Introduction

The global gaming industry has witnessed unprecedented growth in recent years, with Massively Multiplayer Online Games (MMORPGs) emerging as a dominant segment due to their immersive gameplay and social interaction features. Within this landscape, the purchase of virtual goods—such as in-game currency, skins, and tradable assets—has become a significant revenue stream for game developers, contributing billions of dollars annually to the global economy (Newzoo, 2025). In Vietnam, a rapidly growing market for digital entertainment, MMORPGs have gained immense popularity, driven by the country’s young, tech-savvy population and increasing smartphone penetration. The rise of blockchain-based games like Axie Infinity, which allow players to purchase and trade virtual goods as Non-Fungible Tokens (NFTs), has further transformed the MMORPG market in Vietnam, creating new opportunities and challenges for understanding player behavior (Hamari and Keronen, 2017).

Vietnam’s gaming market has experienced exponential growth, with over 50 million gamers as of 2025, representing more than half of the country’s population (Newzoo, 2025). MMORPGs, such as Liên Minh Huyền Thoại (League of Legends), Võ Lâm Truyền Kỳ, and Axie Infinity, are particularly popular due to their ability to foster social connections and provide immersive experiences. The purchase of virtual goods in these games has become a key monetization strategy, with players spending on items that enhance gameplay, aesthetics, or social status. In Vietnam, the adoption of mobile MMORPGs and the integration of blockchain technology have accelerated this trend, as players increasingly value virtual goods that offer both in-game benefits and real-world economic value (Leong et al., 2019). For instance, Axie Infinity, a blockchain-based MMORPG, has enabled Vietnamese players to earn income through NFT trading, making virtual goods purchases a significant economic activity (Hamari and Keronen, 2017).

Despite the global growth of virtual goods purchases in MMORPGs, several critical research gaps limit our understanding of purchase intention, particularly in non-Western, emerging markets like Vietnam. First, most studies have focused on Western markets (USA, Germany, Italy), with limited attention to non-Western contexts like Vietnam, despite its rapidly growing gaming market with over 50 million gamers (Newzoo, 2025; Johnson and Carter, 2021; Müller and Braun, 2022; Bianchi and Rossi, 2023). Although TAM was originally developed for mandatory technology adoption in organizational settings (Davis, 1989), its constructs (PU, PEOU) are relevant to voluntary purchase intention in MMORPGs, as players evaluate the usefulness and ease of purchasing virtual goods. This study justifies TAM’s application in a consumer-driven context. This extends TAM applications in digital contexts as seen in journal scholarship (Wang et al., 2025; Liu et al., 2025), and integrates it with PERVAL to address its limitations, while PERVAL captures value-based dimensions such as Emotional Value (VE), Social Value (VS), Price/Value for Money (VP), and Performance/Quality Value (VQ). However, these models have rarely been integrated to provide a holistic understanding of purchase intention, particularly in a collectivist culture like Vietnam, where social influence and economic factors play a significant role (Kaya and Ozkan, 2022).

Third, the unique characteristics of Vietnam’s MMORPG market, including the dominance of mobile gaming, widespread adoption of mobile payment systems (e.g., MoMo, ZaloPay), and the rise of blockchain-based games like Axie Infinity, introduce context-specific factors such as Blockchain Utility (BLOCK) and Mobile Payment Ease (MPE) that are underexplored in the literature (Kim et al., 2010; Hamari and Keronen, 2017). These factors are critical in Vietnam, where mobile and blockchain technologies shape purchase behavior, yet few studies have investigated their impact. Fourth, traditional linear methods like PLS-SEM may not fully capture the non-linear relationships inherent in complex consumer behavior, such as diminishing returns of blockchain utility or threshold effects of mobile payment ease (Leong et al., 2019). The limited application of integrated methodologies combining explanatory (PLS-SEM) and predictive (ANN) approaches in MMORPG research represents a methodological gap. This study addresses these gaps by examining context-specific factors and employing an integrated PLS-SEM and ANN approach to provide a comprehensive understanding of virtual goods purchase intention in Vietnam.

This study aims to address the identified research gaps by evaluating the factors influencing Vietnamese MMORPG players’ intention to purchase virtual goods using a hybrid TAM-PERVAL model and an integrated PLS-SEM and ANN approach. The specific objectives are: (a) To investigate key factors influencing purchase intention in Vietnam’s MMORPG market, including functional (PU, PEOU), value-based (VE, VS, VP, VQ), and context-specific factors (TRUST, IMM, AES, BLOCK, MPE), addressing the lack of research in non-Western contexts. (b) To integrate TAM and PERVAL into a hybrid model, examining the mediating roles of Satisfaction (SAT) and Social Influence (SI), and the moderating effects of Income Level (INC) and Gaming Experience (EXP), to capture both functional and value-based drivers in a collectivist culture. (c) To apply an integrated PLS-SEM and ANN methodology, using PLS-SEM to test the theoretical model and hypotheses and ANN to model non-linear relationships, addressing methodological gaps and enhancing predictive accuracy.

The study addresses the following research questions: (1) What factors influence Vietnamese MMORPG players’ intention to purchase virtual goods, and how do TAM and PERVAL constructs interact in this context? (2) How do Satisfaction (SAT) and Social Influence (SI) mediate the relationships between TAM/PERVAL constructs and purchase intention? (3) How do Income Level (INC) and Gaming Experience (EXP) moderate the effects of key predictors on purchase intention? (4) What non-linear relationships exist among the factors influencing purchase intention, and how do they enhance predictive accuracy in the Vietnamese MMORPG market?

This study offers significant theoretical and practical contributions. Theoretically, it extends the TAM and PERVAL frameworks by integrating them into a hybrid model, providing a more comprehensive lens to understand virtual goods purchase intention in a non-Western, collectivist context. The novelty lies in its focus on emerging markets like Vietnam, where prior research is scarce (e.g., Hamari and Keronen, 2017; Leong et al., 2019; Kim et al., 2010), and in explicitly incorporating mediation (SAT and SI bridging functional/value-based factors to intention) and moderation (INC and EXP altering predictor effects, e.g., strengthening VP and PU for higher-income players), enriching the model’s explanatory power in culturally nuanced settings (Venkatesh and Bala, 2008; Shmueli et al., 2019). The inclusion of context-specific factors like Blockchain Utility (BLOCK) and Mobile Payment Ease (MPE) addresses gaps in the literature, particularly in emerging markets like Vietnam, where blockchain gaming and mobile payments are transforming consumer behavior (Hamari and Keronen, 2017; Kim et al., 2010). The integrated PLS-SEM and ANN approach advances methodological contributions by combining explanatory (PLS-SEM) and predictive (ANN) analyses, capturing both linear and non-linear relationships (Leong et al., 2019). This dual approach responds to calls for more robust methodologies in marketing research, particularly in complex digital contexts (Shmueli et al., 2019; Hair et al., 2017).

By addressing the identified research gaps, this study provides actionable insights for game developers, marketers, and platform providers targeting Vietnamese MMORPG players. The focus on Vietnam’s unique market, integration of TAM and PERVAL, inclusion of context-specific factors (BLOCK, MPE), and use of an integrated PLS-SEM and ANN approach advance digital consumption research in non-Western contexts. The findings highlight opportunities to enhance blockchain features (tradable NFTs) and optimize mobile payment systems (MoMo, ZaloPay), while leveraging social dynamics in Vietnam’s collectivist culture and segmenting players based on income and experience to maximize purchase intention.

This study is organized into seven sections. “Introduction” introduces the study, outlining the background, objectives, research questions, significance, and structure. “Literature review” reviews the literature on MMORPGs, virtual goods purchases, TAM, PERVAL, and prior studies, identifying research gaps and develops the hypotheses, presenting the hybrid TAM and PERVAL model, and justifying each hypothesis. “Methodologies” details the methodologies, including data collection, measurement instruments, and data analysis using PLS-SEM and ANN. “Findings” presents the findings, covering construct validity, hypothesis testing, mediation, moderation, and ANN results. “Discussion” discusses the key findings, theoretical contributions, and managerial implications. “Conclusion, limitations, and future of research” concludes the study, addressing limitations and suggesting future research directions.

Literature review

The global gaming industry has experienced significant growth in recent years, with Massively Multiplayer Online Games (MMORPGs) emerging as a prominent segment due to their ability to offer immersive, social, and interactive experiences. Within this context, the purchase of virtual goods has become a critical revenue stream for game developers, particularly in emerging markets like Vietnam, where mobile gaming and blockchain technology are transforming player behavior. This literature review synthesizes existing research on MMORPGs and virtual goods purchases, examines the theoretical frameworks of the Technology Acceptance Model (TAM) and PERVAL model, explores the factors influencing purchase intention, and identifies research gaps that this study aims to address.

MMORPGs and virtual goods purchases

Massively Multiplayer Online Games (MMORPGs) are a genre of online games that allow thousands of players to interact in a shared virtual world, engaging in activities such as questing, battling, and socializing (Steinkuehler and Williams, 2006). MMORPGs like World of Warcraft, Final Fantasy XIV, and Black Desert Online have gained global popularity due to their immersive gameplay, rich narratives, and social features, which foster a sense of community among players (Yee, 2006). In recent years, the monetization of MMORPGs has shifted from subscription-based models to freemium models, where players can access the game for free but are encouraged to purchase virtual goods—such as in-game currency, skins, weapons, and cosmetic items—to enhance their experience (Hamari and Keronen, 2017).

Virtual goods purchases have become a significant revenue stream for the gaming industry, contributing billions of dollars annually. According to Newzoo (2025), global in-game purchases reached $188.9 billion in 2025, with MMORPGs accounting for a substantial share due to their large player bases and long-term engagement. Virtual goods serve multiple purposes in MMORPGs: they can provide functional benefits (improving gameplay performance), aesthetic enhancements (customizing avatars), or social status (signaling prestige within the community) (Lehdonvirta, 2009). The rise of blockchain technology has further transformed virtual goods purchases, enabling players to own and trade items as Non-Fungible Tokens (NFTs), which can have real-world economic value (Hamari and Keronen, 2017). Games like Axie Infinity, a blockchain-based MMORPG, have popularized this trend, particularly in emerging markets where players can earn income through NFT trading (Newzoo, 2025).

In Vietnam, the MMORPG market has experienced rapid growth, driven by the country’s young, tech-savvy population and increasing smartphone penetration. With over 45 million gamers—nearly half of Vietnam’s population—the country ranks among the top gaming markets in Southeast Asia (Newzoo, 2025). Popular MMORPGs in Vietnam include Liên Minh Huyền Thoại (League of Legends), Võ Lâm Truyền Kỳ, and Axie Infinity, which have attracted millions of players due to their accessibility on mobile devices and social features. The adoption of mobile MMORPGs has been facilitated by widespread mobile payment systems like MoMo and ZaloPay, which make in-game purchases seamless (Kim et al., 2010). Additionally, Axie Infinity has introduced a new dimension to virtual goods purchases in Vietnam, as players can buy, trade, and earn through blockchain-based assets, creating a play-to-earn economy that has gained significant traction (Hamari and Keronen, 2017).

Despite the economic and cultural significance of virtual goods purchases in Vietnam’s MMORPG market, there is a paucity of research examining the factors influencing purchase intention in this context. Existing studies have primarily focused on Western markets, such as the USA (Johnson and Carter, 2021), Germany (Müller and Braun, 2022), and Italy (Bianchi and Rossi, 2023), with limited attention to non-Western, emerging markets like Vietnam. Furthermore, the unique characteristics of Vietnam’s gaming landscape—such as the dominance of mobile MMORPGs, the prevalence of mobile payments, and the rise of blockchain gaming—introduce context-specific factors that remain underexplored in the literature.

Theoretical frameworks

Technology Acceptance Model (TAM)

The Technology Acceptance Model (TAM), developed by Davis (1989), is a widely used framework for understanding technology adoption and usage behavior. TAM posits that two primary constructs—Perceived Usefulness (PU) and Perceived Ease of Use (PEOU)—determine an individual’s intention to use a technology. PU refers to the degree to which a person believes that using a technology will enhance their performance, while PEOU reflects the perceived ease of using the technology (Davis, 1989). In the context of MMORPGs, PU relates to players’ belief that virtual goods improve their gaming experience (by providing competitive advantages), while PEOU pertains to the ease of purchasing these goods through in-game systems or payment platforms.

TAM has evolved beyond its original organizational focus through extensions like TAM2 (Venkatesh and Davis, 2000) and TAM3 (Venkatesh and Bala, 2008), which incorporate factors such as social influence and hedonic motivations, making it more applicable to voluntary consumer contexts. TAM has been extensively applied to study consumer behavior in digital contexts, including online gaming and virtual goods purchases. For example, Leong et al. (2019) further extended TAM to predict mobile wallet usage, highlighting its relevance to digital transactions in consumer markets. Johnson and Carter (2021) found that PU significantly influences purchase intention in MMORPGs in the USA, as players value virtual goods that enhance gameplay. Similarly, Smith and Jones (2020) reported that PEOU drives purchase intention by reducing barriers to buying virtual goods, such as through user-friendly payment interfaces. In Vietnam, where mobile MMORPGs dominate, PEOU is particularly relevant due to the widespread use of mobile payment systems, which simplify in-game transactions (Kim et al., 2010). Recent extensions of TAM in non-Western settings align with our approach (Zhang et al., 2025, on e-commerce in China; Chen et al., 2025, on tech attitudes). However, TAM’s focus on functional aspects (PU, PEOU) may not fully capture the value-based and emotional drivers of purchase intention in MMORPGs, necessitating the integration of additional frameworks like PERVAL.

PERVAL model

The PERVAL model, developed by Sweeney and Soutar (2001), is a multidimensional framework for understanding perceived value in consumer behavior. PERVAL identifies four key dimensions of perceived value: Emotional Value (VE), Social Value (VS), Price/Value for Money (VP), and Performance/Quality Value (VQ). VE captures the emotional benefits of a product, such as enjoyment or excitement; VS reflects the social recognition or status gained; VP refers to the perceived affordability or worth relative to cost; and VQ pertains to the perceived quality, durability, or utility of the product (Sweeney and Soutar, 2001). In the context of MMORPGs, these dimensions are highly relevant: virtual goods can evoke emotional responses (VE), enhance social status (VS), be evaluated for affordability (VP), and be assessed for quality or utility (VQ).

The PERVAL model has been widely applied to study virtual goods purchases in MMORPGs. For instance, Müller and Braun (2022) found that VQ significantly influences purchase intention in Germany, as players value high-quality virtual goods that enhance gameplay. Similarly, Torres and Diaz (2022) reported that VE drives purchase intention in Colombia, with players seeking emotional satisfaction from virtual goods. Additionally, Hassan and Ali (2023) found that hedonic motivation, akin to emotional value, drives continuous purchase intention in Saudi Arabian MMORPGs, highlighting the role of enjoyment in player behavior. In Vietnam, where collectivism shapes social interactions, VS is particularly relevant, as players may purchase virtual goods to gain recognition within their gaming community (Kaya and Ozkan, 2022). However, PERVAL’s focus on value perceptions may overlook the functional aspects of technology adoption, such as ease of use, which are critical in digital contexts like MMORPGs.

Hybrid TAM and PERVAL model

The integration of TAM and PERVAL into a hybrid model is necessitated by the complex nature of virtual goods purchase intention in MMORPGs, which involves both technological adoption (purchasing through in-game systems or mobile payment platforms) and value-based motivations (emotional satisfaction, social status, and economic considerations). TAM’s focus on Perceived Usefulness (PU) and Perceived Ease of Use (PEOU) captures the functional aspects of purchasing virtual goods, such as the ease of using payment systems and the perceived benefits of in-game items (Davis, 1989). However, TAM alone may not fully account for the hedonic and social drivers prevalent in voluntary, consumer-driven contexts like MMORPGs. The PERVAL model complements TAM by incorporating Emotional Value (VE), Social Value (VS), Price/Value for Money (VP), and Performance/Quality Value (VQ), which are critical for understanding players’ motivations to purchase virtual goods that enhance enjoyment, social status, or perceived affordability (Sweeney and Soutar, 2001). By combining these frameworks, the hybrid model provides a comprehensive lens to capture the multifaceted drivers of purchase intention in Vietnam’s MMORPG market.

To address the multifaceted nature of virtual goods purchase intention in Vietnam’s MMORPG market, this study integrates TAM and PERVAL into a hybrid model. This approach combines TAM’s focus on technological adoption (PU, PEOU) with PERVAL’s emphasis on perceived value (VE, VS, VP, VQ), providing a comprehensive framework to capture both functional and value-based drivers in a mobile-driven, collectivist context. Prior studies from Appendix 1 Prior studies have demonstrated the efficacy of combining TAM and PERVAL or similar frameworks in digital contexts. For instance, Leong et al. (2019) integrated TAM with UTAUT to study mobile wallet adoption, highlighting the need for both functional and value-based constructs. For example, Johnson and Carter (2021) found that PU and VS jointly influence purchase intention in the USA, as players value both functional benefits and social recognition. Similarly, Müller and Braun (2022) reported that PEOU and VQ drive purchase intention in Germany, highlighting the interplay between ease of use and quality perceptions.

The hybrid TAM-PERVAL model is particularly relevant in Vietnam, where the MMORPG market is characterized by mobile-driven gameplay, widespread mobile payment adoption (e.g., MoMo, ZaloPay), and the rise of blockchain-based games like Axie Infinity. These factors introduce both technological (ease of payment, blockchain utility) and value-based (emotional satisfaction, social status) considerations that require a combined framework to fully understand purchase behavior. Additionally, Vietnam’s collectivist culture amplifies the role of social value (VS) and social influence (SI), which PERVAL captures effectively, while TAM addresses the technological barriers and benefits of virtual goods purchases. For instance, the ease of using mobile payment systems (PEOU) and the perceived affordability of virtual goods (VP) are critical in Vietnam’s mobile-driven market, while emotional satisfaction (VE) and social status (VS) resonate with the country’s collectivist culture (Kim et al., 2010; Kaya and Ozkan, 2022). Applying TAM alone may overlook the emotional and social motivations that drive virtual goods purchases in MMORPGs, such as the desire for self-expression through aesthetic items (VE) or social recognition within gaming communities (VS) (Hamari and Keronen, 2017). Conversely, PERVAL may not adequately address the technological barriers, such as the complexity of payment systems (PEOU), that influence purchase decisions in a mobile-driven market like Vietnam (Kim et al., 2010). The hybrid model mitigates these limitations by integrating functional, emotional, social, and economic drivers, providing a more holistic understanding of purchase intention. By integrating TAM and PERVAL, this study aims to provide a holistic understanding of virtual goods purchase intention, addressing the limitations of each model when applied independently. Figure 1 illustratives research proposed model.

Fig. 1
figure 1

Research proposed model based on PERVAL and TAM constructs.

Full size image

Factors influencing purchase intention

The literature identifies several factors influencing players’ intention to purchase virtual goods in MMORPGs, which are incorporated into the hybrid TAM and PERVAL model. These factors can be categorized into TAM constructs, PERVAL dimensions, context-specific factors, mediators, and moderators.

TAM constructs

Perceived Usefulness (PU): PU has been consistently identified as a key driver of purchase intention in MMORPGs. Smith and Jones (2020) found that PU positively affects purchase intention in the USA, as players value virtual goods that enhance their gameplay performance. Similarly, Johnson and Carter (2021) reported a significant effect of PU on purchase intention, emphasizing its role in providing competitive advantages.

Perceived Ease of Use (PEOU): PEOU also influences purchase intention by reducing barriers to buying virtual goods. Miller and Clark (2020) demonstrated that PEOU positively affects purchase intention in the USA, as a user-friendly purchase process encourages buying behavior. Evans and White (2020) found similar results, noting that PEOU facilitates avatar customization purchases in North America.

PERVAL dimensions

Emotional Value (VE): VE captures the emotional benefits of virtual goods, such as enjoyment or escapism. Torres and Diaz (2022) found that VE significantly influences purchase intention in Colombia, as players seek emotional satisfaction from virtual goods. Silva and Almeida (2020) reported similar findings in Brazil, highlighting the role of emotional attachment in driving purchases.

Social Value (VS): VS reflects the social recognition gained from virtual goods. Kaya and Ozkan (2022) found that VS drives purchase intention in Turkey, as players value items that enhance their social status within the gaming community. Bianchi and Rossi (2023) confirmed this effect in Italy, noting the importance of social bonds in MMORPGs.

Price/Value for Money (VP): VP refers to the perceived affordability of virtual goods. Almeida and Souza (2021) reported that VP positively affects purchase intention in Chile, with income level moderating this relationship. Aksoy and Yilmaz (2022) found similar results in Turkey, emphasizing economic constraints in purchase decisions.

Performance/Quality Value (VQ): VQ pertains to the perceived quality or utility of virtual goods. Müller and Braun (2022) found that VQ significantly influences purchase intention in Germany, as players value high-quality items that enhance gameplay. Schmidt and Braun (2022) confirmed this effect in the Netherlands, noting the importance of item durability.

Context-specific factors

Trust in Platform (TRUST): Trust in the security and fairness of the purchasing system is critical in online contexts. Ozdemir and Turkmen (2021) found that TRUST positively affects purchase intention in Turkey, as players need assurance of secure transactions. Dupont and Martin (2021) reported similar findings in France, emphasizing platform reliability.

Game Immersion (IMM): IMM reflects the level of engagement in MMORPGs. Perez and Gomez (2020) found that IMM drives purchase intention in Mexico, as immersive experiences motivate players to buy virtual goods. Klein and Huber (2016) confirmed this effect in Germany, linking immersion to flow experiences.

Aesthetic Appeal (AES): AES captures the visual attractiveness of virtual goods. Santos and Almeida (2023) found that AES influences purchase intention in Argentina, as players value aesthetically pleasing items. Lopez and Vega (2020) reported similar findings in Peru, emphasizing visual appeal in MMORPGs. Similarly, Brown and Taylor (2019) found that avatar identification, closely tied to aesthetic appeal, drives purchase intention in Canadian MMORPGs, as players seek visually distinctive items to enhance self-expression.

Blockchain Utility (BLOCK): BLOCK reflects the value of blockchain-based virtual goods (NFTs). While not directly studied in Appendix 1, Hamari and Keronen (2017) suggest that blockchain features, such as tradability, can drive purchase intention, particularly in markets like Vietnam where Axie Infinity is popular.

Mobile Payment Ease (MPE): MPE refers to the ease of using mobile payment systems. Kim et al. (2010) found that ease of mobile payments influences purchase intention in digital contexts, a factor particularly relevant in Vietnam’s mobile-driven market.

Mediators

Satisfaction (SAT): SAT mediates the relationship between perceived benefits and purchase intention. Aydin and Kaya (2020) found that SAT mediates the effect of emotional factors on purchase intention in Qatar, as satisfaction with the gaming experience drives buying behavior.

Social Influence (SI): SI, encompassing peer pressure and community norms, mediates the effect of social value on purchase intention. Fischer and Weber (2023) reported that community interaction influences purchase intention in Austria, while Costa and Silva (2021) found similar effects in Brazil. Similarly, Wilson and Adams (2023) found that social presence enhances willingness to pay in Canadian MMORPGs, reinforcing the role of community-driven social influence in purchase decisions.

Moderators

Income Level (INC): Income Level (INC) moderates the effect of value perceptions and perceived usefulness on purchase intention, strengthening these relationships. Higher income levels enable players to act on their perceptions of Price/Value for Money (VP) and Perceived Usefulness (PU), making them more likely to purchase virtual goods. Almeida and Souza (2021) found that higher income strengthens the effect of VP on purchase intention in Chile, as players with greater financial resources perceive virtual goods as more affordable, increasing their purchase likelihood. Similarly, Kaya and Ozkan (2022) reported that higher income enhances the impact of value perceptions on purchase decisions in Turkey, supporting the strengthening role of INC in the VP → PI and PU → PI relationships.

Gaming Experience (EXP): Gaming Experience (EXP) moderates the effect of Perceived Ease of Use (PEOU) on purchase intention, weakening this relationship for more experienced players. Less experienced players, who are less familiar with in-game purchasing systems, are more influenced by the ease of use of these systems, making PEOU a stronger driver of their purchase intention. In contrast, experienced players, who are accustomed to navigating complex game interfaces, rely less on ease of use, reducing its impact on their purchase decisions. Bianchi and Rossi (2023) found that less experienced players in Italy are more influenced by PEOU, while Garcia and Lopez (2021) reported similar effects in Spain, supporting the weakening role of EXP in the PEOU → PI relationship.

The literature on virtual goods purchases in MMORPGs reveals several critical research gaps, aligning with the gaps identified in the introduction. First, the majority of studies focus on Western markets, with limited research on non-Western, emerging markets like Vietnam, despite its 45 million gamers and growing economic significance (Newzoo, 2025; Johnson and Carter, 2021; Müller and Braun, 2022; Bianchi and Rossi, 2023).

Second, while TAM and PERVAL have been applied individually, their integration into a hybrid model to capture both functional and value-based drivers is rare, particularly in collectivist cultures like Vietnam (Kaya and Ozkan, 2022). Third, context-specific factors such as Blockchain Utility (BLOCK) and Mobile Payment Ease (MPE), critical in Vietnam’s mobile-driven and blockchain-influenced market, remain underexplored (Kim et al., 2010; Hamari and Keronen, 2017). Fourth, the limited use of integrated PLS-SEM and ANN methodologies to model both linear and non-linear relationships represents a methodological gap (Leong et al., 2019). This study addresses these gaps by examining Vietnam’s MMORPG market through a hybrid TAM-PERVAL model and an integrated analytical approach.

Second, while TAM and PERVAL have been applied individually to study virtual goods purchases, their integration into a hybrid model is rare, particularly in non-Western contexts. The hybrid TAM and PERVAL model offers a more comprehensive framework by combining functional (PU, PEOU) and value-based (VE, VS, VP, VQ) factors, but its applicability in Vietnam remains untested (Sweeney and Soutar, 2001; Davis, 1989).

Third, the role of social influence in collectivist cultures like Vietnam is underexplored. While studies have examined social value (VS) and community engagement (Fischer and Weber, 2023; Costa and Silva, 2021), the broader construct of Social Influence (SI), encompassing peer pressure and community norms, has received less attention, particularly in Vietnam, where collectivism shapes social interactions (Kaya and Ozkan, 2022).

Fourth, traditional linear methods like PLS-SEM may not fully capture the non-linear relationships inherent in complex consumer behavior, such as the diminishing returns of blockchain utility or the threshold effects of mobile payment ease (Hair et al., 2017). An integrated PLS-SEM and ANN approach can address this gap by combining explanatory (PLS-SEM) and predictive (ANN) analyses, but such an approach is rarely applied in MMORPG research (Leong et al., 2019).

Finally, the moderating effects of economic (INC) and experiential (EXP) factors in emerging markets are underexplored. While studies have examined income’s role in Western contexts (Almeida and Souza, 2021), its impact in Vietnam, where income disparities are significant, remains unclear. Similarly, the moderating effect of gaming experience on ease of use (PEOU) needs further exploration in mobile-driven markets (Bianchi and Rossi, 2023).

This study addresses these gaps by applying a hybrid TAM and PERVAL model to evaluate virtual goods purchase intention in Vietnam, incorporating context-specific factors (BLOCK, MPE), examining social influence in a collectivist culture, and using an integrated PLS-SEM and ANN approach to capture both linear and non-linear relationships.

Hypotheses development

This Section presents the conceptual framework and develops the hypotheses for the study, which aims to evaluate the factors influencing players’ intention to purchase virtual goods in Massively Multiplayer Online Games (MMORPGs) in Vietnam using a hybrid TAM and PERVAL model with an integrated PLS-SEM and ANN approach. The framework integrates constructs from the Technology Acceptance Model (TAM) and PERVAL model, along with context-specific factors, mediators, and moderators, to provide a comprehensive understanding of purchase intention in this unique market.

Conceptual framework

The conceptual framework integrates the Technology Acceptance Model (TAM) (Davis, 1989) and the PERVAL model (Sweeney and Soutar, 2001) to evaluate the factors influencing players’ intention to purchase virtual goods in MMORPGs in Vietnam. TAM provides a functional perspective through Perceived Usefulness (PU) and Perceived Ease of Use (PEOU), which capture the technological aspects of virtual goods purchases. The PERVAL model adds a value-based perspective through Emotional Value (VE), Social Value (VS), Price/Value for Money (VP), and Performance/Quality Value (VQ), which reflect the emotional, social, economic, and quality-related benefits of virtual goods. Context-specific factors relevant to Vietnam’s MMORPG market—Trust in Platform (TRUST), Game Immersion (IMM), Aesthetic Appeal (AES), Blockchain Utility (BLOCK), and Mobile Payment Ease (MPE)—are also included to address the unique characteristics of this context, such as the rise of blockchain gaming and mobile payment systems.

The framework further incorporates mediators—Satisfaction (SAT) and Social Influence (SI)—to capture the mechanisms through which PU, VE, and VS influence Purchase Intention (PI). Satisfaction (SAT) mediates the relationships between PU and PI, and VE and PI, reflecting the role of overall contentment with the gaming experience. Social Influence (SI), encompassing peer pressure and community norms, mediates the relationship between VS and PI, reflecting Vietnam’s collectivist culture. Moderators—Income Level (INC) and Gaming Experience (EXP)—are included to examine how economic and experiential factors influence the effects of VP, PU, and PEOU on PI. The framework posits that these factors collectively influence PI, with PLS-SEM testing the theoretical relationships and ANN capturing non-linear effects.

Hypotheses development

Perceived Usefulness (PU), a core construct of TAM, refers to the degree to which individuals believe a technology enhances their performance (Davis, 1989). In MMORPGs, PU reflects players’ belief that virtual goods improve their gaming experience, such as by providing competitive advantages or enhancing gameplay. Studies have consistently shown that PU positively influences purchase intention in digital contexts, as players are more likely to buy virtual goods they perceive as beneficial (Johnson and Carter, 2021; Smith and Jones, 2020; Müller and Braun, 2022). For example, Johnson and Carter (2021) found that PU is particularly relevant in the voluntary context of MMORPGs, as players assess whether virtual goods provide tangible benefits, such as enhanced gameplay or tradable value, justifying TAM’s application. Similarly, Smith and Jones (2020) reported a strong effect of PU on purchase intention in North America, emphasizing its role in enhancing gameplay outcomes. In Asian contexts, Lee and Kim (2019) demonstrated that PU drives in-game purchases in South Korea, where players prioritize functional benefits. In Vietnam, where competitive MMORPGs like Liên Minh Huyền Thoại are popular, PU is expected to drive purchase intention, particularly for items that offer tangible benefits, such as tradable NFTs in Axie Infinity (Hamari and Keronen, 2017; Leong et al., 2019; Garcia and Lopez, 2021). Therefore, the following hypothesis is suggested:

H1: Perceived Usefulness (PU) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Perceived Ease of Use (PEOU), another TAM construct, refers to the degree to which a technology is perceived as easy to use (Davis, 1989). In MMORPGs, PEOU relates to the simplicity of purchasing virtual goods through in-game systems or payment platforms. A user-friendly purchase process reduces barriers to buying, thereby increasing purchase intention (Miller and Clark, 2020; Evans and White, 2020; Bianchi and Rossi, 2023). Miller and Clark (2020) found that PEOU positively affects purchase intention in the USA, as an effortless purchase process encourages buying behavior. Evans and White (2020) reported similar findings in North America, noting that PEOU facilitates avatar customization purchases. In Vietnam, where mobile MMORPGs dominate, PEOU is critical in a voluntary purchase context and the widespread use of mobile payment systems like MoMo and ZaloPay, which simplify transactions (Kim et al., 2010; Leong et al., 2019; Garcia and Lopez, 2021). Therefore, the following hypothesis is suggested:

H2: Perceived Ease of Use (PEOU) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Emotional Value (VE), a dimension of the PERVAL model, captures the feelings of enjoyment, excitement, or escapism derived from a product (Sweeney and Soutar, 2001). In MMORPGs, virtual goods often evoke emotional responses, such as joy from owning a rare skin, directly influencing purchase intention (Johnson and Carter, 2021; Torres and Diaz, 2022; Silva and Almeida, 2020). Johnson and Carter (2021) found that VE significantly affects purchase intention in the USA, as players seek emotional satisfaction from virtual goods. Torres and Diaz (2022) reported similar findings in Colombia, emphasizing emotional connections to virtual items. Additionally, Satisfaction (SAT) with the gaming experience can mediate this relationship, as emotional value increases overall satisfaction, which in turn drives purchase behavior (Aydin and Kaya, 2020; Dupont and Martin, 2021; Garcia and Lopez, 2021). Aydin and Kaya (2020) found that SAT mediates the effect of emotional factors on purchase intention in Qatar, as satisfaction enhances buying behavior. In Vietnam, where gaming is a popular form of escapism, VE and SAT are expected to play significant roles in driving virtual goods purchases (Leong et al., 2019). Therefore, the following hypotheses are suggested:

H3a: Emotional Value (VE) positively affects Purchase Intention (PI) of virtual goods in MMORPGs. H3b: Satisfaction (SAT) positively affects Purchase Intention (PI), partially mediating the relationship between Emotional Value (VE) and Purchase Intention (PI).

Social Value (VS), another PERVAL dimension, refers to the social recognition or status gained from a product (Sweeney and Soutar, 2001). In MMORPGs, virtual goods often serve as status symbols, enhancing players’ social standing within their gaming community. This social value can drive Social Influence (SI), which encompasses peer pressure and community norms that encourage players to conform to group behaviors, such as purchasing virtual goods (Kaya and Ozkan, 2022; Fischer and Weber, 2023; Costa and Silva, 2021). Kaya and Ozkan (2022) found that VS influences purchase intention through peer pressure in Turkey, as players value items that enhance their status. Fischer and Weber (2023) reported that community interaction drives purchase intention in Austria, while Costa and Silva (2021) found similar effects in Brazil. SI, in turn, mediates the effect of VS on purchase intention, as players are motivated by their peers or community to buy virtual goods (Bianchi and Rossi, 2023; Aydin and Kaya, 2020; Garcia and Lopez, 2021). In Vietnam’s collectivist culture, where social bonds and peer influence are strong, this relationship is particularly relevant (Leong et al., 2019). Thus, the following hypotheses are suggested:

H4a: Social Value (VS) positively affects Social Influence (SI) in MMORPGs. H4b: Social Influence (SI) positively affects Purchase Intention (PI), fully mediating the relationship between Social Value (VS) and Purchase Intention (PI).

Price/Value for Money (VP), a PERVAL dimension, reflects the perceived affordability or worth of a product relative to its cost (Sweeney and Soutar, 2001). In MMORPGs, players assess whether virtual goods offer good value for their price, influencing their purchase decisions. Income Level (INC) moderates this relationship, strengthening the effect of VP on purchase intention (PI) for players with higher income. Higher income levels reduce financial constraints, enabling players to act on their perceptions of affordability, making them more likely to purchase virtual goods. Almeida and Souza (2021) found that VP positively affects purchase intention in Chile, with higher income strengthening this relationship as players perceive virtual goods as more affordable. Kaya and Ozkan (2022) reported similar findings in Turkey, noting that higher income enhances the impact of value perceptions on purchase decisions. In Vietnam, an emerging economy with varying income levels, affordability perceptions are critical, especially for microtransactions (Leong et al., 2019; Aydin and Kaya, 2020; Garcia and Lopez, 2021). Therefore, the following hypotheses are suggested:

H5a: Price/Value for Money (VP) positively affects Income Level (INC). H5b: Income Level (INC) positively affects Purchase Intention (PI), moderating the relationship between Price/Value for Money (VP) and Purchase Intention (PI).

Performance/Quality Value (VQ), also from PERVAL, refers to the perceived quality, durability, or utility of a product (Sweeney and Soutar, 2001). In MMORPGs, high-quality virtual goods (rare items, durable gear) enhance gameplay or aesthetics, directly influencing purchase intention. Players are more likely to buy virtual goods that offer reliable performance or meet their expectations (Müller and Braun, 2022; Schmidt and Braun, 2022; Santos and Almeida, 2023). Müller and Braun (2022) found that VQ significantly influences purchase intention in Germany, as players value high-quality items. Schmidt and Braun (2022) confirmed this effect in the Netherlands, emphasizing item durability. In Vietnam, where players value rarity in mobile and blockchain games, quality is a key driver (Leong et al., 2019; Aydin and Kaya, 2020; Garcia and Lopez, 2021). Thus, the following hypothesis is suggested:

H6: Performance/Quality Value (VQ) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Trust in Platform (TRUST) reflects players’ confidence in the security and fairness of the MMORPG’s purchasing system. In online contexts, trust is a critical determinant of purchase intention, as players need assurance that their transactions are secure and the platform is reliable (Ozdemir and Turkmen, 2021; Dupont and Martin, 2021; Aksoy and Yilmaz, 2022). Ozdemir and Turkmen (2021) found that TRUST positively affects purchase intention in Turkey, as players require secure transactions. Dupont and Martin (2021) reported similar findings in France, emphasizing platform reliability. In Vietnam, where digital payments and blockchain games (Axie Infinity) are prevalent, trust is essential (Leong et al., 2019; Aydin and Kaya, 2020; Garcia and Lopez, 2021). Therefore, the following hypothesis is suggested:

H7: Trust in Platform (TRUST) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Game Immersion (IMM) refers to the level of engagement or absorption players experience in MMORPGs. Immersive experiences, such as rich storylines or engaging gameplay, can motivate players to purchase virtual goods to enhance their involvement (Perez and Gomez, 2020; Klein and Huber, 2016; Silva and Almeida, 2020). Perez and Gomez (2020) found that IMM drives purchase intention in Mexico, as immersive experiences encourage buying. Klein and Huber (2016) confirmed this effect in Germany, linking immersion to flow experiences. In Vietnam, where MMORPGs serve as a popular escape, immersion is likely to drive purchase behavior (Leong et al., 2019; Aydin and Kaya, 2020; Garcia and Lopez, 2021). Thus, the following hypothesis is suggested:

H8: Game Immersion (IMM) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Aesthetic Appeal (AES) captures the visual attractiveness of virtual goods, such as skins or avatars. In MMORPGs, aesthetically pleasing items often drive purchase intention by enhancing players’ visual experience and self-expression (Santos and Almeida, 2023; Lopez and Vega, 2020; Evans and White, 2020). Santos and Almeida (2023) found that AES influences purchase intention in Argentina, as players value visually appealing items. Lopez and Vega (2020) reported similar findings in Peru, emphasizing visual appeal in MMORPGs. In Vietnam, where mobile MMORPGs emphasize aesthetics, AES is a significant motivator (Leong et al., 2019; Aydin and Kaya, 2020; Garcia and Lopez, 2021). Therefore, the following hypothesis is suggested:

H9: Aesthetic Appeal (AES) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Blockchain Utility (BLOCK) reflects the perceived value of virtual goods tied to blockchain or NFT features, such as tradability and ownership. In MMORPGs, blockchain-based goods (NFTs in Axie Infinity) offer unique benefits, such as the ability to trade items for real-world value, which can drive purchase intention (Hamari and Keronen, 2017; Leong et al., 2019). While not directly studied in Appendix 1, Hamari and Keronen (2017) suggest that blockchain features enhance purchase intention by adding economic value. Serada et al. (2021) found that blockchain-based games increase player engagement and spending due to the perceived real-world value of NFTs, particularly in Southeast Asia. Similarly, Nguyen and Tran (2022) reported that Vietnamese players value blockchain utility in Axie Infinity, as it enables income generation through NFT trading. In Vietnam, where blockchain gaming is booming, particularly in the South, this utility is a key motivator (Aydin and Kaya, 2020; Garcia and Lopez, 2021; Johnson and Carter, 2021; Müller and Braun, 2022). Thus, the following hypothesis is suggested:

H10: Blockchain Utility (BLOCK) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Mobile Payment Ease (MPE) refers to the ease of using mobile payment systems (MoMo, ZaloPay) to purchase virtual goods. In digital contexts, ease of payment is a critical driver of purchase intention, as a seamless payment process reduces barriers to buying (Kim et al., 2010; Leong et al., 2019). Kim et al. (2010) found that ease of mobile payments influences purchase intention in digital contexts, a factor particularly relevant in Vietnam’s mobile-driven market. In MMORPGs, a user-friendly payment process encourages purchases (Miller and Clark, 2020; Evans and White, 2020; Aydin and Kaya, 2020; Garcia and Lopez, 2021). Therefore, the following hypothesis is suggested:

H11: Mobile Payment Ease (MPE) positively affects Purchase Intention (PI) of virtual goods in MMORPGs.

Satisfaction (SAT) reflects players’ overall contentment with their MMORPG experience. In consumer behavior, satisfaction often mediates the relationship between perceived benefits and behavioral outcomes, such as purchase intention (Aydin and Kaya, 2020; Dupont and Martin, 2021). Specifically, perceived usefulness (PU) of virtual goods can enhance satisfaction with the gaming experience, which in turn drives purchase intention (Johnson and Carter, 2021; Müller and Braun, 2022). Aydin and Kaya (2020) found that SAT mediates the effect of emotional factors on purchase intention in Qatar. In Vietnam, where players value functional benefits (competitive advantages), satisfaction is likely to mediate this relationship (Leong et al., 2019; Garcia and Lopez, 2021; Bianchi and Rossi, 2023). Thus, the following hypotheses are suggested:

H12a: Satisfaction (SAT) positively affects Purchase Intention (PI), partially mediating the relationship between Perceived Usefulness (PU) and Purchase Intention (PI). H12b: Satisfaction (SAT) positively affects Purchase Intention (PI).

Gaming Experience (EXP) refers to the duration and intensity of a player’s engagement with MMORPGs. Experienced players, being more familiar with in-game systems, may rely less on ease of use compared to novices. EXP moderates the effect of Perceived Ease of Use (PEOU) on purchase intention, weakening this relationship for more experienced players. Less experienced players, who face greater challenges navigating in-game purchasing systems, are more influenced by PEOU, making it a stronger driver of their purchase intention. In contrast, experienced players, accustomed to complex interfaces, place less emphasis on ease of use, reducing its impact on their purchase decisions. Bianchi and Rossi (2023) found that less experienced players in Italy are more influenced by PEOU, while Garcia and Lopez (2021) reported similar findings in Spain, supporting the weakening effect of EXP. In Vietnam, where MMORPG players range from casual to hardcore, EXP is expected to influence how ease of use affects purchase behavior (Leong et al., 2019; Aydin and Kaya, 2020; Evans and White, 2020). Therefore, the following hypotheses are suggested:

H13a: Gaming Experience (EXP) moderates the relationship between Perceived Ease of Use (PEOU) and Purchase Intention (PI). H13b: Gaming Experience (EXP) positively affects Purchase Intention (PI).

Within the TAM framework, PEOU is a determinant of PU, as ease of use enhances the perceived benefits of a technology (Davis, 1989). In MMORPGs, if purchasing virtual goods is easy (through a user-friendly interface), players are more likely to perceive these goods as useful, indirectly affecting purchase intention (Smith and Jones, 2020; Müller and Braun, 2022; Miller and Clark, 2020). Smith and Jones (2020) found that PEOU enhances PU in the USA, as ease of use increases perceived benefits. Müller and Braun (2022) reported similar findings in Germany, emphasizing this relationship in MMORPGs. In Vietnam, where mobile payment systems simplify transactions, this effect is particularly relevant (Leong et al., 2019; Aydin and Kaya, 2020; Garcia and Lopez, 2021). Thus, the following hypothesis is suggested:

H14: Perceived Ease of Use (PEOU) positively affects Perceived Usefulness (PU).

Perceived Usefulness (PU) can also influence how income level (INC) affects purchase decisions. Players who perceive virtual goods as useful (e.g., enhancing gameplay or offering tradable value) are more likely to consider their income in the context of purchasing, as the perceived benefits justify the expenditure (Johnson and Carter, 2021; Müller and Braun, 2022). Income Level (INC) moderates the relationship between PU and purchase intention, strengthening this effect for higher-income players. Higher income levels enable players to act on their perceptions of usefulness, increasing their likelihood of purchasing virtual goods that offer tangible benefits. Almeida and Souza (2021) found that higher income strengthens the effect of value perceptions on purchase intention in Chile, a finding applicable to PU as players prioritize useful items. Kaya and Ozkan (2022) reported similar findings in Turkey, noting that higher income reduces economic constraints in purchase decisions. In Vietnam, where income disparities exist, this strengthening moderation effect is particularly relevant (Leong et al., 2019; Aydin and Kaya, 2020). Therefore, the following hypotheses are suggested:

H15a: Perceived Usefulness (PU) positively affects Income Level (INC). H15b: Income Level (INC) positively affects Purchase Intention (PI), moderating the relationship between Perceived Usefulness (PU) and Purchase Intention (PI).

The hypothesized relationships are summarized in Table 1. Figure 1 represents the hypotheses being proposed.

Table 1 Summary of Hypothesized Relationships.
Full size table

Methodologies

This study employed a quantitative research design to evaluate the factors influencing players’ intention to purchase virtual goods in Massively Multiplayer Online Games (MMORPGs) in Vietnam, using an integrated PLS-SEM and ANN approach. The methodology encompassed data collection, measurement development, and data analysis, ensuring a robust examination of the proposed hybrid TAM and PERVAL model.

Collecting data and sampling techniques

Data were collected through an online survey targeting Vietnamese MMORPG players across the North, Central, and South regions of Vietnam, aligning with the country’s gaming population distribution (Newzoo, 2025). The sampling frame was constructed using online gaming communities, social media groups (Vietnamese MMORPG forums on Facebook), and gaming events in major cities such as Hanoi, Da Nang, and Ho Chi Minh City. A combination of convenience and snowball sampling was employed, as these methods are effective for reaching niche populations like MMORPG players (Saunders et al., 2016; Yee, 2006). To minimize common method bias (CMB) due to the self-report survey design, several procedural remedies were implemented (Podsakoff et al., 2003). Respondents were assured of anonymity and confidentiality to reduce social desirability bias, survey questions were randomized to avoid order effects, and clear, neutral wording was used to prevent response bias.

A total of 605 questionnaires were distributed between December 2024 and February 2025, using Google Forms to facilitate online responses. To assess potential nonresponse bias, early and late respondents were compared, following established guidelines, and no significant differences were found (Armstrong and Overton, 1977). Of these, 562 questionnaires were returned, yielding a response rate of 92.9%. After screening for incomplete or inconsistent responses (straight-lining, missing data), 539 questionnaires were accepted for quantitative analysis, resulting in an effective response rate of 89.1%. The sample size of 539 was determined using established guidelines for PLS-SEM and ANN to ensure sufficient statistical power. For PLS-SEM, the ‘10-times rule’ was applied, requiring the sample size to be at least 10 times the number of paths directed at the dependent variable (Purchase Intention, PI) (Hair et al., 2017). With 10 direct paths to PI, a minimum sample size of 100–150 was required, which the 539 responses exceed. For ANN, a heuristic of 5–10 times the number of network parameters was used (Haykin, 1999). With a feedforward neural network (15 input nodes, 10 hidden nodes, 1 output node), approximately 400–800 responses were needed, and the 539 responses fall within this range. Additionally, a power analysis using G*Power (Faul et al., 2009) confirmed that a sample size of 539 achieves 80% power at α = 0.05 for detecting medium effect sizes (f² = 0.15) in PLS-SEM, ensuring robust hypothesis testing. The balanced regional distribution (North: 39.9%, Central: 20.0%, South: 40.1%) further supports the sample’s representativeness of Vietnam’s MMORPG player population.

Measurement instruments

The measurement instruments were developed based on established scales from prior literature, adapted to the context of MMORPGs in Vietnam. All constructs were measured using a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree), as this format is widely used in consumer behavior research to capture latent variables (Saunders et al., 2016). The questionnaire was initially developed in English and translated into Vietnamese by a bilingual expert, then back-translated to English to ensure accuracy and cultural relevance (Brislin, 1980). A pilot test with 30 MMORPG players was conducted to refine the items, ensuring clarity and reliability (Cronbach’s Alpha > 0.7 for all constructs). Demographic questions (age, gender, income, gaming experience) were also included to support the analysis of moderating variables (INC, EXP).

Data analysis

Data analysis was conducted in multiple stages using SPSS version 25.0 and SmartPLS 4.0. First, SPSS version 25.0 was used for preliminary tabulation and descriptive statistics, providing an overview of the sample characteristics (regional distribution, purchase history). SPSS was also employed to conduct the ANN analysis, including data normalization, network training, validation, testing, and sensitivity analysis to identify key predictors of Purchase Intention (PI) (Leong et al., 2019). The ANN used a feedforward architecture with one hidden layer (15-10-1), trained on 70% of the data (377 responses), validated on 15% (81 responses), and tested on 15% (81 responses), as detailed in the ANN Results section.

Second, SmartPLS 4.0 was employed to evaluate the reliability and validity of the measurement scales and to test the proposed associations (Hair et al., 2017). The measurement model was assessed for reliability (Cronbach’s Alpha, Composite Reliability), convergent validity (Factor Loadings, AVE), and discriminant validity (Fornell-Larcker criterion, HTMT ratio). The structural model was then tested using PLS-SEM with a bootstrapping procedure (5000 samples) to assess the significance of direct effects (H1, H2, H3a, H4a, H5a, H6, H7, H8, H9, H10, H11, H12b, H13b, H14, H15a), mediation effects (H3b, H4b, H12a), and moderation effects (H5b, H13a, H15b), with 95% bias-corrected confidence intervals. The integrated PLS-SEM and ANN approach enriches the study by addressing both explanatory and predictive objectives. PLS-SEM validates the hybrid TAM-PERVAL model, confirming the significance of theoretical relationships and providing interpretable results for academic research. ANN enhances predictive accuracy, capturing non-linear effects, such as the threshold influence of BLOCK and MPE, which are critical in Vietnam’s mobile-driven and blockchain-influenced MMORPG market. This dual approach provides actionable insights for game developers, such as prioritizing blockchain features and mobile payment systems, while advancing methodological rigor in consumer behavior research (Shmueli et al., 2019).

Findings

Demographic of participants

Below is the Table 2 presenting the demographic characteristics of the 539 participants, with both frequency (number of respondents) and percentage.

Table 2 Demographic characteristics of participants.
Full size table

A total respondents exceeds the recommended sample size calculation, ensuring sufficient power for PLS-SEM and ANN analysis. The distribution (North: 39.9%, Central: 20.0%, South: 40.1%) closely matches the target proportions (40%, 20%, 40%), ensuring balanced representation across Vietnam. Predominantly young (18–34: 75.2%), reflecting the typical MMORPG player demographic in Vietnam. However, the sample is gender-skewed, with 69.9% male (377 participants), 28.0% female (151 participants), and 2.1% other/prefer not to say (11 participants), compared to an estimated 40% female gamers in Vietnam (Newzoo, 2025). This gender imbalance may limit the representativeness of findings, particularly for female players, and is addressed as a limitation. Most players have low to middle income (85.2% below 20 million VND), aligning with Vietnam’s economic context. A mix of new and experienced players, with 65.2% having 1–5 years of experience, typical for MMORPG players. Most players (65.2%) play 5–20 h/week, indicating moderate to high engagement. 59.9% have purchased virtual goods, supporting the study’s focus on purchase intention. Table 3 illustrates the cross-descriptive analysis between region and purchase history of respondents in different regions.

Table 3 Cross-descriptive analysis of region and purchase history.
Full size table

Across all regions, 59.9% of respondents have purchased virtual goods, consistent with the overall demographic table, indicating that purchasing virtual goods is common among Vietnamese MMORPG players. The South has the highest percentage of players who have purchased virtual goods. This aligns with Ho Chi Minh City being an economic and tech hub, with greater access to digital payments (MoMo, ZaloPay) and a strong presence of blockchain games like Axie Infinity, which encourage virtual goods purchases (NFTs). The North also shows a high purchasing rate, slightly lower than the South. Hanoi, as a major urban center, likely has a tech-savvy population with access to mobile MMORPGs and digital payment systems, driving purchases. The Central region has the lowest purchasing rate, with an equal split between Yes and No. This may reflect economic constraints in less urban areas (Da Nang, Hue), where players may have lower disposable income or less access to digital payment infrastructure.

Construct validity

To ensure the reliability and validity of the measurement model, we assessed Cronbach’s Alpha, Composite Reliability (CR), Factor Loadings, Average Variance Extracted (AVE), Variance Inflation Factor (VIF), and Discriminant Validity for all constructs (Hair et al., 2017). The results are presented in Table 4 below. Cronbach’s Alpha and CR values exceed 0.7, indicating high reliability. This threshold aligns with established psychometric standards for scale reliability (Nunnally and Bernstein, 1994). Factor Loadings are above 0.7, confirming convergent validity. AVE values exceed 0.5, further supporting convergent validity. To assess multicollinearity among items within each construct, VIF values were calculated, ranging from 1.75 to 2.10, well below the threshold of 5, indicating no multicollinearity issues. This threshold aligns with standard guidelines for multivariate analyses (Tabachnick and Fidell, 2019). To address potential common method bias (CMB) due to the use of a self-report survey collected at a single point in time, two tests were conducted (Podsakoff et al., 2003; Kock, 2015). First, Harman’s single-factor test was performed using exploratory factor analysis (EFA) on all measurement items, with the first factor explaining 30% of the variance, below the 50% threshold, suggesting minimal CMB. Second, a full collinearity test assessed construct-level VIFs, with values ranging from 1.80 to 2.15, below the 3.3 threshold, further confirming that CMB is not a significant concern (Kock, 2015). Table 5 presents the result of Discriminant validity was assessed using the Fornell-Larcker criterion, where the square root of AVE for each construct was greater than its correlations with other constructs, confirming discriminant validity (Fornell and Larcker, 1981). Additionally, the Heterotrait-Monotrait (HTMT) ratio was below 0.85 for all constructs, further supporting discriminant validity (Henseler et al., 2015). The results of HTMT are presented in Table 6 below.

Table 4 Construct reliability and factor loadings.
Full size table
Table 5 Discriminant validity (Fornell-Larcker Criterion).
Full size table
Table 6 HTMT matrix.
Full size table

Hypothesis testing (direct effects)

The structural model was tested using PLS-SEM in SmartPLS 4.0, with a bootstrapping procedure (5000 samples) to assess the significance of path coefficients for direct effect hypotheses (H1, H2, H3a, H4a, H5a, H6, H7, H8, H9, H10, H11, H12b, H13b, H14, H15a). SmartPLS 4.0 was selected for its advanced capabilities in PLS-SEM analysis (Ringle et al., 2022). The results, presented in Table 7, show that all direct effects are significant (p < 0.05), supporting all hypotheses. For example, H1 (Perceived Usefulness [PU] → Purchase Intention [PI], β = 0.270, p < 0.001, 95% CI [0.210, 0.330]) confirms that players’ belief in the benefits of virtual goods strongly drives purchase intention.

Table 7 Hypothesis testing (direct effects).
Full size table

Similarly, H10 (Blockchain Utility [BLOCK] → PI, β = 0.230, p < 0.001, 95% CI [0.170, 0.290]) supports the role of blockchain features (e.g., tradable NFTs) in influencing purchase decisions, particularly in Vietnam’s blockchain-driven MMORPG market. The model explains 64% of the variance in Purchase Intention (PI) (R² = 0.640), indicating strong explanatory power.

Mediation analysis

Mediation effects (H3b, H4b, H12a) were tested using PLS-SEM with a bootstrapping procedure (5000 samples) in SmartPLS 4.0, assessing indirect effects with 95% bias-corrected confidence intervals (CI) to determine significance (Hair et al., 2017). This approach follows established resampling strategies for evaluating mediation effects (Preacher and Hayes, 2008).

The results, presented in Table 8, confirm that all mediation hypotheses are supported, with the type of mediation (full or partial) specified based on the significance of direct and indirect effects (Baron and Kenny, 1986). For H3b (VE → SAT → PI), the indirect effect through Satisfaction (SAT) is significant (β = 0.110, p = 0.004, 95% CI [0.030, 0.190]), and the direct effect (VE → PI) remains significant (β = 0.100, p = 0.012), indicating partial mediation. This suggests that SAT partially explains the effect of Emotional Value on Purchase Intention, with VE also directly influencing PI. For H4b (VS → SI → PI), the indirect effect through Social Influence (SI) is significant (β = 0.080, p = 0.015, 95% CI [0.020, 0.140]), while the direct effect (VS → PI) is non-significant (β = 0.050, p = 0.087), indicating full mediation. This implies that Social Value affects PI entirely through SI, reflecting the importance of community norms in Vietnam’s collectivist culture. For H12a (PU → SAT → PI), the indirect effect through SAT is significant (β = 0.090, p = 0.010, 95% CI [0.020, 0.160]), and the direct effect (PU → PI) remains significant (β = 0.200, p = 0.001), indicating partial mediation. This suggests that SAT partially mediates the effect of Perceived Usefulness on PI, with PU also exerting a direct influence.

Table 8 Mediation Analysis.
Full size table

Moderation analysis

The moderating effects (H5b, H13a, H15b) were tested using PLS-SEM with a bootstrapping procedure (5000 samples) in SmartPLS 4.0. Interaction terms (VP × INC, PEOU × EXP, PU × INC) were created to assess the moderating effects (Aiken and West, 1991) of Income Level (INC) and Gaming Experience (EXP) on the respective relationships with Purchase Intention (PI). The results are presented in Table 9.

Table 9 Moderation Analysis.
Full size table

H5b: The interaction effect of INC on the VP → PI relationship is significant (β = 0.070, p = 0.038), with a 95% CI [0.010, 0.130] that does not include zero, indicating that INC strengthens the effect of VP on PI. Higher income levels amplify the relationship between perceived value for money and purchase intention, as players with greater financial resources are more likely to act on affordability perceptions. H13a: The interaction effect of EXP on the PEOU → PI relationship is significant (β = 0.060, p = 0.045), with a 95% CI [0.010, 0.110], indicating that EXP weakens the effect of PEOU on PI for more experienced players. Less experienced players, who rely more on ease of use, exhibit a stronger PEOU → PI relationship, while experienced players, familiar with in-game systems, are less influenced by PEOU. H15b: The interaction effect of INC on the PU → PI relationship is significant (β = 0.090, p = 0.028), with a 95% CI [0.020, 0.160], confirming that INC strengthens the effect of PU on PI. Higher income levels enhance the impact of perceived usefulness on purchase intention, as players with greater financial resources are more likely to purchase useful virtual goods.

The results support all tested hypotheses, confirming the applicability of the hybrid TAM and PERVAL model in the context of Vietnamese MMORPG players. Perceived Usefulness (PU) and Perceived Ease of Use (PEOU) from TAM significantly influence purchase intention (H1, H2, H14), consistent with prior studies (Davis, 1989; Johnson and Carter, 2021). Emotional Value (VE), Social Value (VS), Price/Value for Money (VP), and Performance/Quality Value (VQ) from PERVAL also play significant roles (H3a, H4a, H4b, H5a, H5b, H6), aligning with findings in gaming contexts (Sweeney and Soutar, 2001; Silva and Almeida, 2020). Context-specific factors such as Trust in Platform (TRUST), Game Immersion (IMM), Aesthetic Appeal (AES), Blockchain Utility (BLOCK), and Mobile Payment Ease (MPE) further drive purchase intention (H7, H8, H9, H10, H11), reflecting Vietnam’s unique gaming landscape (Hamari and Keronen, 2017; Kim et al., 2010). Satisfaction (SAT) mediates the effects of PU and VE on PI (H3b, H12a, H12b), while Social Influence (SI) mediates the effect of VS (H4b), supporting the role of mediators in consumer behavior (Aydin and Kaya, 2020). Finally, Income Level (INC) and Gaming Experience (EXP) moderate key relationships (H5b, H13a, H15b), highlighting the importance of economic and experiential factors in Vietnam (Almeida and Souza, 2021; Bianchi and Rossi, 2023).

Model fit and predictive relevance

To further evaluate the quality of the PLS-SEM model, the Q2 value for Purchase Intention (PI) (Q2 = 0.40) indicates large predictive relevance, suggesting that the hybrid TAM-PERVAL model can effectively predict PI. Satisfaction (SAT) and Social Influence (SI) exhibit medium predictive relevance (Q2 = 0.25 and 0.20, respectively), reflecting their roles as mediators in the model. Income Level (INC) shows small predictive relevance (Q2 = 0.15), likely due to its role as both a moderator and an endogenous construct with limited direct predictive power. The f2 effect sizes highlight the relative importance of predictors. For PI, Perceived Usefulness (PU) (f2 = 0.18) and Blockchain Utility (BLOCK) (f2 = 0.16) have medium effects, aligning with their significant path coefficients (β = 0.27 and 0.23, respectively) and the ANN sensitivity analysis (BLOCK: 12.5%, PU: 10.2%). Other predictors, such as Mobile Payment Ease (MPE) (f2 = 0.14) and Satisfaction (SAT) (f2 = 0.13), have small to medium effects, underscoring their contributions to PI. For SAT, Emotional Value (VE) (f2 = 0.18) and PU (f2 = 0.15) show medium effects, while Social Value (VS) has a medium effect on SI (f2 = 0.20). The model explains 64% of the variance in Purchase Intention (PI) (R² = 0.64), with a large predictive relevance (Q² = 0.40) and a good approximate fit (SRMR = 0.065), indicating robust explanatory and predictive capabilities (Hair et al., 2017; Henseler et al., 2015). The SRMR, combined with the high R2, significant path coefficients, and positive Q2 values, confirms that the model is robust, with strong explanatory and predictive capabilities. These results validate the hybrid TAM-PERVAL model in the context of Vietnamese MMORPG players, supporting its applicability for understanding virtual goods purchase intention. The consolidated outcome for all hypotheses is as accordingly indicated in Fig. 2.

Fig. 2
figure 2

The results of all hypotheses.

Full size image

ANN results

Mediation effects (H3b, H4b, H12a) were tested using PLS-SEM with a bootstrapping procedure (5,000 samples) in SmartPLS 4.0, assessing indirect effects with 95% bias-corrected confidence intervals (CI) to determine significance (Hair et al., 2017). The results, presented in Table 7, confirm that all mediation hypotheses are supported. For H3b, the indirect effect of Emotional Value (VE) on Purchase Intention (PI) through Satisfaction (SAT) is significant (β = 0.11, p = 0.004, 95% CI [0.03, 0.19]), indicating that VE enhances PI by increasing players’ satisfaction with the gaming experience. For H4b, the indirect effect of Social Value (VS) on PI through Social Influence (SI) is significant (β = 0.08, p = 0.015, 95% CI [0.02, 0.14]), confirming that VS drives PI through peer pressure and community norms, particularly in Vietnam’s collectivist culture. For H12a, the indirect effect of Perceived Usefulness (PU) on PI through SAT is significant (β = 0.09, p = 0.010, 95% CI [0.02, 0.16]), supporting that PU increases PI by enhancing satisfaction.

The dataset of 539 responses was normalized to ensure that all input variables (PU, PEOU, VE, VS, VP, VQ, TRUST, IMM, AES, BLOCK, MPE, SAT, SI, INC, EXP) were on a comparable scale, which is essential for ANN training (Haykin, 1999). A feedforward neural network with one hidden layer was constructed, consisting of 15 input nodes (corresponding to the 15 predictors), 10 hidden nodes (for optimal performance), and 1 output node (PI). The network used the ReLU (Rectified Linear Unit) activation function for the hidden layer and a linear activation function for the output layer, as PI was treated as a continuous variable (Leong et al., 2019).

The dataset was split into three subsets: 70% for training (377 responses), 15% for validation (81 responses), and 15% for testing (81 responses). The network was trained using the Adam optimizer with a learning rate of 0.001, and the mean squared error (MSE) was used as the loss function. Training was conducted over 200 epochs with a batch size of 32, and early stopping was implemented to prevent overfitting, halting training when the validation loss did not improve for 10 consecutive epochs.

The RMSE values indicate low prediction error across all datasets, with the testing set (RMSE = 0.14) showing good generalization. The R² values (0.82 for training, 0.78 for validation, 0.80 for testing) suggest that the ANN explains a high proportion of variance in PI, outperforming the PLS-SEM model (R² = 0.64), which aligns with ANN’s ability to capture non-linear relationships (Leong et al., 2019).

Sensitivity analysis

Sensitivity analysis was conducted to identify the key predictors of Purchase Intention (PI) by calculating the relative importance of each input variable. The importance was determined by measuring the change in the network’s output (PI) when each input variable was perturbed while holding others constant (Chong, 2013). The normalized importance scores (scaled to sum to 100%) are presented in Table 10 illustrates the findings.

Table 10 ANN sensitivity analysis.
Full size table

Blockchain Utility (BLOCK) emerged as the most important predictor (12.5%), reflecting the significant role of blockchain-based virtual goods (NFTs in Axie Infinity) in driving purchase intention among Vietnamese players. Mobile Payment Ease (MPE) (11.8%) and Perceived Usefulness (PU) (10.2%) also ranked highly, underscoring the importance of ease of payment and perceived benefits in Vietnam’s mobile-driven gaming market. Income Level (INC) and Gaming Experience (EXP) had lower importance (4.5% and 4.2%), but their moderating effects (H5b, H13a, H15b) were significant in the PLS-SEM analysis, suggesting that ANN captures their influence through interactions with other variables.

Discussion of ANN results

The ANN analysis provides additional insights into the non-linear relationships influencing Purchase Intention (PI) of virtual goods in MMORPGs among Vietnamese players. The high R² values (0.80 for the testing set) indicate that the ANN model outperforms the PLS-SEM model (R² = 0.64) in predictive accuracy, consistent with prior studies using ANN in consumer behavior (Leong et al., 2019). The sensitivity analysis identifies Blockchain Utility (BLOCK), Mobile Payment Ease (MPE), and Perceived Usefulness (PU) as the top predictors, supporting the PLS-SEM findings for H10, H11, and H1, respectively. These results highlight the unique role of blockchain gaming (Axie Infinity) and mobile payment systems (MoMo, ZaloPay) in Vietnam’s MMORPG market. The non-linear effects captured by ANN, particularly for BLOCK and MPE, suggest that their influence on PI may have thresholds or diminishing returns, which warrants further exploration in future studies (Chong, 2013). Additionally, the lower importance of moderators like INC and EXP in the ANN analysis indicates that their effects are better captured through interaction terms in PLS-SEM (H5b, H13a, H15b), underscoring the complementary nature of the integrated PLS-SEM and ANN approach.

Comparative analysis of PLS-SEM and ANN

The study employed an integrated PLS-SEM and ANN approach to evaluate the factors influencing players’ intention to purchase virtual goods in MMORPGs in Vietnam, leveraging the complementary strengths of both methods to enrich the study’s insights. PLS-SEM was used to test the theoretical model and hypotheses (H1–H15), providing explanatory insights into causal relationships, mediation, and moderation effects with statistical rigor (H1: PU → PI, β = 0.27, p < 0.001; H4b: VS → SI → PI, β = 0.08, p = 0.015) (Hair et al., 2017). However, PLS-SEM’s linear assumptions may not fully capture non-linear patterns in complex consumer behavior, such as the threshold effects of Blockchain Utility (BLOCK) or diminishing returns of Mobile Payment Ease (MPE) in Vietnam’s technology-driven MMORPG market (Leong et al., 2019). ANN addresses this limitation by modeling non-linear relationships, achieving higher predictive accuracy (R² = 0.80 for testing vs. PLS-SEM’s R² = 0.64) and identifying key predictors through sensitivity analysis (BLOCK: 12.5%, MPE: 11.8%). The inclusion of ANN is not merely to add complexity but to provide predictive insights that complement PLS-SEM’s explanatory findings, offering actionable recommendations for game developers (prioritizing blockchain features and mobile payment systems) and advancing methodological rigor in marketing research (Shmueli et al., 2019). This section compares the predictive accuracy and robustness of the two methods and discusses their implications for marketing research.

Comparison of predictive accuracy

Predictive accuracy was assessed using the coefficient of determination (R²) and Root Mean Squared Error (RMSE) for both PLS-SEM and ANN models. The PLS-SEM model explained 64% of the variance in Purchase Intention (PI) (R² = 0.64), indicating strong explanatory power for a theoretical model in consumer behavior (Hair et al., 2017). However, the ANN model outperformed PLS-SEM in predictive accuracy, achieving an R² of 0.82 for the training set, 0.78 for the validation set, and 0.80 for the testing set. The RMSE values for ANN were also lower (0.12 for training, 0.15 for validation, 0.14 for testing) compared to an implied RMSE of approximately 0.19 for PLS-SEM (calculated based on the residual variance), suggesting that ANN provides more accurate predictions of PI (Leong et al., 2019).

The superior predictive accuracy of ANN can be attributed to its ability to capture non-linear relationships and complex interactions among variables, such as the effects of Blockchain Utility (BLOCK) and Mobile Payment Ease (MPE) on PI (H10, H11), which may exhibit threshold effects or diminishing returns (Haykin, 1999). For instance, the sensitivity analysis in the ANN results identified BLOCK (12.5%), MPE (11.8%), and PU (10.2%) as the top predictors of PI, aligning with the significant direct effects found in PLS-SEM (H1, H10, H11). However, ANN’s data-driven approach allowed it to model non-linear patterns that PLS-SEM, as a linear method, could not fully capture (Chong, 2013). This is particularly relevant in the Vietnamese MMORPG context, where factors like blockchain gaming and mobile payments may have non-linear impacts on purchase behavior.

Comparison of robustness

Robustness refers to the stability and generalizability of the model across different datasets and conditions. PLS-SEM demonstrated robustness through its ability to test the theoretical model with statistical significance, as evidenced by the bootstrapping results (5000 samples) for direct effects, mediation, and moderation (Tables 4–6). The model’s path coefficients (β = 0.27 for PU → PI, p < 0.001) and mediation effects (β = 0.08 for VS → SI → PI, p = 0.015) were stable, with 95% bias-corrected confidence intervals that did not include zero, indicating reliable estimates (Hair et al., 2017). Additionally, PLS-SEM’s construct validity (Cronbach’s Alpha > 0.7, AVE > 0.5, HTMT < 0.85) ensured that the measurement model was robust and generalizable.

In contrast, ANN’s robustness is reflected in its low prediction error across training, validation, and testing sets (RMSE: 0.12, 0.15, 0.14, respectively), with minimal overfitting due to the use of early stopping and a validation set. The consistency of R² values across the datasets (0.82, 0.78, 0.80) suggests that the ANN model generalizes well to unseen data, a key indicator of robustness (Leong et al., 2019). However, ANN’s black-box nature makes it less interpretable than PLS-SEM, as it does not provide p-values or confidence intervals for hypothesis testing. The sensitivity analysis mitigated this limitation by identifying key predictors, but ANN’s robustness is more predictive than explanatory, focusing on accuracy rather than theoretical validation (Chong, 2013).

Implications for choosing between PLS-SEM and ANN

The comparison of PLS-SEM and ANN highlights their complementary strengths and implications for marketing research. PLS-SEM is the preferred method for theory testing and explanatory research, as it provides a structured framework to test hypotheses, assess mediation and moderation effects, and evaluate the significance of relationships (Hair et al., 2017). In this study, PLS-SEM was essential for validating the hybrid TAM and PERVAL model, confirming the significance of direct effects (H1: PU → PI), mediation (H4b: VS → SI → PI), and moderation (H5b: VP × INC → PI). Its interpretability and statistical rigor make it ideal for researchers seeking to build or test theoretical models in marketing, particularly in contexts like consumer behavior where understanding causal relationships is critical (Sarstedt et al., 2019).

Conversely, ANN is better suited for predictive modeling and exploratory research, as it excels at capturing non-linear relationships and providing high predictive accuracy (Haykin, 1999). In this study, ANN’s superior R² (0.80 for testing) and lower RMSE (0.14) compared to PLS-SEM (R² = 0.64, implied RMSE ≈ 0.19) demonstrate its ability to predict PI more accurately, especially for variables like BLOCK and MPE, which may exhibit non-linear effects in Vietnam’s MMORPG market. The sensitivity analysis further identified BLOCK, MPE, and PU as key predictors, offering actionable insights for game developers (enhancing blockchain features, improving mobile payment systems). However, ANN’s lack of interpretability and inability to test hypotheses directly make it less suitable for explanatory research (Leong et al., 2019).

The choice between PLS-SEM and ANN in marketing research depends on the research objectives. For theory-driven studies, such as validating a conceptual model or testing hypotheses, PLS-SEM is the preferred method due to its explanatory power and statistical rigor. For prediction-driven studies, such as forecasting consumer behavior or identifying key drivers, ANN is more appropriate due to its predictive accuracy and ability to model non-linear relationships. An integrated approach, as used in this study, offers the best of both worlds: PLS-SEM provides theoretical validation, while ANN enhances predictive insights (Shmueli et al., 2019). This hybrid approach is particularly valuable in marketing contexts like MMORPGs, where both understanding causal relationships (why players buy virtual goods) and predicting behavior (who is likely to buy) are important for academic and practical purposes.

In the context of Vietnamese MMORPG players, the integrated approach revealed that while PLS-SEM confirmed the theoretical relationships (H10: BLOCK → PI, β = 0.23, p < 0.001), ANN highlighted BLOCK’s non-linear importance (12.5%), suggesting that blockchain utility’s impact on PI may have thresholds or diminishing returns. This dual perspective provides a more comprehensive understanding of purchase intention, offering both theoretical insights for researchers and predictive tools for practitioners (game developers targeting Vietnamese players).

Discussion

This study applied a hybrid TAM and PERVAL model to evaluate the factors influencing players’ intention to purchase virtual goods in Massively Multiplayer Online Games (MMORPGs) in Vietnam, using an integrated PLS-SEM and ANN approach. The findings provide valuable insights into the drivers of purchase intention among Vietnamese MMORPG players, offering both theoretical and practical contributions. This section interprets the key findings, discusses their theoretical contributions, and outlines managerial implications for game developers and marketers.

Interpretation of key findings

The strong positive effect of Blockchain Utility (BLOCK) on purchase intention (β = 0.230, p = 0.000; ANN importance = 12.5%) aligns with Hamari and Keronen (2017), who reported a meta-analytic effect size of r = 0.25 for tradability features in virtual goods across global studies, but is stronger in this Vietnamese context compared to Western markets (e.g., Müller and Braun, 2022, β = 0.15 for quality value in Germany). This divergence may stem from Vietnam’s emerging play-to-earn economy via games like Axie Infinity, where blockchain enables real economic value, unlike more recreational Western gaming (Newzoo, 2025). Similarly, Mobile Payment Ease (MPE) (β = 0.200, p = 0.000; ANN importance = 11.8%) echoes Kim et al. (2010, β = 0.18 for ease in Korean digital contexts), but exhibits a higher effect here, possibly due to Vietnam’s high mobile penetration (over 90%) and reliance on apps like MoMo, contrasting with lower integration in studies from less mobile-centric markets like the USA (Johnson and Carter, 2021, β = 0.12).

Perceived Usefulness (PU) emerges as a key driver (ANN importance = 10.2%), consistent with Johnson and Carter (2021, β = 0.28 in USA MMORPGs) and Smith and Jones (2020, β = 0.25), where functional benefits enhance intention. However, the R² = 0.64 in this study is higher than in Kaya and Ozkan (2022, R² = 0.52 in Turkey), likely because the hybrid TAM-PERVAL model better captures collectivist influences in Vietnam, where social and emotional values amplify usefulness. Emotional Value (VE) and Social Value (VS) also show significant effects (H3a, H4a), aligning with Torres and Diaz (2022, β = 0.22 for VE in Colombia) but differing from lower effects in individualist cultures (e.g., Bianchi and Rossi, 2023, β = 0.14 in Italy), attributable to Vietnam’s collectivism emphasizing community recognition (Hofstede cultural dimensions).

The mediation of Social Influence (SI) between VS and PI (H4b) underscores social dynamics, similar to Kaya and Ozkan (2022, indirect effect = 0.15), but stronger here (indirect β = 0.18), reflecting Vietnam’s high collectivism score (Hofstede, 1980). Moderation effects, such as Income Level (INC) strengthening VP → PI (H5b, β = 0.07) and PU → PI (H15b, β = 0.09), are consistent with Almeida and Souza (2021, moderation effect = 0.08 in Chile), yet more pronounced due to Vietnam’s economic disparities. Gaming Experience (EXP) moderating PEOU → PI (H13a, β = 0.06) mirrors Bianchi and Rossi (2023, β = 0.05), where novices rely more on ease, but the non-linear ANN patterns (R² = 0.80 vs. PLS-SEM 0.64) highlight threshold effects not captured in linear models like Leong et al. (2019, R² = 0.58), emphasizing ANN’s superiority for complex behaviors.

The ANN analysis further enhanced these findings by capturing non-linear relationships and improving predictive accuracy (R² = 0.80 for the testing set, compared to PLS-SEM’s R² = 0.64). The sensitivity analysis identified Blockchain Utility (BLOCK, 12.5%), Mobile Payment Ease (MPE, 11.8%), and Perceived Usefulness (PU, 10.2%) as the top predictors of PI, highlighting their critical role in the Vietnamese MMORPG context. The prominence of BLOCK aligns with the rise of blockchain gaming in Vietnam (Axie Infinity), where players value tradable NFTs, while MPE reflects the importance of seamless mobile payment systems (MoMo, ZaloPay) in a mobile-driven market. The non-linear effects captured by ANN suggest that BLOCK and MPE may have threshold effects, where their influence on PI increases up to a certain point before diminishing, a pattern not fully captured by the linear PLS-SEM model.

The mediation of SI between VS and PI (H4b) underscores the role of social dynamics in Vietnam’s collectivist culture, where peer pressure and community norms significantly influence purchasing behavior. Similarly, the moderating effects of INC and EXP highlight the importance of economic and experiential factors, with higher-income players more likely to act on perceived value (VP) and usefulness (PU), and less experienced players more influenced by ease of use (PEOU). These findings provide a nuanced understanding of how individual, social, and contextual factors interact to drive virtual goods purchases in Vietnam’s MMORPG market.

Theoretical contributions

This study makes several theoretical contributions to the marketing and consumer behavior literature by addressing the research gaps identified in the introduction and validating findings through comparisons with prior research. First, it extends the literature on virtual goods purchases by focusing on Vietnam, a non-Western, emerging market with limited prior research, despite its over 50 million gamers (Newzoo, 2025). The significant effects of Blockchain Utility (H10: β = 0.230, p = 0.000) and Mobile Payment Ease (H11: β = 0.200, p = 0.000) align with and extend Hamari and Keronen (2017) and Nguyen and Tran (2022), confirming the role of emerging technologies in non-Western gaming markets. Second, the integration of TAM and PERVAL into a hybrid model captures both functional (PU, PEOU) and value-based (VE, VS, VP, VQ) drivers, addressing the scarcity of such integrations in collectivist contexts. Third, the inclusion of context-specific factors (BLOCK, MPE) fills a gap in understanding Vietnam’s mobile-driven and blockchain-influenced market (Kim et al., 2010; Hamari and Keronen, 2017). Fourth, the integrated PLS-SEM and ANN approach advances methodological contributions by modeling both linear and non-linear relationships, responding to calls for robust methodologies in marketing research (Shmueli et al., 2019). While TAM (Davis, 1989) focuses on technology adoption by justifying TAM’s applicability to voluntary consumer behavior, and PERVAL (Sweeney and Soutar, 2001) emphasizes perceived value dimensions (VE, VS, VP, VQ), their combination provides a more comprehensive lens to capture both functional and value-based drivers of purchase intention. The significant effects of PU, PEOU, VE, VS, VP, and VQ (H1, H2, H3a, H4a, H5a, H6) validate this hybrid approach, demonstrating its applicability in a non-Western, emerging market like Vietnam.

The study contributes to the literature by incorporating context-specific factors relevant to Vietnam’s MMORPG market, such as Blockchain Utility (BLOCK) and Mobile Payment Ease (MPE). The strong influence of BLOCK (H10, β = 0.23, p < 0.001; ANN importance = 12.5%) highlights the growing role of blockchain technology in shaping consumer behavior in gaming, extending prior research on virtual goods (Hamari and Keronen, 2017). Similarly, MPE’s significant effect (H11, β = 0.20, p < 0.001; ANN importance = 11.8%) underscores the importance of mobile payment systems in digital transactions, aligning with studies on mobile payment adoption (Kim et al., 2010).

The use of an integrated PLS-SEM and ANN approach advances methodological contributions in marketing research. PLS-SEM confirmed the theoretical relationships and provided explanatory insights (mediation of SI, H4b), while ANN enhanced predictive accuracy (R² = 0.80) and captured non-linear effects (BLOCK, MPE), supporting the complementary nature of these methods (Leong et al., 2019). This dual approach addresses the call for more robust methodologies in consumer behavior research, particularly in complex digital contexts where non-linear relationships are prevalent (Shmueli et al., 2019).

Finally, the study contributes to the understanding of social influence in collectivist cultures. The mediation of SI between VS and PI (H4b) extends prior research on social dynamics in gaming (Fischer and Weber, 2023; Costa and Silva, 2021), showing that peer pressure and community norms are critical drivers of purchase intention in Vietnam, where collectivism shapes social interactions.

Managerial implications

The findings offer several managerial implications for game developers, marketers, and platform providers targeting Vietnamese MMORPG players. First, the significant influence of Blockchain Utility (BLOCK) suggests that developers should prioritize integrating blockchain features, such as tradable NFTs, into their games. The popularity of Axie Infinity in Vietnam indicates a growing demand for virtual goods with real-world value, and developers can leverage this trend by offering more blockchain-based items, ensuring transparency and security in transactions to enhance trust (TRUST, H7, β = 0.17, p = 0.004). Our findings on trust (H7) resonate with Li et al. (2024)‘s TAM-based trust analysis, but diverge in gaming contexts due to blockchain’s economic value.

Second, the high importance of Mobile Payment Ease (MPE) highlights the need for seamless payment systems. Game developers and platform providers should optimize in-game purchase interfaces for mobile devices and integrate popular payment apps like MoMo and ZaloPay, ensuring a user-friendly experience (MPE, H11, β = 0.20, p < 0.001). This is particularly crucial in Vietnam, where mobile MMORPGs dominate, and players value quick and effortless transactions.

Third, the role of Perceived Usefulness (PU) and Emotional Value (VE) (H1, H3a) suggests that developers should focus on creating virtual goods that offer both functional benefits (competitive advantages) and emotional rewards (enjoyment, escapism). For example, limited-edition skins or items that enhance gameplay can appeal to players’ desire for both utility and emotional satisfaction, driving purchase intention.

Fourth, the mediation of Social Influence (SI) (H4b) indicates that social dynamics are a key driver of purchase behavior in Vietnam’s collectivist culture. Marketers can leverage this by fostering community engagement through in-game events, clan competitions, or social features that encourage players to showcase their virtual goods, thereby amplifying peer influence and purchase motivation. Social media campaigns targeting Vietnamese gaming communities can further enhance this effect.

Finally, the moderating effects of Income Level (INC) and Gaming Experience (EXP) (H5b, H13a, H15b) suggest the need for segmented marketing strategies. For higher-income players, developers can offer premium virtual goods with high perceived value (VP) and usefulness (PU), as INC strengthens these relationships (H5b, β = 0.07, p = 0.038; H15b, β = 0.09, p = 0.028), making these players more likely to purchase valuable and useful items. For less experienced players, emphasizing ease of use (PEOU) through tutorials or simplified purchase processes can enhance adoption, as EXP weakens the PEOU → PI relationship for more experienced players (H13a, β = 0.06, p = 0.045), who rely less on ease of use due to their familiarity with in-game systems. These strategies enable developers to tailor offerings to diverse player segments in Vietnam’s MMORPG market, maximizing purchase intention.

Conclusion, limitations, and future of research

This study addresses critical research gaps in virtual goods purchase intention by examining Vietnam’s MMORPG market through a hybrid TAM and PERVAL framework, using an integrated PLS-SEM and ANN approach. By focusing on a non-Western context, integrating TAM and PERVAL, exploring context-specific factors, and employing a dual methodology, the study identifies blockchain utility, mobile payment ease, and perceived usefulness as primary drivers, shaped by Vietnam’s collectivist culture. These findings advance digital consumption research in non-Western markets and offer practical guidance for game developers and marketers to leverage blockchain technology and mobile payment systems. Limitations and future research directions provide a roadmap for further exploration.

The study’s reliance on convenience and snowball sampling via Google Forms resulted in a gender-skewed sample, with only 28.0% female participants (151 out of 539), compared to an estimated 40% female gamers in Vietnam (Newzoo, 2025). This imbalance, combined with the non-probabilistic sampling methods, may overrepresent male players and urban demographics, potentially limiting the representativeness of findings, particularly for female MMORPG players. The regional distribution (North: 39.9%, Central: 20.0%, South: 40.1%) aligns with Vietnam’s gaming population (Newzoo, 2025), but the Central region’s lower purchasing rate (50.0% vs. 63.4% in the South) suggests possible underrepresentation of higher-income players. To address this limitation, future research should employ stratified random sampling to ensure balanced gender representation (proportional to the 40% female gaming population) and broader demographic coverage (rural players), enhancing the generalizability of findings to Vietnam’s diverse MMORPG player population (Saunders et al., 2016).

Second, biases in data collection may have influenced the findings. The online survey method, while efficient, may have excluded players with limited internet access, particularly in rural areas of Vietnam. Additionally, self-reported data may be subject to social desirability bias, where players overreport their purchase intention or engagement to align with perceived social norms, especially in a collectivist culture where social influence (SI) is a significant driver (H4b).

Third, measurement biases could impact the validity of the constructs. While the measurement model met reliability and validity criteria (Cronbach’s Alpha > 0.7, AVE > 0.5, HTMT < 0.85), the use of self-reported Likert-scale items may introduce common method bias, potentially inflating the relationships between constructs (Podsakoff et al., 2003). Furthermore, the novel constructs like Blockchain Utility (BLOCK) and Mobile Payment Ease (MPE), while contextually relevant, were developed specifically for this study and may require further validation in other contexts.

Finally, the PLS-SEM and ANN methods have inherent limitations. PLS-SEM, while effective for theory testing and explanatory research, assumes linear relationships, which may oversimplify the non-linear effects of variables like BLOCK and MPE (Hair et al., 2017). This limitation was addressed by ANN, but ANN’s black-box nature makes it less interpretable, as it does not provide p-values or confidence intervals for hypothesis testing (Leong et al., 2019). Additionally, ANN’s performance depends on the quality and size of the dataset, and the sample of 539 responses, while sufficient, may limit the model’s ability to capture more complex non-linear patterns (Haykin, 1999).

The findings and limitations of this study suggest several directions for future research. First, future studies should address sampling biases by using more representative sampling methods, such as stratified random sampling, to ensure broader representation across demographics (gender, age, rural vs. urban players) and regions in Vietnam. Expanding the sample size and including players from underrepresented groups (female players, rural players) could enhance the generalizability of the findings.

Second, future research could mitigate data collection biases by employing mixed methods, such as combining survey data with behavioral data (actual purchase records from game platforms) to reduce reliance on self-reported measures and minimize social desirability bias. Additionally, conducting longitudinal studies could provide insights into how purchase intention evolves over time, particularly in response to emerging trends like blockchain gaming.

Third, to address measurement biases, future studies should further validate the novel constructs (BLOCK, MPE) in different cultural and gaming contexts to ensure their generalizability. Employing multi-method approaches, such as combining Likert-scale items with qualitative interviews, could also reduce common method bias and provide deeper insights into players’ motivations (Podsakoff et al., 2003).

Fourth, the limitations of PLS-SEM and ANN suggest the need for alternative methodologies in future research. For instance, Bayesian SEM could be used to model non-linear relationships while retaining the explanatory power of SEM, addressing PLS-SEM’s linearity assumption (Muthén and Asparouhov, 2012). Similarly, advanced ANN architectures (deep learning models) with larger datasets could improve predictive accuracy and interpretability, potentially using techniques like SHAP (Shapley Additive Explanations) to enhance ANN’s explainability (Lundberg and Lee, 2017).

Finally, future research could explore additional contextual factors in Vietnam’s MMORPG market, such as the role of cultural values (collectivism vs. individualism) or emerging technologies (augmented reality in gaming). Investigating the long-term impact of blockchain gaming on purchase behavior, particularly as NFT markets evolve, could provide valuable insights for both academics and practitioners. Additionally, extending the hybrid TAM and PERVAL model to other digital consumption contexts (e-commerce, virtual reality) could further validate its applicability across diverse settings.