The Role of Celebrity Endorsement on Impulse Buying Behavior with Customer Loyalty as a Moderating in the E-Commerce Industry

Purpose: Numerous studies have been conducted on celebrities to notice their effect on marketing. Celebrity endorsement has become one of the most popular marketing strategies, and Tokopedia as one of the most popular e-commerce websites in Indonesia also using these strategies for increasing their brand awareness. The purpose of this study is to explore other effects of celebrity endorsement on consumers behavior. This research aimed to define the impact of celebrities endorsement on consumer psychology, influencing impulsive buying behavior with consumer loyalty as a mediating variable. Research Methodology: This study applied quantitative survey with data was collected using an online questionnaire with sample characteristics were Tokopedia Customers and also a fan of the celebrity who lived in Indonesia by 442 samples using the purposive sampling method and data was conducted using the SmartPLS program. Result: This article indicated the findings that (1) celebrity endorsement had a positive impact on consumer loyalty, (2) a positive effect either directly or indirectly on impulsive buying behavior, and (3) consumer loyalty itself had a positive effect on impulsive buying behavior. Conclusions : This study proposes significant implications about how celebrity endorsement works and how it impacts customer behavior in the e-commerce Industry.


INTRODUCTION
Celebrity endorsers are frequently used to achived higher sales, greater exposure of advertising, positive attitude toward advertising, etc. In recent decades, numerous studies have been conducted on the impact of celebrity endorsement. Practitioners believe using celebrities in marketing to be one of the most effective strategies for increasing brand awareness (Rahman, 2018). Celebrity endorsement has always been regarded as a reliable marketing tool. The use of celebrities in marketing began in the early 20th century (Proctor & Kitchen, 2019). In various parts of the world, the technique of marketing using celebrity endorsement has been implemented; in Asia alone, the total use of celebrity endorsement as a marketing technique has reached between 25 % and 60 % (China and South Korea, respectively) (Bennett et al., 2022;Bergkvist & Zhou, 2016;Praet, 2008). Marketing through celebrities is also able to influence the effectiveness of an advertisement, brand recognition, brand recall, purchase intentions and even buying behavior (Spry et al., 2011).
The discussion about celebrities in the marketing world continues to this day due to changing marketing approaches and the magnitude of the influence of celebrities on consumer . Various new things will continue to emerge as the current era and technology evolve. These aspects can be investigated further to determine the influence of celebrity endorsements in marketing, particularly on consumers.
Consequently, the objective of this study was to observe the effect of celebrity endorsement on consumers' impulsive purchasing behavior, with consumer loyalty as a intervening variable. It is expected that this study would increase awareness of the impact of celebrity endorsement on impulsive buying and consumer loyalty.
This study aims to examine the effect of the celebrity endorsement strategy implemented by Tokopedia on impulsive purchasing behavior of Tokopedia customer who also a fan of BTS, together with the role of consumer loyalty as a mediator between celebrity endorsement and impulsive purchasing behavior.

Celebrity Endorsement as a Marketing Strategy
The use of celebrity figures as opinion leaders in marketing has been notorious since the late 19th century (Gauns et al., 2018;Roy, 2006). This celebrity endorsement is a form of marketing that creates an association between the endorser (celebrity) and the brand as part of the marketing strategy's communication process (J & Tangen, 2006;Silva et al., 2020;Till, 1998).
The purpose of using celebrities is to connect and transfer (directly or indirectly) the qualities of celebrities that consumers admire and like, such as physical, health, or lifestyle, by clicking consumers with the brands they market and increasing the attractiveness of a product or service, which can influence consumer preferences and purchase intentions (Leslie, 2011;Pramjeeth & Majaye-Khupe, 2016).
One of the models that discuss the endorser and aspects that shape the effectiveness of the source is the source attributes and received processing model. Herber Kelman, a social psychologist, developed this model and explained that the effectiveness of sources is comprised of credibility, attractiveness, and power (Shimp & Andrews, 2013). This model demonstrated that an endorser could be helpful if their credibility inspires listeners or consumers to have faith in the celebrity (Huang, 2021;Ohanian, 1990). Credibility can be established via the qualities of expertise and dependability. The credibility of the endorser will be influenced by the celebrity's knowledge of the subject matter, as well as the integrity and trustworthiness of the endorser (Shimp & Andrews, 2013).
Attractiveness is another factor that affects the effectiveness of celebrities. Here, attractiveness can refer to physical, intellectual, personality, lifestyle, athlete, etc. Similarity, familiarity, and linking are the indicators that make up the attractiveness characteristic. Fans find celebrity figures appealing because they feel a sense of similarity and familiarity with the celebrity and, therefore, like the celebrity figure. When "fans" admire an appealing star, they often adopt the celebrity's attitudes, behaviors, and interests (Shimp & Andrews, 2013). The final factor that affects the effectiveness of stars is their power, which allows them to influence consumers and get them to comply with their requests psychologically.

Impulse Buying Behavior
Impulsivity is the tendency to experience sudden and spontaneous urges to make a purchase on the spot and to act on a perceived desire with little consideration or evaluation of the action's consequences (Beatty & Elizabeth Ferrell, 1998;Zhang et al., 2018). that consumer emotions or emotional states influence impulse purchasing (Chang et al., 2011;Donovan & Rossiter, 1982;Rook, 1987;Weinberg & Gottlieb, 1982) Impulse buying has always been considered important for marketers, because it can account for 40% to 80% of consumer purchases depending on the type and category of the products (Amos et al., 2014;Parmar et al., 2020). Even though impulse buying can have positive or negative effects on consumers, consumers tend to buy unnecessary items, which can lead to post-purchase dissonance, where they may feel joyful or otherwise experience the negative consequences that must be obtained from buying impulsively, such as economic consequences, etc. (Mittal et al., 2018).
In previous studies, impulse buying generally focused on specific variables in the offline retail environment (physical features and social components) that trigger impulse buying (Herabadi et al., 2009). Recently, research on impulse buying is no longer focused on offline but online, and many factors in online purchases can affect impulse buying. The ease of payment, the large variety of choices and the availability of personalized recommendations as desired are potent tools that encourage impulse buying through online media (Aragoncillo & Orus, 2018).
According to previous research, the impulse buying process began with the emergence of a desire to purchase due to sudden curiosity and a sense of connection to a particular product or brand among consumers. The sudden urged to buy is then processed psychologically, resulting in emotional stimuli and manifestations; feelings that suddenly arise to make a purchase cause consumer to feel pleasure and excitement (Mittal et al., 2018).
These emotions lead to psychological conflicts in consumers, where there is a desire to comply with wishes, but a rational side of consumers attempts to resist them. This psychological conflict then develops into ignoring the consequences (Mittal et al., 2018). When consumers follow their desire to make an impulsive buying, they tend to disregard the potential effects of the purchase. The final stage of the psychological dilemma is rationalisation (Mittal et al., 2018), in which consumers attempt to justify their impulsive purchasing behavior to avoid feeling guilty for giving in to the desire to make impulse buying.
Furthermore, this stage was adapted into dimensions, namely the dimension of the impulse to make an impulsive purchase, comprised of indicators of sudden curiosity and connection to the brand, the dimension of emotional stimulus and manifestations, comprised of feelings of excitement and pleasure, and the dimension of psychological dilemma, comprised of indicators of psychological conflict, ignoring consequences and rationality of actions, and the final dimension of post-purchase dissonance . (Mittal et al., 2018).

Consumer Loyalty
Loyalty is a frequent purchase of a product or brand (Oliver, 1999;Tellis, 1988). Consumer loyalty is the result of a marketing strategy's success in a market with significant competitiveness to create value for consumers (M. L. Li & Green, 2011;Marques et al., 2017). Oliver (1999) presented a definition of loyalty that was quite different from the previous understanding; he stated that loyalty is a commitment that is firmly held by consumers to repurchase or re-subscribe a preferred product/service consistently in the future, thereby leading to the purchase of the same brand or set of the same brands, despite situational influences and marketing efforts that have the potential to cause switching consumer behavior.
In previous research, loyalty was believed to be formed in several phases, beginning with a cognitive phase, then an affective phase, then a conative phase, and concluding with an action phase (Oliver, 1999). Cognitive commitment is the obligation to remain loyal to a brand for ethical or moral reasons (Cuesta-Valio et al., 2020). The cognitive phase is also the phase of brand loyalty based solely on brand trust. This phase typically relies on prior knowledge or information gleaned from recent interactions with a brand or product or service (Oliver, 1999). This phase is characterised by shallow loyalty; however, if satisfaction at this stage is successfully processed and integrated into the consumer experience, the phase of consumer satisfaction can advance to the experimental phase.
The second phase of consumer loyalty is affective, which tends to be more profound than the cognitive phase. At this stage, consumer loyalty is formed from feelings of satisfaction and liking for a product or brand and the consumers' overall evaluation of the brand or product. (Ahn & Back, 2018;Han et al., 2011;Li et al., 2020).
The third stage of loyalty is conative loyalty. Conative loyalty is connected to taking any action; in this context, the action is the intention to make a repeat purchase. Conative loyalty is influenced by the positive effect that arises on a brand that motivates consumers to form a greater intention to make a purchase (Almeida-Santana & Moreno-Gil, 2018; Kim et al., 2015;Li et al., 2020;Pedersen & Nysveen, 2001).
The last stage of loyalty is the action stage. Previous research has explained that the intention converted into action is called "action control" (Kuhl & Beckmann, 1985 ;Oliver, 1999). The intention formed on conative loyalty is then converted into an action to make a purchase at this stage.

Conceptual Framework
The concept model of this study itself was taken from the concept of the stimulus-response model (SOR) model (Kotler & Armstrong, 2018). The SOR model derived from psychology. The Stimulus-Response Model, also known as the organism response stimulus theory, assumes that the "stimulator" directs an internal process to produce a response (Mehrabian & Russell, 1974;Moore, 1996;Russell & Pratt, 1980;Zafar et al., 2020). Kotler and Armstrong (2018) explain that the stimulus will produce an internal process called the "black box", where each individual would process the received stimulus differently, which was influenced by various external and internal factors such as consumer characteristics and decision-making processes. Subsequently, consumers would produce a specific response, such as buying behavior, preferences, and consumer relationships with a particular brand or company.
This study had three variables: the independent variable or X 1 , celebrity who do the endrosement acts as a stimulus. Another variable was the dependent variable, where the dependent variable or Y 1 , was a variable influenced by other variables which affect the response to the given stimulus, whereas the dependent variable was impulse buying behavior. The third variable was the intervening variable or Z 1 , an intermediary variable that mediates or bridges the influence of the independent variable on the dependent variable. The intervening variable in this study itself is Consumer Loyalty.

Celebrity Endorsement towards consumer loyalty
According to previous research on brand ambassadors or celebrity endorsements, a sense of trust in celebrities can be transferred into perceptions of brand image, credibility, and brand loyalty (S. S. Kim et al., 2014). Other research indicated that celebrity endorsements can influence consumers' perceptions of brand quality and encourage brand loyalty (Osei-Frimpong et al., 2019). Based on prior research, celebrity endorsement had a positive effect on consumer loyalty; therefore, the following hypotheses were developed for this study: H 1 : There was a significant positive effect between celebrity endorsement on consumer loyalty.

Consumer Loyalty towards Impulse Buying Behavior
It is believed that consumer loyalty has a positive effect on impulse purchasing. According to previous research (Šeinauskienė et al., 2015), brand loyalty has a positive impact on the propensity to engage in impulsive purchasing. Based on this study, the following hypotheses were developed: H 2 : There was a significant positive effect between consumer loyalty on impulse buying behavior

Celebrities Endorsement towards Impulse Buying Behavior
Over the past few decades, numerous studies have been conducted on the role of celebrities in marketing. Several prior studies indicated that celebrity endorsements play a role in the process of impulsive purchasing; celebrity endorsements or brand ambassadors serve as triggers or stimuli for the emergence of feelings to buy suddenly (Mittal et al., 2018). On this basis, the following hypotheses were developed for this study: H 3 : There was a significant positive influence between celebrities endorsement on impulse buying behavior.

METHODS
This study employed a single case study quantitative method approach. In this research, a case study was conducted on the Tokopedia's Brand Ambassador, BTS, and their influence on consumer purchasing behavior; in this context, BTS fans as a consumer, serve as the dependent variable, while consumer loyalty to the brand serves as the mediating variable. This investigation applied both descriptive and explanatory research objectives.
Descriptive research aimed to characterise the relationship between the studied dependent and independent variables (Mahmoud et al., 2018). Moreover, the purpose of the explanatory research permitted researchers to determine the relationship between celebrity endorsement variables, impulse buying behavior variables, and consumer loyalty variables.

Sample & Data Collection
This study's primary data was collected via a self-administered online questionnaire with closed-ended questions. Each Questionnaire item was developed and adapted based on a review of the relevant literature. The sections of this questionnaire are Section A and Section B. Section A consists of respondent-specific information, such as age, gender, and closed-ended questions to ensure that the respondent is a user of the Tokopedia e-commerce platform.
On the other hand, Section B was comprised of questions regarding the variables under investigation: Celebrity Endorsement, Impulsive Buying Behavior, and Loyalty. Each item contained indicators derived from prior research; each indicator was measured using a 5-point Likert scale ranging from strongly disagree (1) to strongly agree (5).

Celebrity Endorsement
Celebrity endorsers are familiar with the brand. (Shimp & Andrews, 2013) Celebrity-provided information can be trusted.

Celebrities effectively communicate brand image and concept
Celebrities generate consumer interest in a brand.
Adoration for celebrities drives consumer consumption.
Consumers will continue to purchase the brand as long as the celebrity endorse it.

Consumer Loyalty
Consumers prefer this brand over competing brands.
Consumers intend to make additional purchases of the brand in the future.
Consumers purchase the same brands repeatedly for every need.

Impulse Buying Behavior
A sudden intrigue that appears to shop (Mittal et al., 2018) The feeling of being suitable for using the brand, so that there is a desire to buy every time it is related to the brand Feeling excited and delighted to make an impulsive purchase using the brand.
The impulse to make an impulsive purchase is contrary to conscience.
Ignoring the repercussions of an impulsive purchase

Justifying impulsive purchases
Feeling of pleasure towards spontaneous purchases made Sampling using a technique of purposive sampling. This research targeted BTS fans who were also active users of e-commerce applications. This study's population was based on the total number of Tokopedia visitors in Q4 2021 (October to December) of 149.61 million (iPrice Group, 2022). Based on research on Indonesian e-commerce users, those aged 26 to 35 account for 48% of Indonesian e-commerce users, followed by those aged 18 to 25 with 28% (Kredivo & KataData Insight Center, 2022). Based on this investigation, the criteria for this study's respondents were that they resided in Indonesia, were fans of BTS, were between the ages of 17 to 35, and used the Tokopedia application at least three times. This survey would be conducted for two months, beginning in December 2021 and ending in January 2022. To target respondents with the appropriate criteria, the survey was published online via the social media platform Instagram. In the meantime, the Questionnaire was filled out using the Google Forms application, whose settings were modified to ensure that each respondent met the specified criteria and answered each question in full.
Instagram as a platform for disseminating questions is motivated by the application's higher level of interaction and engagement than other social media platforms (Silva et al., 2020;Valentini et al., 2018). In addition, the distribution of questionnaires on Instagram media was dominated by Instagram users, assisted by the Indonesian BTS fanbase with the username @7bts.update. This Fanbase account was created in 2015 and had 974 thousand followers as of February 2022. This survey's link is accessible via the IG Story feature, allowing respondents to open the link and access the survey. The spread of this link lasted four times longer than the survey period.
The number of samples in this study was calculated using the Yamane formula, developed by Taro Yamane and known as the formula for calculating sample size (Yamane, 1962). With a precision value of 5% of the total population of 149.61 million, the minimum number of samples required for this study according to Yamane's formula is 400. The results of an online survey distributed via Instagram collected 442 valid respondents who met the criteria for this study's respondents. Furthermore, the collected data was processed using the SMART PLS software.

RESULT AND DISCUSSION
This study's data were collected using a Structural Equation Model (SEM) approach to facilitate the testing of hypotheses. The SEM model consists of two primary components: the measurement and structural models. The measurement model describes the relationship between latent variables and their corresponding indicators (manifest variables), whereas the structural model describes the relationship between each latent variable.

Measurement Model
This research aimed to determine the connection between celebrity endorsement (CE), Impulse Buying Behavior (IB), and Consumer Loyalty (CL). This study had three latent variables: CE, IB, and CL. Each was composed of an indicator (manifest variable). The reliability and validity of the findings were evaluated in these findings. In Partial Least Square-Structural Equation Modeling (PLS-SEM), there is convergent and discriminant validity. In measuring convergent validity, the factor loading and Average Variance Extracted (AVE) analysis can be used, indicators are declared valid if they have a value of Factor Loading greater than 0.70 (Bahadur et al., 2020). While in AVE, Furthermore, the AVE value of each construct must exceed the minimum threshold of 0.50 for convergence validity to be deemed adequate Khan et al., 2021). Based on Table III. the factor loading value in this study ranges from 0.701 to 0.92. And the AVE values for each construct ranged from 0.534 to 0.597. Based on Table III regarding the result of AVE and factor loading analysis, it can be concluded that each variable in this study is considered valid.
In measuring discriminant validity, Fornell-Larcker Criterion can be used. (Fornell & Larcker, 1981;Khan et al., 2021;Vinzi et al., 2010). Based on Table IV. the result of Fornell & Larcker method revealed that the square root of the AVE was more significant in the correlation between constructs, indicating that each variable was an acceptable discriminant for its validity.
In assessing the reliability value, Cronbach's alpha and composite reliability (CR) approaches can be used. According to (Gokmenoglu & Amir, 2021), Cronbach's Alpha values above 0.80 could be categorized as excellent, then the composite reliability value should be greater than 0.7 to be considered as a reliable instrument. (Arghashi et al., 2021;Hair et al., 2017). Based on Table III. Cronbach's Alpha values from all the constructs below had met Cronbach's alpha standards and could be evaluated as reliable, with Cronbach Alpha ranging from 0.921 to 0.948. For Composite reliability (CR) values, based on Table III. ranged between 0.932 and 0.954, and could be evaluated as reliable. Therefore, the validity and reliability of this study's constructs could be accepted based on these values.

Structural Model
In assessing the structural model using Partial Least Square (PLS), the most commonly used to evaluate the structural model is the coefficient of determination (R 2 value), this coefficient is measure of the model's predictive accuracy and is calculated as the squared correlation between a specific endogenous construct's actual and predicted value (Hair, Hult, et al., 2014). The R 2 value of 0.67 can be interpreted as the model is strong, R 2 value of 0.33 can be interpreted as a moderate model, and an R 2 value of 0.19 can be interpreted as a weak model (Chin, 1998).
Based on Table V., it can be concluded that the coefficient of determination (R 2 ) value of 0.303 is interpreted that customer loyalty variable of 30.3% can be explained by the variable of celebrity endorsement, whereas the remaining of 69.7% is explained by other variable. The value of R 2 shows that the model is weak. Then the value of the coefficient of determination (R 2 ) of 0.585 is interpreted as impulse buying behavior variable can be explained by 58.5% of the variable of celebrity endorsement and customer loyalty, while the remaining 41.5% is explained by the other variable. R 2 values show that the model is moderate. In addition to evaluating the magnitude of the R 2 value as a criterion of predictive accuracy, the predictive relevance (Q 2 ) test is conduct to determine the predictive power using the blindfolding procedure. In structural model, Q 2 values larger than zero indicate the path model's has predictive relevance (Hair, Hult, et al., 2014). Based on the predictive relevance (Q 2 ) test using the blindfolding procedure, it can be concluded the predictive relevance (Q 2 ) value of 0.213 and 0.337 indicate that the model has predictive relevance. This study use Path Coefficient to test the relationship between the predetermined hypotheses. The results are shown in Table VII, it can be concluded that celebrity endorsement has the largest coefficient and has a positive direction that affect impulse buying behavior variable by 0.623. In addition to the direct effect of hypothesis testing, also shown that celebrity endorsement has an positive indirect effect to impulse buying behavior by 0.343. The celebrity endorsement has the second largest coefficient and has a positive direction which is affect the consumer loyalty variable of 0.550. Based on the result, consumer loyalty has the least coefficient and positive direction, affecting the impulse buying behavior variable by 0.219.
In this path coefficient and hypothesis testing, if t-statistic values are greater than 1.96 and the p-values are less than 0.05, it can be interpreted that the hypothesis is supported and has a direction of positive correlation (Garson, 2016). Based on Table VII and Table VIII, the result of path coefficient test and hypothesis testing, also indirect effect testing, the result of the test can be concluded that the first hypothesis, the second hypothesis and also the third hypothesis are supported.

Discussion and Implication
Based on the result of testing the Hypothesis 1 (H 1 ) prove that customer endorsement positively affects consumer loyalty. This is proven by the t-statistic value 16.375, greater than 1.96 and the p-value of 0.000, less than 0.05. It can be concluded the first hypothesis (H 1 ) is supported. This result conforms to a previous study by Osei-Frimpong et al. (2019), which shows that celebrity endorsement affects consumer perceptions of brand quality and encourages brand loyalty. This finding also supports the other previous research of S.S Kim et al. (2014), which shows that the relationship between the attributes of celebrities who act as endorsers is heterogeneous on the credibility of the company or loyalty to the brand or company. He further explained that each celebrity has a different ability to influence consumer loyalty to the brand; it is emphasized that not all celebrities are attractive to all potential consumers; in some cases, more than one celebrity is generally used to cover a large target market. With the concept of the SOR model, celebrities here can play a role as a stimulus that can provide a stimulus to consumers who admire them for being able to react in the form of loyalty to the brand they endorse. In this study, BTS as a brand ambassador of Tokopedia is important and positively affects consumer loyalty to Tokopedia. As a result, Consumer, especially those who are also fans of BTS, are loyal to Tokopedia based on their admiration for BTS which makes them ready to support Tokopedia and remain loyal to Tokopedia. Osei-Frimpong et al.(2019) also found that there was also a positive effect of brand loyalty on consumers' purchase intentions. This finding supports this study's second hypothesis (H 2 ) that prove consumer loyalty positively affect impulse buying behavior. This proven by the hypothesis test result, the t-statics value of 16.717, greater than 1.96, and the p-value less than 0.005 which is 0.000. Therefore, it can be concluded that the second hypothesis (H 2 ) is supported. The result of this hypothesis align with the result of previous research by Šeinauskienė et al. (2015), who see that there is a tendency when loyalty to a brand increases, according to this study's findings, loyalty positively impacts impulse buying, The findings in this study are also supported by Gulfraz et al. (2022), who examined the effect of consumer attitudinal loyalty from an e-commerce platform on online impulse purchases. In this study, attitudinal loyalty acts as a mediator of consumers' experience of online shopping with impulse buying. The results show that there is a positive effect of online shopping experience on loyalty and impulse buying, as well as a positive influence of consumer's attitudinal loyalty on impulsive purchases. Positive feelings from consumer experiences can increase feelings of liking that lead to consumer loyalty which can have an impact on consumers' impulsive purchases of brands. In this study, consumer loyalty is important and positively affect impulse buying behavior of customer to shopping at Tokopedia because when consumers feel loyal to a brand, they tend to make impulse purchases more easily. When customer feeling attached to Tokopedia, they are feeling safe so customer are easily to purchase something based on their impulsive behavior.
Based on the result of the last hypothesis testing, is proven that celebrity endorsement positively affect impulse buying either directly or indirectly. Proven by the t-statistic value of directly effect is 6.240 and for the indirectly effect is 11.074, which is greater than 1.96, and the p-value is 0.000 either in directly or indirectly effect, is less than 0.005. Thus, it can be concluded that the third hypothesis (H 3 ) is supported.. the result of this hypothesis are in line to previous research, which explains that celebrity endorsements have a significant influence on the effectiveness of an advertisement, brand recognition, brand recall and purchase behavior of customers (Osei-Frimpong et al., 2019;Spry et al., 2011).
The results of this study also support previous research, which states that celebrity self-disclosure positively affects the attachment between celebrities and their fans, and the expertise and knowledge of celebrities also positively affect impulsive buying behavior . These results are consistent with the study of Meng et al. (2021), which shows that celebrities can influence consumer emotions; celebrities can gain consumer admiration which is then transmitted into products marketed by the celebrity's personality, which increases the intention to make a purchase.
This hypothesis also in line with another previous study by Mittal et al. (2018), which stated that influencers play a role in encouraging emotional stimuli that encourage consumers to take impulsive buying actions in an impulse buying process. The study by Parmar et al. (2020) shows that celebrities whom consumers admire can provide pleasant feelings that motivate them to make impulse purchases; celebrities tend to be more likely to have a more significant influence on encouraging consumers to make impulse purchases when involving products or brands being marketed.
In this study, BTS as a brand ambassador who endorse Tokopedia is important an positively affect Tokopedia consumer, specially those who are BTS fans, impulse buying behavior. BTS can provide special and pleasant feeling to motivate their fans to make impulse purchase. BTS as a celebrity who endorse Tokopedia, have a power to influence consumer emotion and gain their admiration to make their fan have an urge to buying something impulsively from Tokopedia to show their support for BTS.

CONCLUSION
This study aims to examine the effect of the celebrity endorsement strategy implemented by Tokopedia on impulsive purchasing behavior of Tokopedia customer who also a fan of BTS, together with the role of consumer loyalty as a mediator between celebrity endorsement and impulsive purchasing behavior. Respondents in this study were 442 Tokopedia customers, who are also a fan of BTS, who live in Indonesia and aged around 17-18, and have been do a transaction on Tokopedia at least three times. Hypothesis testing and data analysis were undertaken using the Smart PLS version 3.0 application program.
Based on the result of the analysis and discussion above, the conclusion that can be drawn from this research are, celebrity endorsement positively affects consumer loyalty of Tokopedia, celebrity endorsement also positively affects, either direct or indirectly, impulse buying behaviour of consumer on Tokopedia. consumer loyalty of Tokopedia positively affects on impulse buying behavior of consumer.
The implications of this study can provide significant contributions and benefit to the research literature. Although many previous studies have discussed the influence of celebrity endorsement, with the development of the era, the influence of celebrity endorsement on consumers has also change. The theoretical implications of this study are providing academic contribution as an empirical research by examining the variable celebrity endorsement that affect consumer loyalty and impulse buying behaviour on Tokopedia, as one of the biggest e-commerce in Indonesia.
Meanwhile, the practical implication of this study is to contribute to the development of celebrity endorsement as a marketing strategy and to provide useful insight to e-commerce business sector about the impact of celebrity endorsement as a marketing strategy. It is expected that player in the e-commerce industry or other business industry can follow the success of Tokopedia by choosing the best marketing strategies to gain attention, admiration, and loyalty of consumer.
This study was designed with several limitations. First, limitations in terms of cost, effort, and a short time have resulted in researchers limiting research focuses on only one e-commerce platform and is limited geographically to Indonesia; furthermore, the sample size obtained is relatively small, and the scope of the discussion is not overly expansive. This limitation can be considered for future research, where future research can consider conducting a more in-depth discussion about aspects of impulsive purchasing behavior and loyalty, as there are not yet many studies discussing the relationship between these two variables.