Investigating the impact of digital influencers on consumer decision-making and content outreach: using dual AISAS model

Abstract With exponential rise of social media, marketers identify the power and effectiveness of influencer’s advertising on social networking site (SNS). Despite comprehensive understanding of the effects of influencers, their outreach to large audience is yet to be addressed. In this article, we have investigated the effects of fashion influencers on consumers’ decision-making processes and their content outreach on Instagram by embracing new behavioral consumption model ‘dual AISAS model’, which is upgraded version of AISAS Model. It is based on theoretical grounding theory of buying behavior and multi-step flow theory. Both offline and online surveys were conducted involving 969 Pakistan Instagram users following digital influencers. Valid data was assessed and analyzed through structural equation modeling. Our findings demonstrate that every path in dual AISAS model is found significant and have profound effect. It reveals that fashion influencers exert powerful influence on consumers’ decision-making process. Being so influential, they grab the consumers’ attention immediately, engage them, and get wider outreach by upturn in consumer intention in order to spread the fashion content within private networks as well as extended networks. The findings hold robust implications to both theory and practice. Some limitations of the present study offer boulevards to future scholars.

and services with others (Merz et al., 2018). Moreover, social media is playing vital role in digital market as the digital influencer (Mavroudis, 2018). On social networking sites (SNS), online branding of products through business accounts and advertisements has proven to be an efficient marketing strategy (De Vries & Carlson, 2014). On social media, certain users actively create online content (user generated content) and become an opinion leader. Their shared post enriched with reviews, emotions, and personal experience can influence the products, brands, and potential audience (Hsu et al., 2013). Therefore, we have named them as 'digital influencers' (Susarla et al., 2016). Digital influencers have been recognized as more influential (Thakur et al., 2016) thanks to large number of followers (Jin & Phua, 2014). They are ubiquitous in all fields like food (Sony Kusumasondjaja & Tjiptono, 2019), travel (Song et al., 2017), and fashion (Casal o, 2018), but they have seized a mammoth attention specifically in fashion industry (Delisle & Parmentier, 2016). In fashion industry, the opinion of influencers is worthwhile (Wiedmann et al., 2010) because the fashion posts act as a source of inspiration for consumers and have high propensity to influence the consumers' purchase decision (Susarla et al., 2016;Zhu et al., 2015) or shopping behavior. Moreover, it is evident that fashion seekers aid the influencers to spread the content or diffuse the fashion trends within their private network and extended networks (Goldsmith & Clark, 2008) by leveraging the power of social media (Lipsman et al., 2012).
Fashion related products urge the consumers to use visual centric-platform such as Instagram (Mull & Lee, 2014), which reported that its monthly active users were 1 billion in 2018 (Statista-The Statistics Portal, 2018). In this context, Instagram seems reasonable to showcase the products (Locowise, 2017;Streimikien_ e et al., 2021). The basic premise behind gaining the popularity of Instagram is that it, unlike other social networking sites, provides an opportunity to brands to post the content in an aesthetic appealing way (image-sharing, photos, videos, stories etc.) (Lyfe Marketing, 2018). The nature of Instagrammer influencer posts encompasses images and videos with textual description. Prior studies have focused primarily on textual platforms such as blogs (Li & Du, 2011), Twitter (Park & Kaye, 2017), and other online communities (Li et al., 2013) with regards to influencers . Despite gaining popularity due to its aesthetic visual appeal, which may have a greater impact on users' purchase decision (Mediakix, 2017), studies conducted on visual social networking site i.e. Instagram (Abed, 2018;Li et al., 2018) are limited.
Nevertheless, prior studies have been dedicated to digital influencers and their effects on consumers' decision-making processes need to be explored further. We intended to fulfill the slit by enhancing the understanding about the effects of digital influencers on consumers' decision-making processes. Our study has focused on Instagram bloggers of fashion industry as influencer marketing is intensively used in this industry (Garland & Reed, 2018). Additionally, we employed a novel consumption model 'The Dual AISAS Model' instead of traditional AISAS model because 'traditional AISAS TM ' only captures the 'purchasing desire' resulting mostly from the posts/photos shared by influencers. Since social influence theory corroborates that consumers' decision-making process is greatly influenced by the reviews, remarks, and suggestions of people connected with them (Zhang et al., 2014), in the prevalent digital network era, consumers can easily seek diverse or detailed information not only in the shape of ads but also in the form of posts or pics from groups and networks etc. in order to get maximum utility from their purchases (Crandall et al., 2008). Dual AISAS model has two flows; first flow is 'traditional AISAS', and the second flow is new 'A þ ISAS', which captures the 'sharing desire' spreading of information by consumers within entwined network (Shintara and Yuji, 2018). We have considered this model since it enables the researcher to measure the impact of digital influencers on consumers' decision-making processes (i.e. capturing the offline movement of consumers by assessing their purchasing desire) as well as to identify their content outreach (i.e. capturing the online movement by assessing the reaction of consumers to advertisement or consumer sharing desire). Moreover, it is in line with Chahal (2016) who stated that influencer outreach will get an increasing focus in the future.
The study is designed with following objectives: 1) on the basis of dual AISAS Model, we intend to investigate the impact of digital influencers on consumers' decision-making processes, and 2) on the basis of dual AISAS Model, we intend determine the influencers' published promotional content outreach (coverage). The current study makes the contributions in following ways. This study contributes to the literature of consumers' behavior, digital influencers, and consumers' decision-making processes by examining the effects of digital influencers and determining their content outreach. Furthermore, our study considered 'Dual AISAS Model' as research framework to get better insight on influencer communication and resultant engagement because influencer has the ability to get people engaged on shared image and drive them into action (Freberg et al., 2011) and ultimate outreach of promotional content to larger audience (Nabi et al., 2019). However, we used survey method/technique to complete this research. Finally, the findings of the study assists both brands and digital influencers to identify the right influencers that best resonate with the targeted audience. Additionally, through the lens of this model they can measure the consumer reactions (or consumer behavioral and purchasing data (e.g. monthly basis and quarterly basis etc.)) towards Insta fashionista photos.
The rest of the study is organized as follows. Section 2 presents theoretical grounding, followed by research model development (Section 3). Section 4 describes the employed methodology followed by analysis and findings (Section 5). The last section (Section 6) deals with discussion and concluding remarks about findings and theoretical as well as practical implications of the findings in addition to the limitations and boulevards for future researchers.

Theoretical grounding of dual AISAS model
While consumer's behavior is in debate, extant studies have found that consumer's activity processes or purchase decision processes have remained its prime concern (Zhang et al., 2018). To evaluate the processes of consumer decision, a model of consumer decision was initially developed by Howard (Du Plessis et al., 1991). Later on, in 1969, an improvement was made in traditional model by Howard and Sheth (Du Plessis et al., 1991). This model is termed as 'Theory of Buying Behavior'. The theory of buying behavior contained psychological, marketing, and social features. The assimilation of these factors affects the consumers' decision-making processes, and transforms them sequentially into information processing. The five-resultant outcomes of buying behavior model are attention, comprehension, attitude, intention, and purchase (Kavak et al., 2015). In numerous business buying studies (Kavak et al., 2015), especially in retailing sector (De Oliveira & Gosling, 2015), hospitality, tourism (Karl & Reintinger, 2016), and social media (Cheah et al., 2019), theory of buying behavior has been adopted on theoretical grounding.
The present study also utilizes another theory as theoretical foundation titled as the multi-step flow theory. This theory is basically an extension of the two-step flow theory (Katz & Lazarsfeld, 1955). As the name implies, this paradigm has multidimensional flow; goes through multiple 'channels', i.e. from mass media to online opinion leaders (i.e. digital influencers), and then eventually from opinion leader to wider target audience (Ognyanova, 2017). Successful opinion leaders (i.e. influencers) have an extended network or are connected with more concentrated network (Basille, 2009). Albeit they convey the notion to their followers/subscribers, who in turn spread those notions to other audience (Brosius & Weimann, 1996). Transferring this notion into the context of present study, social networking sites like Instagram, YouTube, Facebook and Twitter provide their audience an option to forward the content to their connections with the click of button. A number of studies have employed multi-step flow theory in media (Brosius & Weimann, 1996) and social media studies (Ognyanova, 2017;Stansberry, 2017).
Therefore, this study embraced two theories as theoretical foundation; theory of buying behavior to investigate the impact of digital influencers on decision-making processes of users (Opinion leaders have the potential to influence the purchasing decision of others across specific product categories (Flynn et al., 1996), and multistep flow theory to determine their content outreach.

Digital influencers
The rapid increase in content generated on microblogging platforms compels the network participants to make an attempt to grab the attention, and thus have an influence on peers (Trusov et al., 2010). These opinion leaders have the tendency to influence the members of their social networks by disseminating the information (Chaney, 2001). The studies on information diffusion have particularly concerns about how influential people interact and influence their interpersonal networks. These opinion leaders have now turned into online opinion leaders in the form of bloggers and vloggers (Agarwal et al., 2014), and have gained a power to influence the behavior of others on social media (Ki & Kim, 2019). These influential people on social media are named as social media influencers/digital influencers/digital celebrities. They have accumulated a huge fan base by creating and sharing content on multiple social networking sites. Moreover, brand focuses on identifying these influential people to communicate a brand's message to mass consumers (Talavera, 2017).
They engender a meaning transfer process, wherein they stimulate interest and influence the consumers' attitude towards products or brands and purchases (Sokolova & Kefi, 2020). Furthermore, extant literature found the influence of promotional content on different e-commerce platforms, such as Facebook (Aswani et al., 2017b), Twitter (Aswani et al., 2017a), Instagram (Ki & Kim, 2019), and other prevalent domains. It revealed that influence is manifested in fashion (Wiedmann et al., 2010), education (Tess, 2013), and healthcare sector (McNeill & Briggs, 2014). Moreover, in virtual communities, it is evident that digital influencers have a momentous effect on consumers' pre-purchase behavior (Sokolova, 2019) and post-purchase behavior (Song et al., 2017). Since they are influential opinion leaders (Agarwal et al., 2014), they have the tendency to persuade their peers (Weeks et al., 2017) to take a specific action (Clow & Baack, 2016) in order to further spread the content to other users (Goldsmith & Clark, 2008). These influential personalities are familiar with their own follower preference, thus they can spread content to more intended consumers with definite consumer psychology that conforms to their admirers (Wei, 2017), and provokes to share it in their own social networks, thus extending the dissemination of content (Kown, 2016). The quicker dissemination or wide outreach of content facilitates in accessing the novel information in network and drives the user engagement (Yu et al., 2015). Moreover, these personalities assist the firms to discern novel consumers' trends and upkeep the new needs of consumers (Aldhaheri & Bach, 2013).
Despite gaining the popularity, a little effort has been made to investigate the effects of digital influencers on consumers' decision-making. Moreover, a little effort has been made to determine the influencers' content outreach (Chahal, 2016). Therefore, we have investigated the effects of digital influencers on consumers' decision-making and have determined their content outreach.

New consumption behavior model (dual AISAS), decision processes, & hypotheses development
With the passage of time, consumers ' pattern of buying behavior has changed with the evolution and development of media, and the purchasing behavior model has undergone a change. Various models about consumers' decision-making processes have been used by advertising agencies. The widely used consumer buying behavior model and consumer activity model are AIDA and AIDMA model (Hawaldar & Ullal, 2018;Wei & Lu, 2013). AIDA incorporates four stages: attention, interest, desire, and action. AIDMA incorporates five stages: attention, interest, desire, memory, and action (Hassan et al., 2015). Subsequently, improvement was made in AIDMA model by well-recognized Japanese advertising agency 'Dentsu' (2008). This agency proposed another consumer activity model AISAS (Dentsu Incorporated, 2008). The AISAS model has been created to meet the current needs of digital era, and due to its interactive nature of internet, it will better to elucidate the behavior/ activity of consumer on the internet (Wei & Lu, 2013). This model incorporates five processes; attention, search, interest, action, and share. This model has some similarities with AIDMA model. Initially, the advertisement grabs the attention of customer, and the customer becomes interested in it. The third path of AISAS model is novel, search, and gather information from different sources on the internet. The fourth phase is following consumers' action to make a purchase and subsequently consumers share their experiences on the internet (Cheah et al., 2019). The process after the action has expanded, which illustrates the current consumer behavior. However, this model is not enough to explain the phenomenon which Instagram triggered.
Nowadays, people firmly believe on word-of-mouth communication (WOM) rather than ads. So, the mode we used earlier to get people's attention has become obsolete. On daily basis, we can find information about products or services in the form of photos on Instagram. The photos are not ads, but they are shared by friends or people whom we follow on Instagram. They have a great effect on us because they include their emotions as well (Rashidin et al., 2020a). If they are attractive, they will arise our purchasing-desire or sharing-desire. These two desires are diverse, so they should be detached. The new consumption behavior model, 'The Dual AISAS Model' will help elucidate it further. In October 2015, Dentsu brought a new version of AISAS model named as Dual AISAS model, thanks to the cooperation of digital marketing company Atara LLC, in order to advance the digital communication that further maximizes the sales (Dentsu, 2015). Moreover, it further clarifies the information flowing in the model and the contents of the consumer's interest. Additionally, this model puts the interest into two categories (interest in products and interest in photos/images) by recognizing the worth of activation which leads towards purchases in both categories. It makes the AISAS model as 'Purchasing desire', whereas A þ ISAS is 'Sharing desire' or diffusing the information, wherein 'I' stands for 'Interest', 'S' stands for 'Share', 'A' stands for 'Accept' and 'S' stands for 'Spread' as demonstrated in Figure 1 below (i.e. corresponds to the purchasing made through Instagram).
The model begins with an attention phase. It is defined by Venkatraman et al. (2015) as 'a person's 'ability to focus on certain aspects of the environment while ignoring others'. In the era of Web 2.0, the scholars suggested that brands are partnering with digital influencers, wherein they help the brands grab the attention of masses more easily as compared to the celebrities (Ki & Kim, 2019). The attention phase highlights that the influence is instigated from the outwards (influencers), therefore the extent of the influence depends on the user's number of influencers/ digital celebrities following. The vertical flow of Figure 1 demonstrates that user attention is seized though exposure of promotional messages in the form of photos posted by bloggers and vloggers (digital influencers) on Instagram. These influential people attract a lot of attention from internet users and play a key role in word-ofmouth advertising, generating messages and content of use for other people (Meng et al., 2011). For instance, Instagram is one of the digital social media platforms that is being used mostly by opinion leaders to express their opinions about products and services via sharing photos and short videos (Silva et al., 2013). The user rate of influencers following and trust on them (experience) decide the user interest arousal rate. The higher the number of following and trust on influencers, the more the interest will provoke (Javed et al., 2021). The attention and interest that customers have towards the digital influencer's posts create an outcome-searching activities (Hung, 2014;McCartney & Pinto, 2014). According to Korgaonkar et al. (2006), users are more worried about whether only one review is true or manipulated, so while engaged in shopping to experience goods, they will read more reviews in order to verify the quality of the products being considered. This is supported by Park et al. (2009), who found that users tend to search significantly more kinds of information, including both product and customer reviews, when shopping to experience goods than when shopping to search goods. Hence, the use of digital influencers in promotions transforms the user's attention and interest into a high involvement action because influencers have the tendency to deepen the user engagement on shared post (Freberg et al., 2011), and change the attitudes of followers, consequently influencing their intention to buy the recommended products (Renata et al., 2018). These finding are in line with Ki and Kim (2019) study, which revealed that followers who hold positive attitude towards digital influencers are most likely to imitate them, and their mimicry desire drives towards purchase action. Subsequently, the user becomes a transmitter to share his/her experience about product consumption. According to previous studies, the positive attitude towards digital influencer content is linked to content sharing intention at social network (Choi & Lee, 2019). Based on the aforementioned literature, we postulate following hypotheses.
H1: The post of digital influencers has positive impact on user's attention. H2: The post of digital influencers has positive impact on user's interest rate. H3: The post of digital influencers has positive impact on user's intention to search more.
H4: The post of digital influencers has positive impact on user's intention to purchase or to act on it.
H5: The post of digital influencers has positive impact on user's intention to share. On the other hand, last phase 'share' of AISAS begins A þ ISAS. Figure 1 depicts that the influence to inwards begins. The influence on followers depends on number of user's followers. Once the user shares the following post, he/she gets reactions in the form of 'Likes', 'Follows', 'Comments', and 'Share'. The post shared here is related to the product itself, service, and brand experience (Amaly & Hudrasyah, 2012;Wei & Lu, 2013). The higher the number of followers, the more received reactions on certain posts. Therefore, it will activate the user to be attention of others. At this point, online searching behavior is not as consistent as before. This could be due to the fact that in instead of searching for information online, other users could just drop a message to their friends or comment on their friends' post to inquire about the product. This may explain why their attention and interest towards friend's posts would likely lead them to communicate with high involvement (Cheah et al., 2019). For instance, though the user's followers get a lot of information on Instagram, they only share the valuable information without making purchase. Additionally, the user's followers get the original information along with emotion, reviews, and comments, so that information is 'unable to ignore ¼ with a strong attention', or in other words, they 'accept' that information. Cheung et al. (2008Cheung et al. ( , 2009) and Cheung and Thadani (2012) defined the acceptance of information as 'a process in which people intentionally engage in the use of information, i.e. it is an intrinsic aspect of individual, in which the individual purposefully judges whether the information received is reliable and can be used in his/her consumption decision-making'. Moreover, it is widely studied and discussed in academia (Cheung et al., 2009;Cheung & Thadani, 2012;Teng et al., 2014). Cheah et al. (2019) conducted a study, in which authors found that information is widely accepted when it is communicated by close friends than celebrities, and the effect of former is greater than latter, because friend's recommendation is based on multi source information fusion (Cheng et al., 2019). Shintara and Yuji (2018) used 'The Dual AISAS Model' to understand the consumption behavior on SNS and found that consumers tend to trust the information transmitted by other consumers. Moreover, authors found that consumers have ever recommended by sharing and spreading via SNS than before. In addition to this, content spreading plays an important role in motivating followers to express their opinions through the social network, because the goal of sharing information is for others to receive (Lin et al., 2014), and can better leverage the power of social media by paying attention to the content outreach along with private networks (Yu et al., 2015) and extended networks (Lipsman et al., 2012). With a preceding example, this follower becomes the originator and 'spreads' the information within network friends, therefore the information is diffused at an accelerated pace. Particularly, this activity may not affect the purchase of the user, but it affects others', because the spread information can get someone's attention, and in this way, bunch of people make purchases through the platform of Instagram. Therefore, Instagram has an enormous influence to alter the consumer activity/behavior (Sano, 2017). On the basis of above arguments, we develop following hypotheses: H6: The repost of digital influencers has positive impact on user's follower activation.
H7: The repost of digital influencers has positive impact on user's follower attention.
H8: The repost of digital influencers has positive impact on user's follower interest rate.
H9: The repost of digital influencers has positive impact on user's follower intention to share.
H10: The repost of digital influencers by user's follower has positive impact on other user's acceptance.
H11: The repost of digital influencers by user's follower has positive impact on another user's intention to spread.
In a nutshell, red color looped in our research model around horizontal flow is displaying A þ ISAS ¼ Sharing/Dissemination Desire; the interest in product switches from product to communication about goods, services, and spread to outreach (Yajima, 2015). Moreover, it exhibits the flow of diffusion of the brand information, such as advertising and promotion, through the digital network, which assists to determine the influencer content outreach (Chahal, 2016).

Instrument and measures
The scale was adapted from 15-item scale of AISAS (a three-item scale of attention, a three-item scale of interest, a three-item scale of search, a three-item scale of action, and a three-item scale of share). Moreover, a scale of active, accept and spread was developed with the help of existent literature (Erz et al., 2018, Teng et al., 2014. Additionally, for more clarity, the study denotes the two types of interest in our model differently, like the post of digital influencers is labeled as 'interest pd ', and the interest in posts of friends is labeled as 'interest pf '. The scale used for the measurement of constructs was 5-point Likert scale, ranging from 'strongly disagree ¼ 1' to 'strongly agree ¼ 5'. We validated our scales and measurement model using multiple methods. Firstly, following Ruyter and Bloemer (1999), we executed EFA to confirm the underlying factors using varimax rotation and extracted 11 factors that include attention (Att), interest pd (Int), search (Srch), action (Act), Share (Sha), Active (Actv), Attention (Attnt), Interest pf (Inter), Share (Shre), Accept (Accpt), and Spread (Sprd).
Most of the factor loadings (FLs) were above .80 except Actv3 (FL ¼ .62) and Sprd3 (FL ¼ .51), relevant to active and spread, thus excluded for further proceedings. Moreover,eigen values,5.921,3.809,2.792,2.546,2.368,2.262,1.688,1.612,1.456,1.202, and 1.025 of factors cumulatively explained 85.605% of the total variance respectively. Secondly, the face and content validity of the scale was ensured by taking the comments from marketing professors. Furthermore, a pilot study was conducted in order to test the content validity of instrument. The survey recorded 62 data points. Among them, 14 data points were ruled out due to disqualification (i.e. screening questions, incomplete questionnaire (i.e. missing values), or extreme uniform values (i.e. all 1 or all 5). As a result, 48 data points remained valid for analysis, representing a response rate of 70.30%. In the pilot study, female respondents had higher representation (52.1%) than male (47.9%). Majority of the participants (50%) lies under the age category of 24-29, followed by 18-23 (39.6%) and above 29-35 (10.4%). With respect to qualification, majority of the respondents (43.8%) have earned a bachelor's degree, followed by masters (37.5%), higher secondary (16.7%), and high school (2.1%). As far as profession is concerned, majority of participants were students (54.2%), followed by job holders (43.8%), and business (2.1%). Majority of the respondents having Instagram accounts were private (79.2%) than having public accounts (20.8%). Half of the participant accounts were linked with SNS (70.8%) than only linked with Twitter (18.8%) and Facebook (10.4%). In addition, reliability analysis was performed to confirm the robustness of measurement. The Cronbach's alpha (a) values were found to be attention ¼ .

Contextual research setting
Pakistan has been selected as the setting of this study due to the elevated increase in the usage of social media. According to Ali (2020), 37 million people accessed social media regularly from video sharing platforms like TikTok and YouTube to microblogging sites like Facebook, Twitter, and Instagram. According to NepleonCat (2020), in first half of 2020, Instagram had 6,394,000 active users with a net addition of approximately 4,158,000 users on yearly basis. Among all microblogging platforms, Instagram is becoming the eye-catching platform for fashion industry (Yesmail, 2015;Socialbakers, 2016). It has been rapidly expanding and has gained much popularity in Pakistan (Babar, 2020).
Pakistan fashion industry is boosting rapidly and contributes significantly to the economy (Batool, 2018). It turns to digital media for product reviews and endorsement from common people/influencers instead of celebrities whom they can relate to them (Jamil, 2020). In the context of fashion industry, these influential people create content about skincare, clothing, and makeup and post them on their personal accounts. They have the tendency to attract a great number of followers, and enjoy a massive fanbase (Jamal, 2020). Since the concept of influencer marketing is gaining a significant momentum in Pakistan (Bradri.com, 2019), therefore it has been used as contextual setting for our study to substantiate our study hypotheses.

Sampling and data collection procedures
Prior to the execution of data collection, Free Statistical Power Calculator version 4.0 was used to determine the minimum required sample size for structural equation modeling (SEM). The findings revealed that in order to achieve statistical power of 80% with number of observed variables n ¼ 30, number of latent variables n ¼ 11, probability level ¼ .05, and medium effect size ¼ .15, recommended minimum sample size was n ¼ 900. Black et al. (2020) suggested a sample of 100, whereas  suggested a sample of 200 to employ structural equation modeling (SEM) method, so our study sample fulfills this threshold for justifying the use of SEM.
Purposive sampling technique was used to select the respondents. It permits the researchers to choose the valid and informative respondents (Sekaran & Bougie, 2016). We conducted both offline and online surveys to collect the data. For offline survey, five doctoral candidates were trained to distribute the English version questionnaire. Moreover, an online questionnaire was designed on Survey Monkey (i.e. host online questionnaire), and the respondents were contacted by posting a web-link of survey on social media pages of different Pakistani public universities. It must be noted that an online survey helps the researcher reach larger numbers of participants besides reaching more relevant respondents by eliminating irrelevant one through filter/screening questions (Denscombe, 2014). Based on survey research, the instrument commenced with an open-ended statement following the critical incident method (Seckler et al.,2015). The open-ended statement asked the participants to read a statement carefully before filling out the rest part of questionnaire: 'Please take you as a viewer of the posts/reposts of fashion influencers you followed on Instagram. Try to explain your perception.' The underlying reason behind this open-ended question was to get a description about the perception of followers. Purposive sampling has been criticized not being able to generalize research outcomes, although a good deal of online research uses this technique because of subjects' accessibility and proximity to the researchers (Roman, 2007). Moreover, purposive sampling is deemed valid and appropriate when the study is exploratory and when the instrument's items are related to the respondents (Elbeltagi & Agag, 2016). In this regard, our study depends on constructs that have not been widely studied in the past, and the targeted respondents were Instagram users who followed fashion influencers, so the instrument was closely related to the respondents. Thus, the research fulfills both requirements for using purposive sampling.
The data was gathered during a period of 6 weeks, from mid of February till March 2020. Moreover, we ensured with the help of screening questions that 'participants had been the user of Instagram for at least 1 year', 'logged in to Instagram at least once in last 30 days', 'made post on Instagram in last 30 days', and had been the 'follower of at least one fashion digital influencer. Participants who met these selection criteria were allowed to further pursue the survey. The questionnaire commenced with cover letter which narrated the broad purpose of study, and ethical statement to keep their personal information confidential (Rashidin et al., 2020b). The operational definition of digital influencers was clearly stated. It had two sections. First section of survey was particularly related to personal traits of the participants, which contained five close ended questions concerning their gender, age, profession, and educational attainment. Before proceeding to second section, we provided a statement 'Please fill out the rest part of questionnaire (Q1-Q15) keeping in mind the digital fashion influencers (bloggers, vloggers, etc.) you followed on Instagram.' Second section of survey instrument contained 31 close ended questions on five stages of AISAS (to measure purchasing desire) and six stages of A þ ISAS (to measure sharing desire). Upon completion of five stages of AISAS, we again provided a statement to participants 'Please fill out the remaining part of questionnaire (from Q 16-Q31) keeping in mind the repost of fashion influencers make by your friends'. The questionnaire took approximately 10-15 minutes for completion. A total of 525 respondents participated in online survey. In the offline survey, a total of 545 questionnaires were distributed, and 502 questionnaires were returned back, representing a response rate of 92.11%. Fifty-eight questionnaires were discarded due to duplication and incompletion, so 444 questionnaires remained valid for data analysis. Therefore, a total of 969 responses (525 responses from online survey, and 444 responses from offline survey) were valid for data analysis.

Demographic of respondents
The qualified participants who took part in this study were n ¼ 969 as shown in Table 1. The majority of the respondents were females (52.2%) and rest were males (47.8%), which indicates a fair distribution of gender. 59.8% respondents were found to be in 24-29 years of age. Over half of the participants were students (54.9%), followed by job holders (24.8%), and business person (17.2%). Approximately half of the respondents held bachelor's degree (49.6%). The majority of participants had private accounts (n ¼ 862, 89.0%), and rest of participants had public accounts (n ¼ 107, 11.0%). Additionally, the detail of the participants whose Instagram accounts are synchronized with social media is as follows: Facebook (n ¼ 412, 42.5%), Twitter (n ¼ 256, 26.4%), and both Facebook and Twitter (n ¼ 244, 25.2%). Some participants have not connected their accounts with other SNS (n ¼ 57, 5.9%) (see Table 1).

Measurement model
The statistical analysis was carried out by SPSS AMOS graphics version 21. We checked the reliability and the validity of unobserved constructs through measurement model by performing confirmatory factor analysis. The results of confirmatory factor analysis indicate a good fit of the model (v 2 (455.588) ¼ 270, v 2 /df ¼ 1.  Table 2 demonstrates that all constructs are considered reliable by having CR scores above the cut off level .70 (Hair et al., 2011), ranging from 0.837-0.942. Also, a scores surpassed the suggested cut-off level >.70 (Hair et al., 2010), ranging from 0.865-0.943. Convergent validity was tested by employing Fornell and Larcker (1981) approach, which implies that factor loadings of all items should be above .70 and significant. Also, value of average variance extracted from each variable should be greater than .50. Table 2 demonstrates that factor loadings of all construct items are above the threshold level (>.70), with values ranging from 0.778-0.945, and values of AVE are ranging from 0.630-0.844, demonstrating that it is above the recommended cut-off level >.50 (Fornell & Larcker, 1981;Hair et al., 2011), thus the convergent validity is established (Hair et al., 2010). The discriminant validity was ensured by using two tests; first, the square root of average variance extracted (AVE)/diagonal elements should be above the correlations among the variables at the corresponding columns and rows values (Fornell & Larcker, 1981), and second, the correlation among the constructs should not exceed .85 ) (see Tables 2 and 3).

Structural model
The results of the measurement model demonstrate a good model fit.   The fashion related image posted by digital celebrity attracts me. Wei & Lu (2013) and Jun-Hwa Cheah et al. (2019) .909 ÃÃÃ

Att2
The fashion related image posted by digital celebrity draws my full attention.

Att3
The fashion related image posted by digital celebrity catches my eye.
.950 ÃÃÃ Interest, After viewing the fashion related image posted by digital celebrity, I am interested in it.

Int2
After viewing the fashion related image posted by digital celebrity, I like the fashion related products.

Int3
The image posted by digital celebrity gives me a good impression of the fashion items.
.898 ÃÃÃ Search, After viewing the image posted by digital celebrity, I will search for information about the fashion products .873 ÃÃÃ

Srch2
After viewing the image posted by digital celebrity, I will search for online word-of-mouth about the fashion items.

Srch3
After viewing the images posted by digital celebrity, I will compare prices of products within fashion brands.
.897 ÃÃÃ Action, After viewing the images posted by digital celebrity, I believe the offered products of fashion brands is worth trying.

Act2
After viewing the images posted by digital celebrity, I am willing to buy the fashion related items.

Act3
After viewing the image posted by digital celebrity, I think the product will best satisfy my need.
.945 ÃÃÃ After viewing the image posted by digital celebrity, I will forward this advertisement to my friends After viewing the image posted by digital celebrity, I will share the information about the fashion products with my friends .861 ÃÃÃ

Sha3
After purchasing and using fashion products, I will share my experiences and do comments about the fashion products.

Actv2
The customize #hashtags used by my friends on repost of fashion influencers activates me.
.906 ÃÃÃ Attention, The repost of fashion influencers by my friends attracts me.

Attnt2
The repost of fashion influencers by my friends draws my full attention.

Attnt3
The repost of fashion influencers by my friends catches my eye. .

Shre2
I will intend to forward the repost of fashion influencers to my friends.
.890 ÃÃÃ Shre3 I will do share the repost of fashion influencers by adding my comments.

Accpt2
After viewing the repost of fashion influencers, my friends will be greatly influenced.

Sprd2
My friends will expect to diffuse the fashion influencer repost by adding creative #hashtags .  The bold digits in the diagonal are square root of AVE.
These findings indicate an adequate fit to the model. All fit indices were found to be in acceptable range as recommended by several authors (Hooper et al., 2008;MacCallum & Hong, 1997).

Hypotheses testing
We employed structural equation modeling technique (SEM) to test our research model. In SPSS, we initially checked multicollinearity in our dataset. The score of variance inflation factor (VIF) implies that all the constructs had VIF scores within suggested range <3, therefore multicollinearity is not an issue in our dataset (see Table 3). Moving ahead, the results of path coefficients are exhibited in Table 4. The results suggest that on visual centric platform, consumer attention is seized by fashion influencers and has significant impact on consumer interest pd (b ¼ .054 Ã , t ¼ 2.014, p < .05), therefore H1 was supported. Also, consumer interest pd has powerful influence to search more information ( ). The received original information along with friend's emotion gets acceptance from friends in the form of likes, and comments (b ¼ .060 Ã , t ¼ 2.002, p < .05). They spread the information (share) further, and start acting as originator of information for their network friends (b ¼ .396 ÃÃÃ , t ¼ 13.454, p < .001). Therefore, we accepted H5, H6, H7, H8, H9, and H10. Inclusively, all the findings report the first and second inquiry of the present research. Moreover, all the phases involved in AISAS and A þ ISAS have profound effect. Figure 2 also exhibits the predictive power (R 2 ) and the effect size f 2 . We followed Hair (1998) approach for assessing R 2 score, which tells us that the variance explained should be >50% in all outcome variables due to predictor variables. The results of current study show that 0.3% interest pd due to attention, 15.8% search due to interest pd , 5.4% action due to search, 10.4% share due to action, 0.6% activate due to share, 9.1% attention due to activate, 12.8% interest pf due to attention, 1.3% share due to interest pf , 0.4% accept due to share, and 15.7% spread due to accept indicate that R 2 score (71.8%) is above the suggested criteria of 50%. Moreover, by following Cohen (1988) approach, we also assessed the substantive effect size (f 2 ) of research model. The suggested threshold for effect sizes by Cohen (1988) is as follow: small effect is .02, medium effect is .15, and large effect is .35. We found that effect size of activate (f 2 ¼ .006), accept (f 2 ¼ .004), share in A þ AISAS(f 2 ¼ .013), action (f 2 ¼ .057) and interest pd (f 2 ¼ .003) did not lie under the suggested threshold of Cohen's (1988), but gets support by Sawilowsky (2003) threshold having small and very small effect size. However, attention (f 2 ¼ .100), share in AISAS (f 2 ¼ .116), and interest pf (f 2 ¼ .146) have medium effect size (up to the criteria of Cohen's (1988)), and search (f 2 ¼ .187) and spread (f 2 ¼ .186) have large effect size (up to the criteria of Cohen's (1988)).

Discussion
The exponential growth of social media has entirely revamped the way folks interact, communicate, engage, and spread the information (Hansen et al., 2011) due to transmutation of information platform in order to influence them (Hanna et al., 2011). The present research studied the effects of digital influencers on consumers' decisionmaking embracing new behavioral consumption model 'dual AISAS Model' has particular value for domestic and international fashion industry of Pakistan. Pakistan fashion industry has seen briskly mounting since the last decade. Our study findings offer valuable contribution to influencers and fashion brands, because consumers are mounting to seek influencers reviews/information. They base their decisions on influencers reviews/information, because consumers mostly go through influencers reviews/opinions before making their buying decision (Casal o et al., 2020).
Our study revealed that presence of digital influencers is influential because they act as a pull factor to fashion brands in Pakistan. Consumers are more likely to seek the information about latest fashion trends from fashion influencers in order to avoid risk in purchase decision (Flynn et al., 1996). Hence, the influencers seized attention of consumers who are seeking fashion news, reviews, and suggestions. The generated content of fashion influencers has significant influence on consumer attention and interest, and prompts them to search more information about fashion brands either by visiting company insta-page or brand website (McCartney & Pinto, 2014). Additionally, digital influencers have the ability to transform attention, interest and drive into action, and get consumer engagement on their shared post (Freberg et al., 2011).
It is evident from our findings that fashion influencers exert a robust influence on consumers' sharing attitude, which is in line with prior literature (Cheah et al., 2019). Additionally, consumers immediately grab the attention of friends by sharing the post/image either through Instagram stories or they go Live on Instagram with their connections (Lipsman et al., 2012) or with online friends (acquaintances, real world friends) (Chu & Kim, 2011;Jin & Phua, 2014). Evidently, according to our findings, sharing begins the activation of network peers and becomes eye-catching for them, because when they see a close friend using fashion products (by wearing those items), they take interest in them and want to buy them (Frosh, 2015). The matter of the fact is that network peers do not get involved in online searching activities because the needed information is asked from friends by dropping personal message and writing comments on their shared post to inquire it further (Cheah et al., 2019), because information conveyed personally can expedite two-way communication (Herr et al., 1991). Evidently, opinion/reviews of consumers are worthwhile for their peers (Paek et al., 2011) although these reviews can be positive or negative (Kwon et al., 2014). Our findings also suggest that consumers provoke their pals to disseminate/share the image/ post within their private or extended networks (Chatterjee, 2011) that further leads towards communication about fashion brands by activating your friends (Jin & Phua, 2014, Muntinga et al., 2011, seizes their involvement (to grab a reach, intimacy, and engagement) and gets acceptance at wider level (in the form of 'likes', 'comments', and 'share' on same platform or extended networks such as by linking accounts with Facebook, Twitter, Snapchat, and other SNS). Moreover, an individual is more likely to spread the content/image in his small circle of friends or whole network, after getting a sense of identification from cherished comrade (Yu et al., 2015). Our findings are in line with prior notion that information seekers assist the influencers to spread the content from one consumer to another (Goldsmith & Clark, 2008).

Conclusion
A concluding summation is that in contemporary era of digitization, influencers' impact on fashion consumers is an efficient promotional tool to elucidate consumers' decision-making processes. Our findings based on significant association conferred the fact that on visual centric platform/Instagram, digital influencers facilitate the powerful influence on consumers' decision-making processes (Goldsmith & Clark, 2008), particularly in fashion industry (Casal o et al., 2020). Moreover, they offer novel ways to make decisions (Kapitan & Silvera, 2016) and get wider outreach by upturn in consumers intention (awakening their sharing desire) to spread the fashion content within private networks and extended networks (Goldsmith & Clark, 2008). In this regard, consumers (followers) are spreading influencers' promotional content along with their additional knowledge (reviews, comments) to others, thus promoting the influencers, which will eventually escalate their following base (Goldsmith & Clark, 2008).

Theoretical implication
To the best of our knowledge, this study has made remarkable contributions to theory and practice pertaining to consumers' decision-making processes. Firstly, from the theoretical standpoint, the empirical study embraced multi step flow paradigm and theory of buying behavior to support the research model; to examine the effects of fashion influencers on Pakistani consumers decision-making as well as to determine the content outreach of these influencers. Secondly, the literature has examined the effects of digital influencers on consumers' decision-making processes using AISAS model (Cheah et al., 2019;Wei & Lu, 2013), but in the digital world, this study delved digital influencers and explicated their respective effects on consumers' decision-making processes by using an innovative consumer behavior model 'The Dual AISAS Model', which is an updated version of AISAS (Atara & Denstu, 2015). Thirdly, this study advances our knowledge beyond prior research on effects of digital influencers on consumers' decision-making (Sokolova & Kefi, 2020) by focusing the contexts outside the US and European countries. Earlier research indicates that very few studies have paid attention to influencers marketing particularly in South East Asia region. Therefore, we responded to the call of earlier research, and presented/validated a behavioral consumption model in the contextual setting of Pakistan. Specifically, this model enriches our understanding about effects of digital influencers on consumer attention, interest pd , search, action, share, active, attention, interest pf , share, accept, and spread. Additionally, it also gives an insight into the extent of this effect on five phases of AISAS and six phases of A þ ISAS, which was missing in extant marketing literature, and opens up new avenues for future attempts. It does not only imply how digital influencers affect the individual's buying decision, but also gives an understanding about the fashion influencer content outreach through activating consumer sharing desire to diffuse the fashion trends from one user to another user (i.e. within private network) on the same platform or on different SNS by linking up Instagram accounts (i.e. extended networks) with other SNS (Facebook, Twitter, Snapchat etc.). Therefore, our empirical study particularly contributes to the extant literature of consumer decision-making, consumer behavior, and digital influencers.

Managerial implications
The findings of this study are critical for both brands and digital influencers because influencer marketing is growing more dynamic in Pakistan (Nasrullah, 2020) with a growing use of Instagram (Eman, 2019). Particularly, Instagram fashionista capitalized on the growth to expand their reach to the audience (Babar, 2020). Moreover, the rapid expansion of fashionista in Pakistan has altered the behavior of consumers (Babar, 2020). The findings revealed that fashion influencers exert a powerful influence on Pakistani consumers' decision-making. Therefore, we suggest fashion influencers to make their posts more influential; original, unique, and attractive content/ image by adding creative hashtags and emoji, weave a product in Instagram personal stories, and go Live on Instagram. In this way, the content will grab the attention of existing as well as potential customers, and deepen the consumer engagement (Hashoff, 2017) followed by subsequent positive behavior (Herrando et al., 2018). In addition to this, a positive behavior leads to activation/digital dissemination of information to a greater extent on private networks/closely connected network, and extended network along with an evaluative remark. The reactions/evaluative remarks on a shared post will further influence the friends' purchase decision albeit it is amalgam of influences (Katz & Lazarsfeld, 1955). Moreover, we suggest fashion brands to identify the right influencers that best resonate with the targeted audience. Additionally, we recommend the brands, through the lens of this model, to collect the consumer behavioral and purchasing data (e.g. monthly basis and quarterly basis etc.) or to measure the consumer reactions towards Insta fashionista photos. The effective use of this data helps the brands do right investment in promotions and capitalize their growth.

Limitations
Like other empirical studies, this empirical research has also some limitations. Firstly, this study used the online promotional method that's why our model corresponds to the purchasing caused by Instagrammers. Additionally, this study is limited to single platform Instagram. Moreover, it is limited to fashion industry solely therefore, findings cannot be completely construed to other industries or products (electronics, food etc.). Our study is limited to a single country Pakistan and collected the data from Pakistanis to validate our model. Finally, our study is limited to cross-sectional data.

Future research directions
This study is limited in several aspects and gives suggestions to future researchers. Future attempts can be made to test this model on other e-commerce platforms to get better insight into influencer advertising and masses engagement (decision-making) for greater coverage/outreach. The future researchers should test 'The dual AISAS model' in other industries like beauty, food, travel, health, and fitness to increase the robustness of this model. Thirdly, due to the cultural differences and consumer behavior, we recommend future researcher to cross validate the findings of current study by employing this model in other emerging countries as well as developed countries. Future attempt can use longitudinal data to measure eleven phases of dual AISAS model at different point in time or can use experimental design to see how consumers respond to stimuli in diverse scenarios and its effects on consumers' decision-making processes.
These effects are much warranted by taking into consideration customer psychographic (generation, lifestyle, and values) and demographic aspects.

Disclosure statement
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.