without considering cost performance, enterprises can prioritize selecting Yellow V hosts, which have the highest topic heat contribution, to maximize the increase in topic heat. The highest-ranking big V on social platforms does not necessarily guarantee the best topic heat effect. Enterprises should change the traditional “rank-first” perspective and pay more attention to lower-ranked hosts who have the grassroots advantage.

The third recommendation is based on the experimental results shown in Table 7, Table 8, Table 9 and Table 10. When hiring others to conduct publicity on social platforms, enterprises should adjust the activity index assessment standards according to the attributes of the selected users. For example, in positive topic publicity, enterprises should further optimize the selection of specific hosts based on the optimal heat contribution levels of hosts from each attribute (Non-V: Rank IV; Blue V: Rank I; Gold V: Rank II; Yellow V: Rank IV). While recognizing the “deception” behind high activity indicators, enterprises should also consider the “potential” behind low activity indicators.

The last recommendation is based on the experimental results shown in Table 11. Considering cost and performance, enterprises should prioritize selecting ordinary users with the highest cost performance as topic hosts and increase the number of hosts employed. By leveraging the high cost performance of ordinary users, enterprises can minimize cost expenditure while ensuring topic heat, thus achieving the best topic heat effect under cost constraints. A low-cost publicity strategy on social platforms may imply high cost performance, and increasing the number of times low-cost strategies are implemented may lead to a “double win” of excellent topic heat effect and low-cost expenditure. This offers a novel approach for enterprises to optimize their strategic choices aimed at enhancing positive topic heat.

5.3. Research Limitations

This research focuses on three key dimensions: topic host, the attributes of the host, and the hosts’ activity levels. It uses empirical analysis models and the quantification of hosts’ cost performance to deeply analyze the influencing factors of enterprise positive topic heat and their correlation with topic heat. However, the specific mechanisms and diminishing marginal effects of hosts’ impact on topic heat, as well as their overall utility, warrant further study.

5.4. Future Research Directions

Additionally, the scope of enterprise publicity in the new media era is not limited to a single online social platform. Guided by the identification of influencing factors of positive topic heat, future research will focus on analyzing enterprises’ strategies for enhancing positive topic heat on other social media platforms such as TikTok, in order to cross-validate the generalizability of the present findings across different online social platforms.