537 American Journal of Economic and Management Business e-ISSN: 2835-5199 Vol. 4 No. 5 May 2025 Analysis of the Influence of Social Media Marketing Features on Consumer Purchase Decisions with Brand Trust as a Mediating Variable in the Electric Vehicle Industry Mochamad Amirudin1, Maria Apsari Sugiat2, Fera Shinta Maria Silalahi3 Telkom University, Indonesia1,2,3 Email: mochamadamir@student.telkomuniversity.ac.id, mariasugiat@telkomuniversity.ac.id, ferasilalahi@365.telkomuniversity.ac.id Abstract The growth of the electric vehicle (EV) industry in Indonesia presents opportunities and challenges, especially for new market entrants like PT Neta Auto Indonesia. One of the primary obstacles is low consumer trust, which impedes purchase decisions despite favorable government policies. This study investigates the influence of social media marketing featuresβ€”interactivity, informativeness, entertainment, and perceived relevanceβ€”on consumer purchasing decisions, mediated by brand trust. Employing a quantitative approach, data were collected from 389 verified Neta electric vehicle consumers using an online questionnaire. Structural Equation Modeling (SEM) with AMOS 22 was utilized to analyze direct and indirect effects among the variables. Findings reveal that interactivity and informativeness significantly and positively affect purchase decisions both directly and through brand trust, while entertainment shows a significant indirect effect via trust. Conversely, perceived relevance does not significantly influence trust or decisions. These results emphasize the critical role of trust-building content in social media strategies, underscoring that high-quality, interactive, and informative digital content strengthens brand trust and facilitates consumer decision-making. The study provides actionable insights for automotive marketers in emerging markets and offers a foundation for future research to examine longitudinal and cross-cultural dynamics or integrate technological innovations such as AI-driven personalization and AR content. This research highlights the strategic importance of optimizing social media marketing features to support sustainable market growth in the EV sector. Keywords: Social Media Marketing Features, Brand Trust, Purchase Decisions, Digital Marketing, Electric Vehicles, Interactivity, Entertainment, Informativeness, Perceived Relevance. INTRODUCTION Hozon New Energy Automobile Co., Ltd. (HOZON) was established in 2014, is an innovative technology company that integrates hardware products and software services based on innovative mailto:mochamadamir@student.telkomuniversity.ac.id mailto:mariasugiat@telkomuniversity.ac.id mailto:ferasilalahi@365.telkomuniversity.ac.id Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 538 technology research and development, intelligent manufacturing, and omni-channel sales services (Druehl et al., 2018; Pistrui, 2024). Neta Auto is a car brand under HOZON that adheres to the development concept of "electrification, intelligence, and networking", making high-quality and affordable intelligent electric vehicles. In April 2017, Neta Auto obtained a vehicle production license approved by NDRC, and in June 2018 obtained a production license approved by MIIT China (Baciarlini, 2021; Paba, 2022). Neta Auto implemented a global restructuring strategy with a total investment of more than RMB 20 billion (Chap & Liu, 2024). At present, the company has more than 9000 employees and more than 3000 technology patents. Since 2021, Neta Auto has proposed a "go global" strategy and has established overseas subsidiaries in more than 20 countries and regions around the world, such as ASEAN, Central and South America, the Middle East, and Africa. On March 10, 2023, construction of Neta's factory in Thailand officially began, marking Neta's first overseas factory (Dal Maso, 2024; Neuman, 2023). At the same time, Neta signed a cooperation agreement with BGAC Company in Thailand (Bath & Nottage, 2020). On April 22, 2023, PT Neta Auto Indonesia officially established itself in Indonesia and declared the brand and launched the model V electric vehicle at GIIAS 2023. PT Neta Auto Indonesia then started local production in the form of CKD in collaboration with PT Handal Indonesia Motor. Figure 1.1 shows the Neta V-II electric vehicle model, a product from PT Neta Auto Indonesia designed with a modern, stylish, and contemporary design. This vehicle has dimensions of 4,070 mm x 1,690 mm x 1,540 mm and can accommodate up to five people. Featured features include advanced technology such as Advanced Driver Assistance Systems (ADAS) with nine functions, as well as a 36.1 kWh Lithium Ferro-Phosphate (LFP) battery that allows up to 401 km of mileage on a single charge. With DC Fast Charging technology, the battery can be charged from 30% to 80% in just 30 minutes, offering users convenience and efficiency (Ghasemi-Marzbali, 2022; Suarez & Martinez, 2019; Tu et al., 2019). On September 27, 2024, at The Grand Launch of Neta X Xperience Tech With EV, which was held at Hutan Kota in Plataran, Central Jakarta, PT Neta Auto Indonesia officially released a medium electric SUV with premium features, Neta X, after previously being introduced at the GIIAS 2024 event. Neta X offers a wide range of the latest features and technologies with the concept of "Xperience Tech With EV" to improve driving comfort and safety and meet the needs of a dynamic modern society. Neta X, a medium SUV electric vehicle launched by PT Neta Auto Indonesia with the concept of "Xperience Tech With EV." This vehicle carries premium features such as cutting-edge technology for driving comfort and safety. As part of the company's strategy, Neta X is produced locally with a Domestic Content Level (TKDN) of 44%, demonstrating the company's commitment to the development of Indonesia's automotive industry. The model also offers technological innovations that appeal to the dynamic modern consumer, making it one of the flagship products in the premium electric vehicle segment. The digital age has brought significant changes in corporate marketing strategies, with social media becoming one of the most effective tools to reach consumers widely (Berthon et al., Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 539 2012). Platforms like Instagram, YouTube, TikTok, and Facebook have revolutionized marketing, creating opportunities for more interactive communication, personal, and Real-time between companies and consumers (Alfajri et al., 2019). Increased internet penetration and payment adoption digital also driving this growth, making social media a key element in the modern marketing ecosystem (Zulfikar & Elsamara, 2024). Social media allows for interactive communication and Real-time, as well as providing flexibility through a variety of content formats, such as images, videos, and live broadcasts. With 78.5% of Indonesia's 221.56 million internet users active on social media (APJII, 2024), this platform is a strategic space to build Brand Trust and strengthen consumer relationships with brands (Alfajri et al., 2019; Rogers, 2014) In the electric vehicle industry, social media plays an important role in introducing new technologies to the public who still need clear and reliable information before making a purchase decision. Through content such as feature reviews, simulations, and testimonials, companies can increase consumer awareness and trust. For Neta Auto Indonesia's electric cars, the effective use of social media can help drive electric vehicle adoption, build brand loyalty, and expand market share amid increasingly fierce competition. The volume of electric car sales globally increased from 2013 to 2023, this electric vehicle (EV) phenomenon has become a hot topic around the world. In 2023, electric vehicle sales will reach more than 9 million units globally. China leads the market with nearly 60% of global EV sales . Europe and the United States are also experiencing rapid growth, with Europe accounting for 25% of global sales and the US about 10%. The phenomenon of electric vehicles is also growing rapidly in Indonesia, in line with the government's commitment to reduce emissions and increase energy security. Based on a press release from the Coordinating Ministry for Economic Affairs on August 29, 2023, the government targets the production of 13 million electric motorcycles and 2.2 million electric cars by 2030, this target is supported by abundant nickel resources. Indonesia has the world's largest nickel reserves, equivalent to 23% of global reserves, and is key in the global battery supply chain. Based on data from the Ministry of Energy and Mineral Resources as of October 18, 2023, Indonesia has a total of 17.68 billion tons of ore and 177.8 million tons of metals, with reserves of 5.2 billion tons of ore and 57 million tons of metals. These nickel reserves are mostly spread across Southeast Sulawesi, Central Sulawesi, and North Maluku (Maulina Ulfa, IDX Channel, 2023). The Coordinating Ministry for Economic Affairs of the Republic of Indonesia, in its press release hm.4.6/278/set.m.ekon.3/08/2024 on August 7, 2024, stated Indonesia's commitment to becoming a leading global producer of lithium-ion battery anodes to support the electric vehicle (EV) industry. The government has implemented strategic policies, including Presidential Regulation No. 79/2023, which offers incentives like import duty exemptions for EVs committed to local production, and Presidential Regulation No. 55/2019, which serves as the legal foundation for EV development, covering production, tax incentives, and charging infrastructure. Additional measures include Tax Incentive Regulation (BKPM No. 6/2023), providing tax holidays and allowances, and the Ministry of Energy and Mineral Resources Regulation No. 13/2020, which Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 540 supports charging infrastructure development, with PT PLN operating 624 public charging points by 2023 (Coordinating Ministry for Economic Affairs, 2024). Despite these efforts, PT Neta Auto Indonesia faces significant challenges in meeting its sales targets, as reported by cnnindonesia.com (2024), with sales trends showing instability from 2023 to 2024. PT Neta Auto Indonesia initially sold 100 units of Neta V EVs in November 2023, followed by 66 units in December, totaling 166 units for the year, meeting the 250-unit CBU import quota. However, in 2024, sales fell sharply, achieving only 26% of the Whole Sales target (615 of 2,408 units) and 28% of the Retail Sales target (612 of 2,153 units), with monthly ratios as low as 0% in February and 4% in June (Neta internal data, 2024). High inventory levels (4,247 units in warehouses and 792 at dealerships) indicate poor market absorption. A survey at GIIAS 2024 revealed that 27% of respondents hesitated to purchase Neta EVs due to low brand trust, highlighting the need for improved marketing strategies, particularly in social media features like interactivity and informativeness, to enhance brand trust and purchasing decisions (Neta internal survey, 2024). These challenges underscore the gap between policy support and market reality, requiring Neta to address brand perception and consumer confidence to achieve future targets. Interactivity allows for two-way communication that strengthens consumer and brand relationships. Research (Hanaysha, 2022) shows that interactivity has a significant positive influence on Brand Trust and purchase decisions. Informativeness help consumers understand the product and reduce uncertainty. Studies on research (Febriyani & Indriani, 2023) Indicates that informative content increases brand trust and supports purchasing decisions. Entertainment creates an emotional appeal that motivates consumers to seek out more information. (Hanaysha, 2022) Finding that entertaining marketing activities have a positive impact on brand trust and consumer attitudes. High relevance makes consumers feel that the content suits their needs. Studies on (Hanaysha, 2022) and (Zaenudin & Harto, 2024) shows that perceived relevance improves emotional relationships and purchasing decisions. Thus, the identification and analysis of these elements becomes important to understand the obstacles in achieving sales targets. An in-depth analysis of the influence of Social Media Marketing Features on brand trust and purchase decisions is expected to produce a more effective marketing strategy formulation, thereby improving the sales performance of Neta's electric vehicles in Indonesia. This study aims to (1) examine the influence of brand trust on purchase decisions for Neta electric cars, (2) analyze the effects of interactivity, entertainment, perceived relevance, and informativeness on both brand trust and purchase decisions, and (3) assess the mediating role of brand trust in the relationship between these social media marketing factors (interactivity, entertainment, perceived relevance, and informativeness) and purchase decisions. The research benefits institutions by providing insights for more effective marketing strategies to enhance consumer confidence and grow Indonesia's electric car market, while also identifying optimal digital marketing approaches. For academics, it serves as a reference for future studies on purchase decisions mediated by brand trust and influenced by interactivity, informativeness, entertainment, and perceived relevance. Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 541 While previous studies have explored the impact of social media marketing features (such as interactivity, informativeness, entertainment, and perceived relevance) on purchase intentions across various industries (Hanaysha, 2022; Febriyani & Indriani, 2023; Azizpuddin & Shahrom, 2024), this research provides a distinct contribution by focusing specifically on actual purchase decisions in the electric vehicle (EV) industry in Indonesia, using real purchase data from verified consumers of Neta electric cars. Unlike prior studies which often measured consumer intention or engagement, this study integrates brand trust as a mediating variable and applies Structural Equation Modeling (SEM) on a large sample of 389 verified buyers, enhancing the reliability of its conclusions. This empirical approach bridges the gap between theoretical constructs and actual market behavior in a sector critical to Indonesia's green economy transition. RESEARCH METHOD This study adopts a quantitative research method with a causal relationship approach, aiming to investigate how independent variables such as interactivity, informativeness, entertainment, and perceived relevance influence purchase decisions, mediated by brand trust. The analysis utilizes Structural Equation Modeling (SEM) through AMOS 22, which enables the examination of complex variable relationships in a single framework. This approach aligns with the study's objective of testing hypothesized relationships derived from existing theories. The research population comprises 76,662 individuals, including Neta Auto Indonesia's social media followers (Instagram, Facebook, TikTok) and customers who purchased vehicles between 2023 and 2024. Using Krejcie and Morgan's (1970) sampling table, a minimum sample size of 382 respondents was determined, ensuring statistical reliability while remaining feasible for data collection. The study employs non-probability convenience sampling, selecting participants based on accessibility and willingness to complete the survey. Eligible respondents must meet specific criteria: they must follow Neta's social media accounts, have engaged with its digital marketing content, be at least 18 years old, and reside in Indonesia. Data collection relies on primary sources, specifically online questionnaires distributed through Neta's official WhatsApp Call Center and social media platforms (Instagram, Facebook, TikTok). Collaboration with the company's Marketing and PR team facilitated respondent recruitment, ensuring sufficient participation. The questionnaire design incorporates screening questions to verify respondent relevance, aligning with established consumer behavior theories. This methodology ensures the collected data effectively measures the impact of social media marketing factors on brand trust and purchase decisions. RESULT AND DISCUSSION Characteristics of Respondents/Data/Sources/Descriptive Data This research was conducted with a quantitative approach and involved actual consumers who have purchased Neta brand electric vehicles in Indonesia. The main purpose of data collection Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 542 was to analyze the influence of social media marketing features on purchase decisions, with brand trust as a mediating variable. The total number of respondents in this study is 389 people, all of whom are buyers of Neta electric cars in the period from 2023 to early 2025. Thus, the data obtained comes from consumers' real experience in making purchases, not just perceptions or purchase intentions, which makes the results of this study have a high validity value. Data collection is carried out through two channels: 1. Distribution of survey links by Neta's official call center to consumers who have made purchases, and 2. The distribution of Google Form through the Neta car buyer community group in Indonesia. The sampling technique uses a non-probability sampling approach, with the convenience sampling method, because the characteristics of the respondents are very specific and cannot be selected randomly. Respondents in this study have the following characteristics: a. Age All respondents are over 18 years old, in accordance with the eligibility requirements for filling out the questionnaire. Most fall into the category of productive age, which has a high level of engagement in digital activities and access to social media. b. The domicile of the Respondents comes from various regions in Indonesia, with a distribution that reflects the geographical distribution of electric vehicle users. This shows that Neta's digital marketing strategy has reached an audience nationwide. c. Jobs From the open fields provided, respondents came from various work backgrounds, including private employees, entrepreneurs, civil servants/SOEs, and professionals. This variation shows that Neta electric vehicle users come from active economic circles. d. Income Income Information is optionally submitted by respondents. In general, the majority of respondents have income levels that reflect the purchasing power of the middle to upper class, which is in line with the price and market segment of electric vehicles in Indonesia. e. Vehicle Ownership All respondents already own Neta electric vehicles as part of their personal ownership. Some of them even stated that they had more than one vehicle, both electric vehicles and conventional vehicles. f. Based on open responses, various reasons that encourage the purchase of Neta electric cars include: operational cost efficiency, interest in environmentally friendly technology, and increasingly well-known brand image. This data also enriches the understanding of consumer behavior in the context of purchasing decisions. The characteristics of the respondents that have been described show that the data collected has a high suitability with the focus of the research. As an actual buyer of Neta's electric vehicle, respondents' perceptions and experiences are a very valid source of information to analyze the effectiveness of social media marketing and its influence on brand trust and purchase decisions. Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 543 Research Results Validity Test The following are the results of the analysis using AMOS in Phase 1 or Step 1: Figure 1. Standardized Regression Weight Or Factor Loadings Output Chart Explanation = CFA on X1 Based on the Standardized Output Diagram diagram above, it appears that all the values of the loading factor indicator are more than 0.5 so it can be stated that all indicators are valid. When detailed in the table is as follows: Table 1. Standardized Regression Weights: (CFA at X1) Estimate Validity INT4 <--- Interactivity_X1 0,693 > 0.5 Valid INT3 <--- Interactivity_X1 0,677 > 0.5 Valid INT2 <--- Interactivity_X1 0,679 > 0.5 Valid INT1 <--- Interactivity_X1 0,636 > 0.5 Valid Source : Taken from the Output of the AMOS SEM Analysis Results: Output Step 1 Based on the diagram above, it appears that all the values of the loading factor indicator are more than 0.5 so it can be stated that all indicators are valid. Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 544 CFA on X2 Table 2. Standardized Regression Weights: (CFA at X2) Estimate Validity INF5 <--- Informativeness_X2 0,779 > 0.5 Valid INF4 <--- Informativeness_X2 0,776 > 0.5 Valid INF3 <--- Informativeness_X2 0,775 > 0.5 Valid INF2 <--- Informativeness_X2 0,711 > 0.5 Valid INF1 <--- Informativeness_X2 0,701 > 0.5 Valid Source : Taken from the Output of the AMOS SEM Analysis Results: Output Step 1 Based on the diagram above, it appears that all the values of the loading factor indicator are more than 0.5 so it can be stated that all indicators are valid. CFA on X3 Table 3. Standardized Regression Weights: (CFA at X3) Estimate Validity ENT4 <--- Entertainment_X3 0,861 > 0.5 Valid ENT3 <--- Entertainment_X3 0,703 > 0.5 Valid ENT2 <--- Entertainment_X3 0,649 > 0.5 Valid ENT1 <--- Entertainment_X3 0,810 > 0.5 Valid Source : Taken from the Output of the AMOS SEM Analysis Results: Output Step 1 Based on the diagram above, it appears that all the values of the loading factor indicator are more than 0.5 so it can be stated that all indicators are valid. CFA on X4 Table 4. Standardized Regression Weights: (CFA at X4) Estimate Validity PR4 <--- PerceivedRelevance_X4 0,853 > 0.5 Valid PR3 <--- PerceivedRelevance_X4 0,581 > 0.5 Valid PR2 <--- PerceivedRelevance_X4 0,755 > 0.5 Valid PR1 <--- PerceivedRelevance_X4 0,567 > 0.5 Valid PR5 <--- PerceivedRelevance_X4 0,855 > 0.5 Valid PR6 <--- PerceivedRelevance_X4 0,856 > 0.5 Valid Source : Taken from the Output of the AMOS SEM Analysis Results: Output Step 1 Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 545 Based on the diagram above, it appears that all the values of the loading factor indicator are more than 0.5 so it can be stated that all indicators are valid. CFA on Z Table 5. Standardized Regression Weights: (CFA at Z) Estimate Validity BT1 <--- BrandTrust_Z 0,864 > 0.5 Valid BT2 <--- BrandTrust_Z 0,807 > 0.5 Valid BT3 <--- BrandTrust_Z 0,759 > 0.5 Valid BT4 <--- BrandTrust_Z 0,725 > 0.5 Valid Source : Taken from the Output of the AMOS SEM Analysis Results: Output Step 1 Based on the diagram above, it appears that all the values of the loading factor indicator are more than 0.5 so it can be stated that all indicators are valid. CFA on Y Table 6. Standardized Regression Weights: (CFA at Y) Estimate Validity PD1 <--- PurchaseDecision_Y 0,794 > 0.5 Valid PD2 <--- PurchaseDecision_Y 0,676 > 0.5 Valid PD3 <--- PurchaseDecision_Y 0,824 > 0.5 Valid PD4 <--- PurchaseDecision_Y 0,682 > 0.5 Valid PD5 <--- PurchaseDecision_Y 0,868 > 0.5 Valid Source : Taken from the Output of the AMOS SEM Analysis Results: Output Step 1 Based on the diagram above, it appears that all the values of the loading factor indicator are more than 0.5 so it can be stated that all indicators are valid. Reliability Tests Construct reliability can be measured using composite/construct reliability measure (CR) and average variance extracted (AVE) (Hair et.al, 2020). A good construct reliability measure is β‰₯ 0.70. While AVE β‰₯ 0.5. The CR value is obtained from the calculation: 𝐢𝑅 = (βˆ‘ π‘†π‘‘π‘Žπ‘›π‘‘π‘Žπ‘Ÿ πΏπ‘œπ‘Žπ‘‘π‘–π‘›π‘”)2 (βˆ‘ π‘†π‘‘π‘Žπ‘›π‘‘π‘Žπ‘Ÿ πΏπ‘œπ‘Žπ‘‘π‘–π‘›π‘”)2 + βˆ‘ πœ€π‘— While the AVE value is obtained from the calculation: 𝐢𝑅 = (βˆ‘ π‘†π‘‘π‘Žπ‘›π‘‘π‘Žπ‘Ÿ πΏπ‘œπ‘Žπ‘‘π‘–π‘›π‘”2) (βˆ‘ π‘†π‘‘π‘Žπ‘›π‘‘π‘Žπ‘Ÿ πΏπ‘œπ‘Žπ‘‘π‘–π‘›π‘”2) + βˆ‘ πœ€π‘— Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 546 Based on the CR and AVE analysis, the reliability results on X1 are shown in the calculation table below: Table 7. CR and AVE Interactivity Standardized Regression Weights: (Group number 1 - Default model) Estimate Quadratic Estimate 1 - Squared Estimate INT4 <--- Interactivity_X1 0.693 0.480249 0.519751 INT3 <--- Interactivity_X1 0.677 0.458329 0.541671 INT2 <--- Interactivity_X1 0.679 0.461041 0.538959 INT1 <--- Interactivity_X1 0.636 0.404496 0.595504 Total Construct 2.685 Number of Measurement Errors 2.195885 Construct reliability 2.685^2/(2.685^2+2.195885) 0.766522135 Total Construct Squares 1.804115 AVE 1.804115^2/(1.804115^2+2.195885) 0.597138243 So the CR value is 0.76652 where > 0.7 and the AVE value is 0.59713 > 0.5, then it can be concluded that X1 is reliable. Based on the CR and AVE analysis, the reliability results on X2 are shown in the calculation table below: Table 8. CR and AVE Informativeness Standardized Regression Weights: (Group number 1 - Default model) Estimate Quadratic Estimate 1 - Squared Estimate INF5 <--- Informativeness_X2 0.779 0.606841 0.393159 INF4 <--- Informativeness_X2 0.776 0.602176 0.397824 INF3 <--- Informativeness_X2 0.775 0.600625 0.399375 INF2 <--- Informativeness_X2 0.711 0.505521 0.494479 INF1 <--- Informativeness_X2 0.701 0.491401 0.508599 Total Construct 3.742 Number of Measurement Errors 2.193436 Construct reliability 3.742^2/(3.742^2+2.193436) 0.864569276 Total Construct Squares 2.806564 AVE 2.806564^2/(2.806564^2+2.193436) 0.782186269 So the CR value is 0.8645 where > 0.7 and the AVE value is 0.7821 > 0.5, then it can be concluded that X2 is reliable. Based on the CR and AVE analysis, the reliability results on X3 are shown in the calculation table below: Table 9. CR and AVE Entertainment Standardized Regression Weights: (Group number 1 - Default model) Estimate Quadratic Estimate 1 - Squared Estimate ENT4 <--- Entertainment_X3 0.861 0.741321 0.258679 Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 547 Standardized Regression Weights: (Group number 1 - Default model) Estimate Quadratic Estimate 1 - Squared Estimate ENT3 <--- Entertainment_X3 0.703 0.494209 0.505791 ENT2 <--- Entertainment_X3 0.649 0.421201 0.578799 ENT1 <--- Entertainment_X3 0.81 0.6561 0.3439 Total Construct 3.023 Number of Measurement Errors 1.687169 Construct reliability 3.023^2/(3.023^2+1.687169) 0.844151481 Total Construct Squares 2.312831 AVE 2.312831^2/(2.312831^2+1.687169) 0.760221208 So the CR value is 0.8441 where > 0.7 and the AVE result is 0.7602 > 0.5, so it can be concluded that X3 is reliable. Based on the CR and AVE analysis, the reliability results on X4 are shown in the calculation table below: Table 10. CR and AVE Perceived Relevance Standardized Regression Weights: (Group number 1 - Default model) Estimate Quadratic Estimate 1 - Squared Estimate PR4 <--- PerceivedRelevance_X4 0.853 0.727609 0.272391 PR3 <--- PerceivedRelevance_X4 0.581 0.337561 0.662439 PR2 <--- PerceivedRelevance_X4 0.755 0.570025 0.429975 PR1 <--- PerceivedRelevance_X4 0.567 0.321489 0.678511 PR5 <--- PerceivedRelevance_X4 0.855 0.731025 0.268975 PR6 <--- PerceivedRelevance_X4 0.856 0.732736 0.267264 Total Construct 4.467 Number of Measurement Errors 2.579555 Construct reliability 4.467^2/(4.467^2+2.579555) 0.885524285 Total Construct Squares 3.420445 AVE 3.420445^2/(3.420445^2+2.579555) 0.81934623 So the CR value is 0.8855 where > 0.7 and the AVE result is 0.8193 > 0.5, then it can be concluded that X4 is reliable. Based on the CR and AVE analysis, the reliability results on Z are shown in the calculation table below: Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 548 Table 11. CR and AVE Brand Trust Standardized Regression Weights: (Group number 1 - Default model) Estimate Quadratic Estimate 1 - Squared Estimate BT1 <--- BrandTrust_Z 0.864 0.746496 0.253504 BT2 <--- BrandTrust_Z 0.807 0.651249 0.348751 BT3 <--- BrandTrust_Z 0.759 0.576081 0.423919 BT4 <--- BrandTrust_Z 0.725 0.525625 0.474375 Total Construct 3.155 Number of Measurement Errors 1.500549 Construct reliability 3.155^2/(3.155^2+1.500549) 0.869000017 Total Construct Squares 2.499451 AVE 2.499451^2/(2.499451^2+1.500549) 0.806325903 So the CR value is 0.8690 where > 0.7 and the AVE value is 0.8063 > 0.5, then it can be concluded that Z is reliable. Based on the CR and AVE analysis, the reliability results on Y are shown in the calculation table below: Table 12. CR and AVE Purchase Decision Standardized Regression Weights: (Group number 1 - Default model) Estimate Quadratic Estimate 1 - Squared Estimate PD1 <--- PurchaseDecision_Y 0.794 0.630436 0.369564 PD2 <--- PurchaseDecision_Y 0.676 0.456976 0.543024 PD3 <--- PurchaseDecision_Y 0.824 0.678976 0.321024 PD4 <--- PurchaseDecision_Y 0.682 0.465124 0.534876 PD5 <--- PurchaseDecision_Y 0.868 0.753424 0.246576 Total Construct 3.844 Number of Measurement Errors 2.015064 Construct reliability 3.844^2/(3.844^2+2.015064) 0.879994283 Total Construct Squares 2.984936 AVE 2.984936^2/(2.984936^2+2.015064) 0.815553211 So the CR value is 0.8799 where > 0.7 and the AVE value is 0.8155 > 0.5, then it can be concluded that Y is reliable. Normality Test One of the assumption tests that must be met in SEM analysis is the normality test. The normality test of this data is intended to determine whether the research distribution of each variable is normal or not (Ghozali, 2021). Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 549 In a study using SEM analysis techniques, if the assumption of normality is met, the researcher tends to use the usual Maximum Likelihood estimation method, but nevertheless the model estimation with the Maximum Likelihood (ML) method is very sensitive to abnormal data scattering (Hair et al., 2021), so that if the assumption of normality is not met, the researcher will tend to use other estimation methods that are more resistant (Robust) against data abnormalities. The requirement for estimation using ML (Maxium Likelihood) as explained above, is normal distributed data. The following are the results of the normality analysis: Table 13. Assessment of normality Variable Min Max Skew Q.C. Kurtosis Q.C. PR6 1,000 5,000 -,837 -6,743 ,128 ,514 PR5 1,000 5,000 -,655 -5,278 ,552 2,223 PR1 1,000 5,000 -,110 -,889 ,051 ,204 PR2 1,000 5,000 -,291 -2,340 ,267 1,074 PR3 1,000 5,000 -,469 -3,775 1,210 4,873 PR4 1,000 5,000 -,767 -6,177 1,172 4,717 ENT1 3,000 5,000 -,257 -2,070 -,726 -2,923 ENT2 3,000 5,000 -,367 -2,952 -,946 -3,809 ENT3 3,000 5,000 -,281 -2,259 -,676 -2,723 ENT4 3,000 5,000 -,289 -2,324 -,724 -2,916 INF1 2,000 5,000 -,380 -3,062 -,044 -,176 INF2 2,000 5,000 -,234 -1,888 ,262 1,054 INF3 2,000 5,000 -,423 -3,404 ,439 1,766 INF4 2,000 5,000 -,668 -5,378 -,114 -,460 INF5 3,000 5,000 -,140 -1,130 -,593 -2,388 INT1 3,000 5,000 -,156 -1,256 -,567 -2,283 INT2 3,000 5,000 -,043 -,346 -,265 -1,067 INT3 2,000 5,000 -,415 -3,340 ,656 2,640 INT4 3,000 5,000 -,128 -1,030 -,589 -2,371 BT4 2,000 5,000 -,406 -3,266 -,240 -,965 BT3 2,000 5,000 -,481 -3,873 -,281 -1,129 BT2 3,000 5,000 -,365 -2,939 -,882 -3,553 BT1 3,000 5,000 -,420 -3,381 -,923 -3,715 PD5 3,000 5,000 -,480 -3,865 -1,110 -4,467 PD4 2,000 5,000 -,539 -4,338 -,599 -2,412 PD3 3,000 5,000 -,328 -2,639 -1,072 -4,314 PD2 3,000 5,000 -,293 -2,362 -,930 -3,746 PD1 3,000 5,000 -,202 -1,626 -1,037 -4,176 Multivariate 10,137 2,439 Data Source: Output of SEM AMOS Analysis Results: Output Step 1 Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 550 Based on the results of the above analysis, not all CR (Critical Ratio) indicator or variable manifest values are in the range of Β± 2.58 so that they accept H0 which means that there is a univariate normal distributed indicator which means that all data is normally distributed. Because it is distributed normally or univariately, this model meets the assumption of normality to use the ML (Maximum Likelihood) estimator. Researchers can use estimation methods that are more resistant to data abnormalities if there are problems with normality assumptions, namely by using GLS, ULS or ADF techniques. However, not all research data can be analyzed with this method due to sample count issues. Deep (Ferdinand, 2021), the recommended sample sizes for each analysis method are as follows: Maximum Likelihood (ML) Method: 100 – 200, ML or GLS Method: 200 – 500, ULS Method: 500 – 2500 and ADF Method: above 2500. Because the number of samples or observations in the study is 389 samples where the number is between 200 to 500 and meets the assumption of normality, the estimate to be used is still the ML estimate as explained previously according to (Ferdinand, 2021). Hypothesis Testing Direct Effects (Efek Langsung) The direct effects in SEM analysis will be used to answer the research hypothesis. The results based on SEM analysis using the AMOS application are as follows: Table 14. Direct effects based on SEM analysis PerceivedRel evance_X4 Entertain ment_X3 Informativ eness_X2 Interacti vity_X1 BrandT rust_Z PurchaseDe cision_Y BrandTrust _Z -,014 ,367 ,145 ,471 ,000 ,000 PurchaseDe cision_Y ,030 ,017 ,265 ,311 ,412 ,000 Data Source: Output of SEM AMOS Analysis Results: Step 2 boots Based on table 14 above, it can be seen that the values of the estimated direct effects coefficient or direct effect of one exogenous latent variable on one endogenous latent variable, for example, the direct influence of X1 on Z is 0.471 which means that X1 has a positive direct effect on Z of 47.1%. This value can be interpreted as every change in one unit of X1 can directly change the value of Z by 47.1%. The results of the analysis of the significance of direct effects based on the p value of bootstrapping results are as follows: Table 15. Direct effects based on p value PerceivedReleva nce_X4 Entertainme nt_X3 Informativene ss_X2 Interactivit y_X1 BrandTru st_Z BrandTrust_Z ,712 ,001 ,227 ,001 ... Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 551 PerceivedReleva nce_X4 Entertainme nt_X3 Informativene ss_X2 Interactivit y_X1 BrandTru st_Z PurchaseDecis ion_Y ,273 ,875 ,004 ,008 ,011 Source: SEM AMOS Step 2 Bootstrapping Analysis Output Based on table 15 above, it can be seen that the values of p value direct effects or the direct effect of one latent variable on another latent variable, for example, the direct influence of X1 on Z with a p value of 0.001 < 0.05 which means that accepting H1, namely X1, has a significant positive direct effect on Z. So based on both 14 and 15 above, it can be concluded among others: 1. Based on Tables 9 and 10, we conclude the following: 2. X1 has a positive and significant direct effect on Z (47.1%, p = 0.001 < 0.05) β†’ Accept H1. 3. X2 has a positive but not significant effect on Z (14.5%, p = 0.227 > 0.05) β†’ Accept H0. 4. X3 has a positive and significant effect on Z (36.7%, p = 0.001 < 0.05) β†’ Accept H1. 5. X4 has a negative and not significant effect on Z (-1.4%, p = 0.712 > 0.05) β†’ Accept H0. 6. X1 has a positive and significant effect on Y (31.1%, p = 0.008 < 0.05) β†’ Accept H1. 7. X2 has a positive and significant effect on Y (26.5%, p = 0.004 < 0.05) β†’ Accept H1. 8. X3 has a positive and not significant effect on Y (1.7%, p = 0.875 > 0.05) β†’ Accept H0. 9. X4 has a positive and not significant effect on Y (3%, p = 0.273 > 0.05) β†’ Accept H0. 10. Z has a positive and significant effect on Y (41.2%, p = 0.011 < 0.05) β†’ Accept H1. Indirect Effects and Total Effects According to (Haryono, 2016) Indirect effect is to analyze how strong the influence of a variable is with other variables, both exogenous and endogenous. The following are the results of the indirect effects analysis: Table 16. Indirect Effects based on SEM PerceivedRelevance_ X4 Entertainment_ X3 Informativeness_ X2 Interactivity_ X1 BrandTrust_Z ,000 ,000 ,000 ,000 PurchaseDecision _Y -,006 ,151 ,060 ,194 Source: AMOS SEM Analysis Output: Step 2 boots Based on table 16 above, it can be seen that the values of the estimated indirect effects coefficient of the exogenous latent variable, namely X1 on the endogenous latent variable (Y) through the intermediate variable (Z), for example the indirect influence of X1 on Y through the intermediate variable (Z) is 0.194 which means that X1 has a positive indirect effect on Y through the intermediate variable (Z) of 19.4%. This value can be interpreted as that every change in one unit of X1 can change the value of Y indirectly through an intermediate variable (Z) increasing by 19.4%. Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 552 The results of the analysis of the significance of direct effects based on the p value of bootstrapping results are as follows: Table 17. Indirect Effects based on p value PerceivedRelevance _X4 Entertainment_ X3 Informativeness _X2 Interactivity_ X1 BrandTrust_Z ... ... ... ... PurchaseDecision _Y ,719 ,011 ,218 ,011 Source: SEM AMOS Step 2 Bootstrapping Analysis Output Based on table 17 above, it can be seen that the values of p value indirect effects or indirect effects of one latent variable on another latent variable, for example the indirect influence of X1 on Y with a p value of 0.011 < 0.05 which means that accepting H1, namely X1 has a significant positive indirect effect on Y through an intermediate variable (Z). So based on both tables 4.23 and 4.24 above, it can be concluded among others: 1. The direct influence of X1 on Y through the intermediate variable Z is positive by 19.4% and significant or accepts H1 because the p value bootstrapping is 0.011<0.05. 2. The direct influence of X2 on Y through the intermediate variable Z is positive by 6% and is not significant or accepts H0 because the p value bootstrapping is 0.218>0.05. 3. The direct influence of X3 on Y through the intermediate variable Z was positive by 15.1% and significant or accepted H1 because the p value bootstrapping was 0.011<0.05. 4. The direct influence of X4 on Y through the intermediate variable Z is negative by 0.6% and is insignificant or accepts H0 because p value bootstrapping 0.719>0.05 Determination Coefficients and Regression Equations The following are the results of the AMOS analysis used to assess the Coefficient of Determination and Regression Equation: The Determination Coefficient and significance test results based on bootstrapping are as follows: Table 18. Coefficient of Determination Parameters Estimate Lower Upper P BrandTrust_Z ,762 ,656 ,916 ,001 PurchaseDecision_Y ,831 ,769 ,914 ,001 Source: SEM AMOS Step 2 Bootstrapping Analysis Output Based on Table 18 and the previous direct effect analysis, we conclude: The simultaneous influence of X1, X2, X3, and X4 on Z is 76.2%. Since this value exceeds 33%, the simultaneous effect is considered moderate and significant, with a bootstrapping p-value of 0.001 < 0.05. Therefore, the model is statistically significant β†’ Accept H1. Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 553 The simultaneous influence of X1, X2, X3, X4, and Z on Y is 83.1%. Since this value exceeds 16.9%, the simultaneous effect is considered moderate and significant, with a bootstrapping p- value of 0.001 < 0.05. Therefore, the model is statistically significant β†’ Accept H1. Regression Equations: To predict Z (Brand Trust): Z = 0.471 * X1 + 0.145 * X2 + 0.367 * X3 – 0.014 * X4 To predict Y (Purchase Decision): Y = 0.311 * X1 + 0.265 * X2 + 0.017 * X3 + 0.030 * X4 + 0.412 * Z CONCLUSION This This study aims to analyze the impact of various social media marketing featuresβ€” namely interactivity, informativeness, entertainment, and perceived relevanceβ€”on consumers’ purchase decisions for Neta electric vehicles, with brand trust serving as a mediating variable. The findings reveal that brand trust significantly and positively influences purchase decisions and is shaped by consumers' experiences, consistent product quality, and effective communication. Each marketing element contributes differently: interactivity enhances engagement, informativeness improves consumer understanding, entertainment fosters emotional connection, while perceived relevanceβ€”if not carefully managedβ€”can have unintended negative effects on trust and purchase intention. Furthermore, the analysis confirms that brand trust mediates the relationship between social media marketing elements and purchase decisions, acting as a crucial link that amplifies the impact of marketing efforts. These findings highlight the importance of designing social media strategies that are not only appealing but also trustworthy and value-driven. In practical terms, marketers in the electric vehicle industry should prioritize building brand trust through interactive, transparent, and meaningful content on social media platforms. Especially in a high-involvement product category like electric vehicles, trust is a key determinant of consumer behavior and loyalty. Future research is encouraged to explore these dynamics across different consumer demographics or product categories. REFERENCES Alfajri, M. ., Viranda, & Aini, Q. (2019). Pemanfaatan Social Media Analytics pada Instagram dalam peningkatan interaksi. Jurnal Ilmu Komunikasi, 8(1). Baciarlini, B. (2021). The Italian automotive industry after COVID-19 emergency: Internationalization strategy for the new electric Fiat 500 in China. Bath, V., & Nottage, L. (2020). International investment agreements and investor-state arbitration Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 554 in Asia. Handbook of International Investment Law and Policy, 1–36. Berthon, P. R., Pitt, L. F., Plangger, K., & Shapiro, D. (2012). Marketing meets Web 2.0, social media, and creative consumers: Implications for international marketing strategy. Business horizons, 55(3), 261–271. Chap, S., & Liu, J. (2024). Debt Risk Analysis of Automotive Enterprises. Economics & Management Information, 1–15. Dal Maso, G. (2024). Exploiting time in Green Visions for Thailand: How Green Finance Leverages Past Infrastructure for Future Returns. Tijdschrift voor economische en sociale geografie, 115(5), 615–627. Druehl, C., Carrillo, J., & Hsuan, J. (2018). Technological innovations: Impacts on supply chains. Springer. Febriyani, A. K., & Indriani, F. (2023). Pengaruh Fitur Pemasaran Media Sosial Terhadap Keputusan Pembelian Pada Industri Cepat Saji Dengan Kepercayaan Pada Merek Sebagai Mediator. Diponegoro Journal of Management, 12(1), 1–14. Ferdinand, A. (2021). Structural Equation Modeling: Dalam Penelitian Manajemen (5th Edition). Semarangβ€―: Badan Penerbit Universitas Diponegoro. Ghasemi-Marzbali, A. (2022). Fast-charging station for electric vehicles, challenges and issues: A comprehensive review. Journal of Energy Storage, 49, 104136. Ghozali, I. (2021). Aplikasi Analisis Multivariate Dengan Program IBM SPSS 26. Badan Penerbit Universitas Diponegoro. Hair, J. ., Black, W. ., Babin, B. ., Anderson, R. ., & Tatham, R. . (2021). Partial Least Squares Structural Equation Modeling (PLSSEM). Research and Evaluation. Hanaysha, J. R. (2022). Impact of social media marketing features on consumer’s purchase decision in the fast-food industry: Brand trust as a mediator. International Journal of Information Management Data Insights, 2(2), 100102. https://doi.org/10.1016/j.jjimei.2022.100102 Haryono, S. (2016). Metode SEM untuk Penelitian Manajemen dengan AMOS LISREL PLS. PT Intermedia Personalia Utama. Islami, R. B., Wardhana, A., & Pradana, M. (2021). The Influence of Social Media Influencer and Product Quality on Purchase Decisions (Case Study on Promotion of Hand & Body Lotion Scarlett Whitening in Instagram). Proceedings of the International Conference on Industrial Engineering and Operations Management, 2019, 560–567. https://doi.org/10.46254/eu04.20210339 Neuman, E. (2023). Arieh Sharon and Modern Architecture in Israel: Building Social Pragmatism. Routledge. Paba, S. (2022). The Chinese automotive industry at a turning point. An Overview. DEMB WORKING PAPER SERIES, 1–36. Pistrui, B. L. (2024). Effective implementation of Industry 4.0 technologies in the retail sector through theexample of service robots. Budapesti Corvinus Egyetem. Pramana, Y., Sugiat, M. A., & Azis, A. M. (2022). Evaluation of Customer Satisfaction on Mochamad Amirudin, Maria Apsari Sugiat, Fera Shinta Maria Silalahi 555 Indonesian Banking Chatbot Services During the COVID-19 Pandemic. CommIT Journal, 16(1), 69–85. https://doi.org/10.21512/commit.v16i1.7813 Rogers, D. L. (2014). The network is your customer: five strategies to thrive in a digital age. Yale University Press. Rachmawati, I., Sary, F. P., & Perdani, D. R. (2016). Influence of the use of brands in a foreign language (english) on the purchase decision process of cosmetic products made in Indonesia. Pertanika Journal of Social Sciences and Humanities, 24(July), 51–62. Suarez, C., & Martinez, W. (2019). Fast and ultra-fast charging for battery electric vehicles–a review. 2019 IEEE Energy Conversion Congress and Exposition (ECCE), 569–575. Tu, H., Feng, H., Srdic, S., & Lukic, S. (2019). Extreme fast charging of electric vehicles: A technology overview. IEEE Transactions on Transportation Electrification, 5(4), 861–878. Wardhana, A., Kartawinata, B. R., Akbar, A., & Muslimin, I. (2021). THE EFFECT OF THE USE OF INFLUENCER ON THE PURCHASE DECISION OF MSME CULINARY PRODUCTS IN INDONESIA (Study On Snack Product β€œKripik Belings” On Instagram). Proceedings of the International Conference on Industrial Engineering and Operations Management, 190–198. https://doi.org/10.46254/ap01.20210098 Winarno, K. O., & Indrawati, I. (2022). Impact of Social Media Marketing and Electronic Word of Mouth (E-Wom) on Purchase Intention. Jurnal Aplikasi Manajemen, 20(3). https://doi.org/10.21776/ub.jam.2022.020.03.15 Zaenudin, O. S., & Harto, B. (2024). Mengungkap Peran Media Sosial dalam Shopping Lifestyle Generasi Z: Insight Kualitatif dari Pengalaman Belanja Online. Innovative: Journal Of Social Science …, 4, 5624–5641.