The initial post should be 200-500 words and is expected to be substantive, scholarly, and original0.

Make sure you use proper grammar, word choice, syntax (arrangement of words to create well-formed sentences), and writing mechanics (capitalization, punctuation, and spelling). Please refer to and cite relevant work.

Questions

The assigned reading, titled "Effects of paid search advertising on product sales: a Chinese semantic perspective," by Yang et al. (2020) talks about the brand and non-brand keywords hypothesis. It proposes (a) Brand keywords generate higher product sales than non-brand keywords when the brand market share is large. However, (b) this effect is eliminated when the brand market share is small. Briefly explain with examples what are brand and non-brand keywords. Do you agree with this proposal? Why? Why not?

Discuss some of the privacy concerns of users in the world of programmatic advertising. Refer to one incident about an organization/brand invading customers privacy that you are aware of. Explain how they could have done a better job to avoid privacy concerns.

Effects of paid search advertising on product sales: a Chinese semantic perspective Zhi Yang a, Yueyan Wua, Chongyu Lub and Y angjun Tua 'Business School, Hunan University, Changsha, China; bLubin School of Business, Pace University, New York, NY, USA ABSTRACT Prior research on the impact of brand keywords on product sales has produced contradictory findings. Thus, one purpose of this study was to examine how brand keywords affect product sales when brand equity is considered. The other purpose was to explore how hedonic and utilitarian keywords interact with product type to impact product sales. The results of analyses of two secondary datasets and one lab experiment showed that brand keywords yielded more product sales than non-brand keywords. However, this effect disappeared when brand market share was small or consumer brand knowledge was high. A coding system was devel- oped for Chinese keywords based on Chinese semantic features. Results showed a matching effect in which hedonic keywords generated higher product sales than utilitarian keywords for hedo- nic products, and utilitarian keywords generated higher product sales than hedonic keywords for utilitarian products. Introduction Paid search advertising now accounts for more than 50% of e-retailers' ad spending (Dai \& Luca, 2017). In paid search advertising, keywords serve as an essential bridge linking e-retailers and search users (Yang et al., 2016). Thus, many studies have explored the effects of keyword features on impressions, click-through rates, conversions and product sales. A large number of these studies have focused on the external features of keywords. These extrinsic features include keyword frequency, length, cost, rank, popularity and specificity (e.g. Agarwal et al., 2011; Jerath et al., 2011; Wang et al., 2019; Yang et al., 2016). Other studies have examined the specific information communicated by keywords, such as brand name, retailer name and location, which are referred to as the intrinsic features of keywords (e.g. Ghose \& Yang, 2009; Jansen et al., 2011; Kim et al., 2012). Despite the extensive literature on paid search advertising (see Table 1), there are two matters that should be further addressed. The first is the influence of intrinsic brand information (i.e. brand keywords) on keyword performance. Brand information is so important to keyword advertising that almost all prior research on the intrinsic features of keywords has discussed its impact. Nonetheless, the findings from these studies have CONTACT Yangjun Tu tuyangjun@163.com E Business School, Hunan University, Changsha 410082, People's Republic of China been contradictory. For example, several studies have shown that brand keywords are positively associated with impressions (Kim et al., 2012), click-through rates (e.g. Du et al., 2017; Rutz et al., 2012), conversion rates (e.g. Klapdor et al., 2014; Simonov et al., 2018), return visitations (Rutz et al., 2011) and orders (Spilker-Attig \& Brettel, 2010). Wolk and Theysohn (2007) also found a positive link between brand information in paid content and the number of visitors to a website. Other studies have shown that brand keywords have a negative impact on search volume (Yang \& Ghose, 2010), click-through rates (Ghose \& Yang, 2009; Im et al., 2016) and conversion rates (Ghose \& Yang, 2009). Further, only a few studies have linked brand keywords with product sales, the critical indicator of e-retailers' survival (Jansen et al., 2011; Lu \& Zhao, 2014). Therefore, a more comprehensive study is needed to explore the impacts of brand keywords on e-retailers' product sales. The second matter to be addressed is how to extract and mine keywords' other effective intrinsic features and their impacts on product sales. Intrinsic features communicate a product's key benefits and value to consumers. They are the essential bridge that links consumers' needs and products (Gopal et al., 2011). Therefore, choosing keywords that accurately represent a product's benefits and value should be an effective way to gain sales. Nonetheless, beyond brand and retailer names and location information (Klapdor et al., 2014; Yang \& Ghose, 2010), prior work has rarely addressed how the intrinsic features that communicate a product's benefits and value influence product sales. Such benefits and value vary greatly from product to product. For example, hedonic products tend to provide experience value, whereas utilitarian products are more likely to provide functional value. Therefore, there is a need to address how such intrinsic features of keywords interact with product types to affect product sales. Based on the foregoing discussion, the first aim of this study is to comprehensively examine how brand keywords affect product sales. Brand market share (i.e. large and small) and consumer brand knowledge (i.e. high and low) are considered. The former is an objective measurement and the latter is a subjective self-reported measurement based on consumer-based brand equity. We propose that brand keywords generate higher product sales than non-brand keywords. However, this effect is eliminated when the brand market share is small (versus large) or consumer brand knowledge is high (versus low). The second aim is to explore how keywords that communicate a product's hedonic or utilitarian benefits and value affect sales. We call these keywords attribute keywords and classify them as hedonic keywords and utilitarian keywords. We suggest that there is a matching effect between attribute keywords and product types. In detail, we propose that hedonic keywords generate higher product sales than utilitarian keywords for hedonic products, whereas utilitarian keywords generate higher product sales than hedonic keywords for utilitarian products. Figure 1 outlines the research framework. Analyses of two secondary datasets and one lab experiment support our proposals. Our findings make several contributions. First, by examining brand market share and consumer brand knowledge, we show how brand keywords impact sales in a comprehensive way and explain why previous findings on the impact of brand keywords on sales have been contradictory. Second, previous studies have investigated the intrinsic features of kevwords including hrand name retailer name and location informa- suggest that hedonic keywords generate higher product sales than utilitarian keywords for hedonic products, whereas utilitarian keywords generate higher product sales than hedonic keywords for utilitarian products. Third, we test the matching effect of attribute keywords and product type and examine its effects on product sales. We demonstrate that hedonic (utilitarian) keywords matched with hedonic (utilitarian) products increase e-retailers' product sales. We also offer some insights into and guidelines for this effort. Finally, by obtaining two secondary datasets from two Chinese e-sellers, we extend theory by developing a coding system for Chinese keywords based on Chinese semantics. Through this system, the product attribute information contained in keywords can be identified. Although some prior work (Klapdor et al., 2014; Rutz et al., 2011) has explored keyword information content from a semantic perspective, the information under study has been limited to brand name, retailer name and location, and the paradigm has only applied to English keywords. We code and analyse products' hedonic and utilitarian information from the perspective of Chinese semantics. We hope that our work provides an example of semantic analysis of Chinese keywords because China's e-commerce market is the largest in the world. Its volume is currently 1 USD.5 trillion and will exceed 1 USD. 8 trillion by 2022 (Forrester, 2018). To the best of our knowledge, research on paid search keywords in the Chinese market has been limited (Huang et al., 2016; Lu \& Zhao, 2014; Wang et al., 2019). We believe that in the future, more comprehensive and detailed research on keywords in the Chinese e-commerce market will be needed. In addition, our work provides practical insights into optimising e-retailers' bidding strategy for paid search keywords at auctions in terms of their brand market share, consumer brand knowledge and product type. In the following sections, we first review previous work on paid search advertising and develop our hypotheses. We then report on our two secondary data analyses and one lab experiment, which test our hypotheses. Finally, we discuss the theoretical contribution, managerial implications, limitations and future directions of our work. JOURNAL OF MARKETING MANAGEMENT 7 Literature review and hypotheses development Paid search advertising Keywords serve as an essential bridge between e-retailers and search users in paid search advertising (Yang et al., 2016). There are two perspectives of paid search advertising on e-commerce platforms: the e-retailer bidding perspective and the consumer journey perspective (see Figure 2). For consumers, their goal is to buy products that meet their needs. Thus, they first search keywords that meet their purchasing intent (i.e. search), then click on a product ad that directs them to the landing page of the focal product (i.e. click). They make a purchasing decision after browsing the details of the product (i.e. purchase), and finally they may engage in some post-purchase activities (e.g. rating). For e-retailers, their goal is to present and sell their products to consumers. Thus, their first task is to select and bid on the keywords (i.e. select \& bid) that best describe their products' characteristics (e.g. product name, function, brand). Thereafter, the e-commerce platform exposes the sponsored ads to consumers based on the outcome of the e-retailers' auction (i.e. impress). If the advertised product is sold, e-retailers gain sales (i.e. sales). The final stage for e-retailers is to provide after-sales service. This study explores how keyword selection impacts product sales from the e-retailer bidding perspective. Brandon-brand keywords and product sales Brand keyword and non-brand keyword defined Each keyword may consist of one or more words that reflect a product's characteristics (Ghose \& Yang, 2009). The characteristics can be brand, shape, colour and so on. It is widely accepted that brand represents a product's fundamental information (Rahman et al., 2008; Wootten, 2003). Brand information is also a crucial attribute contained in keywords (Kim et al., 2012). As shown in Table 2, previous studies have defined brand keywords in a variety of ways. Based on prior work (Jerath et al., 2014; Kim et al., 2012; Rutz et al., 2012; Yang \& Ghose, 2010), this study defines brand keyword as a keyword that contains a brand name, whereas a non-brand keyword is a keyword that does not contain a brand name. For example, in the case of shoes, 'Yijiabao comfortable shoes' is a brand keyword in which 'Yijiabao' is a Chinese brand of shoes. 'Fashion design shoes' is a non-brand keyword because there is no brand information. Likewise, in the case of water purifiers, 'Qinyuan water purifier' is a brand keyword in which 'Qinyuan' is a Chinese brand of water purifier, whereas Figure 2. E-retailer bidding perspective and consumer journey perspective in paid search advertising. Takla Drand tarm and dafinitione in tha litaration 'small white water purifier' is a non-brand keyword because there is no brand information. Brandon-brand keywords and product sales Brand information plays an important role in the performance of paid search advertising. First, compared to non-brand keywords, brand keywords contain brand names. These provide clues into a product's quality. With the signal of quality, consumers are more likely to trust the search results of brand keywords when they make purchasing decisions (Rahman et al., 2008; Wootten, 2003). According to Klapdor et al. (2014), brand keywords increase the keyword click-through rate and conversation rate, both of which can positively affect sales (Kim et al., 2012; Rutz \& Bucklin, 2007). Second, brand keywords facilitate sales by arousing brand awareness and attitudes. In paid search advertising, consumers are aware of the searched-for brands and intend to purchase products from specific brands when they use brand keywords (versus non-brand keywords; Drze \& Hussherr, 2003; Fang et al., 2015; Gallagher et al., 2001; Ghose \& Yang, 2009; Rutz \& Bucklin, 2011). Brand awareness increases the level of subsequent visitations, which may positively affect sales (Rutz et al., 2011). Searching brand keywords also indicates that consumers are in a later stage of the purchasing process (Jansen \& Schuster, 2011), when they are more likely to make purchasing decisions. Brand keywords can affect sales as much as 15 times more than non-brand keywords (Jansen et al., 2011). Empirical findings have suggested that brand keywords significantly outperform non-brand keywords, Therefore, this study makes the following hypothesis: H1: Brand keywords generate higher product sales than non-brand keywords. The moderating role of brand market share Brand market share adds values to a product and affects consumers' responses to the product (Goodhardt et al., 1984; Romaniuk et al., 2007; Sharp et al., 2012). For brand awareness and association, a large share brand is always easier to access and gain more responses from than a small share brand (Romaniuk, 2006). Research has shown that a product's brand equity positively affects both consumers' willingness to pay premium prices (Keller, 1993) and the product's profits (Srivastava \& Shocker, 1991). In paid search advertising, because brand keywords signal product quality, they stimulate brand awareness and association. Thus, a keyword with a brand name is more likely than a non-brand keyword to induce a consumer response. However, Romaniuk (2006) found that for smaller share brands, an unprompted approach is less likely to elicit associations. Shopping online is a situation in which consumers spontaneously search keywords for their intended products. Consumers are less likely to associate products with small share brands, which in turn decreases their likelihood of searching for or buying them through brand keywords in e-commerce. Thus, we propose: H2: (a) Brand keywords generate higher product sales than non-brand keywords when the brand market share is large. However, (b) this effect is eliminated when the brand market share is small. The moderating role of consumer brand knowledge All brand-image associations are related to consumers' prior experience and knowledge. Consumer brand knowledge increases through buying, consuming, viewing the brand's advertising or through word of mouth (Romaniuk, 2006). Romaniuk et al. (2012) explained that brand knowledge is a key driver of brand-image associations (see also Bird et al., 1970; Romaniuk, 2006; Romaniuk \& Nenycz-Thiel, 2013). For example, a former user of a brand is more likely to make a brand association than someone who has never tried the brand (Romaniuk et al., 2012). Ku et al. (2019) also observed that brand familiarity increases recall and the association with that brand. A branded product often provides unique benefits and value to consumers. Thus, consumers with high brand knowledge are more likely to associate a brand with its products' unique attributes. Similarly, by assigning unique attributes to products, consumers associate with specific brands. When shopping on an e-commerce platform, consumers with high brand knowledge can search either by a specific brand name or a unique attribute of a brand. When the unique attribute matches the product, they are more likely to purchase it because the brand offers consumers a compelling reason to do so (Aaker \& Shansby, 1982; Keller, 1993; Ku et al., 2019). That is, for consumers with high brand knowledge, their intention to purchase a particular product will increase, compared to non-brand keywords. However, consumers with low brand knowledge tend to evaluate a product based on external clues, such as the brand (Dou et al., 2010; Narayanan \& Kalyanam, 2015). Therefore, they tend to search by brand name instead of by the unique attributes of a brand. In addition, compared with brand keywords, products associated with non-brand keywords may increase consumers' confidence in their decisions if they have low brand knowledge. Hence, we propose: H3: (a) Brand keywords generate higher product sales than non-brand keywords when consumer brand knowledge is low. However, (b) this effect is eliminated when the consumer brand knowledge is high. Hedonic/utilitarian keywords and product sales Hedonic keyword and utilitarian keyword defined Similar to 'brand keyword', we first define 'attribute keyword' as a keyword that contains a product's attribute information. Such attributes can be classified as hedonic or utilitarian based on the benefits and value the product provides (Batra \& Ahtola, 1991; Chitturi et al., 2007; 2008; Strahilevitz \& Myers, 1998). Utilitarian attributes have utilitarian benefits and value, whereas hedonic attributes have hedonic benefits and value (see Dhar \& Wertenbroch, 2013; Jones et al., 2006). For example, for products like shoes, 'round toe' (shape), 'retro design' (fashion trend) and 'net surface' (style) are hedonic attributes, whereas 'ventilated', 'keep warm' and 'antiskid' (related to function) are utilitarian attributes. For products like water purifiers, 'white' (colour), 'mini' (size) and 'wall hanging' (style) are hedonic attributes, whereas 'straight drink', 'reverse osmosis' and 'ultrafiltration' (related to function) are utilitarian attributes. Based on these two well-documented types of product attributes (hedonic and utilitarian), we classified attribute keywords as hedonic keywords and utilitarian keywords. A hedonic keyword mainly describes the aesthetic, experiential and enjoyment-related attributes of a product, whereas a utilitarian keyword mainly describes the functional, instrumental and practical attributes of a product. Keywords such as 'fashion round toe shoes' and 'retro design shoes' are hedonic keywords, whereas 'comfortable toe protection shoes' and 'antiskid shoes' are utilitarian keywords. Hedonic/utilitarian keywords and product sales Products are designed and produced to satisfy various consumer demands (Chitturi et al., 2007; 2008). Products can be classified as hedonic or utilitarian (Bridges \& Florsheim, 2008; Jones et al., 2006; Kempf, 1999; Woods, 1960). Previous studies have indicated that consumers expect different benefits from different types of products. They tend to seek function-related benefits from utilitarian products and experience-related benefits from hedonic products (Dhar \& Wertenbroch, 2013; Jones et al., 2006). When searching online, consumers often type in relevant keywords to describe what they want from a product (Chernev, 2006; Klein \& Melnyk, 2016; van Osselaer \& Janiszewski, 2012). That is, they instinctively input hedonic (utilitarian) keywords when seeking experience (function) related benefits from hedonic (utilitarian) products. Based on regulatory theory, Higgins (2000) pointed out that 'people experience a regulatory fit when they use goal pursuit means that fit their regulatory orientation, and this regulatory fit increases the value of what they are doing' (p. 1217). Regulatory fit makes individuals feel 'right' (Hamstra et al., 2013; Higgins, 2004), thereby enhancing their certainty about their initial goals and increasing their decision-making confidence (Hamstra et al., 2013; Higgins, 2000; Zheng et al., 2015). If the search results for hedonic keywords match the experience benefits that consumers seek, they can be confident in making decisions about hedonic products. Similarly, when consumers have expectations of the functional benefits of a utilitarian product, they focus on utilitarian attributes and input utilitarian keywords. Based on the regulatory fit (Hamstra et al., 2013; Higgins, 2000; 2004; Zheng et al., 2015), once the search results from utilitarian keywords fulfil consumers' function-related needs, the consumers can make their purchasing decisions with confidence. In paid search advertising, it can be assumed that searching keywords that their goals. Hence, we propose: H4: (a) For hedonic products, hedonic keywords generate higher product sales than utilitarian keywords, whereas (b) for utilitarian products, utilitarian keywords generate higher product sales than hedonic keywords. Overview of the studies Two secondary datasets were analysed and one lab experiment was conducted to test our hypotheses. First, we obtained two datasets from two online sellers on Taobao.com, China's largest consumer-to-consumer (C2 C) e-commerce platform. This platform offers sellers a chance to sell their products (including shoes, clothes and electronics) to individual consumers. It also offers keyword auction services to sellers. Sellers in this platform can create and bid for keywords related to their products. Based on a pre-test (see the Results section of Studies 1 and 2 below), we finally chose men's leisure shoes as the hedonic product and water purifiers as the utilitarian product for this study. Both sellers marketed products from multiple brands. The results of the analyses of the two secondary datasets supported H1,H2a,H2b,H4a and H4b. Second, because consumer brand knowledge is hard to measure using secondary data, we conducted a lab experiment using a real mobile phone brand in China to test H3a and H3b. Studies 1 and 2: secondary data analysis Within a three-month window, we downloaded two secondary datasets from two sellers on China Taobao.com. The first dataset was downloaded with the cooperation of a men's shoe seller 1 and, the second dataset was downloaded from a water purifier seller. The men's shoe seller sold several bands of men's leisure shoes, and the water purifier seller sold multiple brands of water purifiers. Overall, 10,966 records of shoe data and 53,701 records of water purifier data were obtained. Data coding Based on the definitions of 'hedonic keyword' and 'utilitarian keyword', four researchers coded the keywords according to the semantic features of Chinese. First, each Chinese semantic group consists of two or more single Chinese characters composed of two or more radicals with a particular semantic feature (Taft et al., 1999). For instance, '' (at leisure, free and having spare time) consists of the two characters '' and '', each of which comprises two semantic radicals. The character '' contains two radicals; the first, ' 1 ', means a man and the second, '', means wood. Therefore, '' refers to a man leaning against wood, feeling pleasant and comfortable. The other character, ' '', contains two radicals; the first, '', meaning a door and the second, '', meaning wood. Therefore, '' refers to a wooden door and to closing the door to sleep. Thus, the semantic unit and clues are the key factors in recognising Chinese words regardless of their radicals and characters. During the coding procedure, the coders needed to bear in mind that each result had to be a complete semantic unit; for example, '' (shoes) and ' ' (shoes + a nonsense syllable) had to be coded as the same unit because each one was a complete semantic unit. Second, hedonic products are different from utilitarian products in their sensory and functional characteristics (Woods, 1960). For instance, hedonic products are dependent on their sensory characteristics (and the visual features of any product, such as colour and design). To a large extent, shoes are appealing to consumers because of their sensory features, such as design, colour and type, whereas their appeal depends to a lesser extent on functional features such as their ability to protect the feet from injury. Conversely, water purifiers mainly attract consumers through their special or powerful functions, not through their design, colour or type. During the coding procedure, the coders referred to the category of hedonic/functional character as a basic coding framework. Third, the meaning of a lexical term can be distinguished according to the attributes of its semantic features, which may be defining or characteristic (Smith et al., 1974). In the keywords '' (a kind of leisure shoe), for example, the defining character '' (shoes) is an essential or defining aspect of the Chinese semantic group, and '' (leisure) indicates a non-essential or characteristic feature of the group. In the coding procedure, the coders considered the difference between defining and characteristic features. Fourth, after each Chinese semantic group was coded (see Table 3), the coders calculated and compared the number of hedonic and utilitarian groups for each keyword, then classified them as hedonic or utilitarian keywords. If the hedonic value outnumbered the utilitarian value, the keyword was classified as hedonic. Likewise, if the utilitarian value outnumbered the hedonic value, the keyword was classified as utilitarian. The keywords were classified as neutral if the numbers for each value were equal. This classification method was adopted from previous studies by Goh et al. (2013), Healey and Kassarjian (1983), and You et al. (2017) in other disciplines. Table 3. Coding results of Chinese semantic group in keywords for the men's shoes and water purifier data. N/A not applicable. Due to the confidentiality agreement, both sellers do not wish to disclose this information. We told the coders the name of the brand with the largest market share. The brand with the largest market share should be coded as 1 and the others as 0. Effects of paid search advertising on product sales: a Chinese semantic perspective Zhi Yang a, Yueyan Wua, Chongyu Lub and Y angjun Tua 'Business School, Hunan University, Changsha, China; bLubin School of Business, Pace University, New York, NY, USA ABSTRACT Prior research on the impact of brand keywords on product sales has produced contradictory findings. Thus, one purpose of this study was to examine how brand keywords affect product sales when brand equity is considered. The other purpose was to explore how hedonic and utilitarian keywords interact with product type to impact product sales. The results of analyses of two secondary datasets and one lab experiment showed that brand keywords yielded more product sales than non-brand keywords. However, this effect disappeared when brand market share was small or consumer brand knowledge was high. A coding system was devel- oped for Chinese keywords based on Chinese semantic features. Results showed a matching effect in which hedonic keywords generated higher product sales than utilitarian keywords for hedo- nic products, and utilitarian keywords generated higher product sales than hedonic keywords for utilitarian products. Introduction Paid search advertising now accounts for more than 50% of e-retailers' ad spending (Dai \& Luca, 2017). In paid search advertising, keywords serve as an essential bridge linking e-retailers and search users (Yang et al., 2016). Thus, many studies have explored the effects of keyword features on impressions, click-through rates, conversions and product sales. A large number of these studies have focused on the external features of keywords. These extrinsic features include keyword frequency, length, cost, rank, popularity and specificity (e.g. Agarwal et al., 2011; Jerath et al., 2011; Wang et al., 2019; Yang et al., 2016). Other studies have examined the specific information communicated by keywords, such as brand name, retailer name and location, which are referred to as the intrinsic features of keywords (e.g. Ghose \& Yang, 2009; Jansen et al., 2011; Kim et al., 2012). Despite the extensive literature on paid search advertising (see Table 1), there are two matters that should be further addressed. The first is the influence of intrinsic brand information (i.e. brand keywords) on keyword performance. Brand information is so important to keyword advertising that almost all prior research on the intrinsic features of keywords has discussed its impact. Nonetheless, the findings from these studies have CONTACT Yangjun Tu tuyangjun@163.com E Business School, Hunan University, Changsha 410082, People's Republic of China been contradictory. For example, several studies have shown that brand keywords are positively associated with impressions (Kim et al., 2012), click-through rates (e.g. Du et al., 2017; Rutz et al., 2012), conversion rates (e.g. Klapdor et al., 2014; Simonov et al., 2018), return visitations (Rutz et al., 2011) and orders (Spilker-Attig \& Brettel, 2010). Wolk and Theysohn (2007) also found a positive link between brand information in paid content and the number of visitors to a website. Other studies have shown that brand keywords have a negative impact on search volume (Yang \& Ghose, 2010), click-through rates (Ghose \& Yang, 2009; Im et al., 2016) and conversion rates (Ghose \& Yang, 2009). Further, only a few studies have linked brand keywords with product sales, the critical indicator of e-retailers' survival (Jansen et al., 2011; Lu \& Zhao, 2014). Therefore, a more comprehensive study is needed to explore the impacts of brand keywords on e-retailers' product sales. The second matter to be addressed is how to extract and mine keywords' other effective intrinsic features and their impacts on product sales. Intrinsic features communicate a product's key benefits and value to consumers. They are the essential bridge that links consumers' needs and products (Gopal et al., 2011). Therefore, choosing keywords that accurately represent a product's benefits and value should be an effective way to gain sales. Nonetheless, beyond brand and retailer names and location information (Klapdor et al., 2014; Yang \& Ghose, 2010), prior work has rarely addressed how the intrinsic features that communicate a product's benefits and value influence product sales. Such benefits and value vary greatly from product to product. For example, hedonic products tend to provide experience value, whereas utilitarian products are more likely to provide functional value. Therefore, there is a need to address how such intrinsic features of keywords interact with product types to affect product sales. Based on the foregoing discussion, the first aim of this study is to comprehensively examine how brand keywords affect product sales. Brand market share (i.e. large and small) and consumer brand knowledge (i.e. high and low) are considered. The former is an objective measurement and the latter is a subjective self-reported measurement based on consumer-based brand equity. We propose that brand keywords generate higher product sales than non-brand keywords. However, this effect is eliminated when the brand market share is small (versus large) or consumer brand knowledge is high (versus low). The second aim is to explore how keywords that communicate a product's hedonic or utilitarian benefits and value affect sales. We call these keywords attribute keywords and classify them as hedonic keywords and utilitarian keywords. We suggest that there is a matching effect between attribute keywords and product types. In detail, we propose that hedonic keywords generate higher product sales than utilitarian keywords for hedonic products, whereas utilitarian keywords generate higher product sales than hedonic keywords for utilitarian products. Figure 1 outlines the research framework. Analyses of two secondary datasets and one lab experiment support our proposals. Our findings make several contributions. First, by examining brand market share and consumer brand knowledge, we show how brand keywords impact sales in a comprehensive way and explain why previous findings on the impact of brand keywords on sales have been contradictory. Second, previous studies have investigated the intrinsic features of kevwords including hrand name retailer name and location informa- suggest that hedonic keywords generate higher product sales than utilitarian keywords for hedonic products, whereas utilitarian keywords generate higher product sales than hedonic keywords for utilitarian products. Third, we test the matching effect of attribute keywords and product type and examine its effects on product sales. We demonstrate that hedonic (utilitarian) keywords matched with hedonic (utilitarian) products increase e-retailers' product sales. We also offer some insights into and guidelines for this effort. Finally, by obtaining two secondary datasets from two Chinese e-sellers, we extend theory by developing a coding system for Chinese keywords based on Chinese semantics. Through this system, the product attribute information contained in keywords can be identified. Although some prior work (Klapdor et al., 2014; Rutz et al., 2011) has explored keyword information content from a semantic perspective, the information under study has been limited to brand name, retailer name and location, and the paradigm has only applied to English keywords. We code and analyse products' hedonic and utilitarian information from the perspective of Chinese semantics. We hope that our work provides an example of semantic analysis of Chinese keywords because China's e-commerce market is the largest in the world. Its volume is currently 1 USD.5 trillion and will exceed 1 USD. 8 trillion by 2022 (Forrester, 2018). To the best of our knowledge, research on paid search keywords in the Chinese market has been limited (Huang et al., 2016; Lu \& Zhao, 2014; Wang et al., 2019). We believe that in the future, more comprehensive and detailed research on keywords in the Chinese e-commerce market will be needed. In addition, our work provides practical insights into optimising e-retailers' bidding strategy for paid search keywords at auctions in terms of their brand market share, consumer brand knowledge and product type. In the following sections, we first review previous work on paid search advertising and develop our hypotheses. We then report on our two secondary data analyses and one lab experiment, which test our hypotheses. Finally, we discuss the theoretical contribution, managerial implications, limitations and future directions of our work. JOURNAL OF MARKETING MANAGEMENT 7 Literature review and hypotheses development Paid search advertising Keywords serve as an essential bridge between e-retailers and search users in paid search advertising (Yang et al., 2016). There are two perspectives of paid search advertising on e-commerce platforms: the e-retailer bidding perspective and the consumer journey perspective (see Figure 2). For consumers, their goal is to buy products that meet their needs. Thus, they first search keywords that meet their purchasing intent (i.e. search), then click on a product ad that directs them to the landing page of the focal product (i.e. click). They make a purchasing decision after browsing the details of the product (i.e. purchase), and finally they may engage in some post-purchase activities (e.g. rating). For e-retailers, their goal is to present and sell their products to consumers. Thus, their first task is to select and bid on the keywords (i.e. select \& bid) that best describe their products' characteristics (e.g. product name, function, brand). Thereafter, the e-commerce platform exposes the sponsored ads to consumers based on the outcome of the e-retailers' auction (i.e. impress). If the advertised product is sold, e-retailers gain sales (i.e. sales). The final stage for e-retailers is to provide after-sales service. This study explores how keyword selection impacts product sales from the e-retailer bidding perspective. Brandon-brand keywords and product sales Brand keyword and non-brand keyword defined Each keyword may consist of one or more words that reflect a product's characteristics (Ghose \& Yang, 2009). The characteristics can be brand, shape, colour and so on. It is widely accepted that brand represents a product's fundamental information (Rahman et al., 2008; Wootten, 2003). Brand information is also a crucial attribute contained in keywords (Kim et al., 2012). As shown in Table 2, previous studies have defined brand keywords in a variety of ways. Based on prior work (Jerath et al., 2014; Kim et al., 2012; Rutz et al., 2012; Yang \& Ghose, 2010), this study defines brand keyword as a keyword that contains a brand name, whereas a non-brand keyword is a keyword that does not contain a brand name. For example, in the case of shoes, 'Yijiabao comfortable shoes' is a brand keyword in which 'Yijiabao' is a Chinese brand of shoes. 'Fashion design shoes' is a non-brand keyword because there is no brand information. Likewise, in the case of water purifiers, 'Qinyuan water purifier' is a brand keyword in which 'Qinyuan' is a Chinese brand of water purifier, whereas Figure 2. E-retailer bidding perspective and consumer journey perspective in paid search advertising. Takla Drand tarm and dafinitione in tha litaration 'small white water purifier' is a non-brand keyword because there is no brand information. Brandon-brand keywords and product sales Brand information plays an important role in the performance of paid search advertising. First, compared to non-brand keywords, brand keywords contain brand names. These provide clues into a product's quality. With the signal of quality, consumers are more likely to trust the search results of brand keywords when they make purchasing decisions (Rahman et al., 2008; Wootten, 2003). According to Klapdor et al. (2014), brand keywords increase the keyword click-through rate and conversation rate, both of which can positively affect sales (Kim et al., 2012; Rutz \& Bucklin, 2007). Second, brand keywords facilitate sales by arousing brand awareness and attitudes. In paid search advertising, consumers are aware of the searched-for brands and intend to purchase products from specific brands when they use brand keywords (versus non-brand keywords; Drze \& Hussherr, 2003; Fang et al., 2015; Gallagher et al., 2001; Ghose \& Yang, 2009; Rutz \& Bucklin, 2011). Brand awareness increases the level of subsequent visitations, which may positively affect sales (Rutz et al., 2011). Searching brand keywords also indicates that consumers are in a later stage of the purchasing process (Jansen \& Schuster, 2011), when they are more likely to make purchasing decisions. Brand keywords can affect sales as much as 15 times more than non-brand keywords (Jansen et al., 2011). Empirical findings have suggested that brand keywords significantly outperform non-brand keywords, Therefore, this study makes the following hypothesis: H1: Brand keywords generate higher product sales than non-brand keywords. The moderating role of brand market share Brand market share adds values to a product and affects consumers' responses to the product (Goodhardt et al., 1984; Romaniuk et al., 2007; Sharp et al., 2012). For brand awareness and association, a large share brand is always easier to access and gain more responses from than a small share brand (Romaniuk, 2006). Research has shown that a product's brand equity positively affects both consumers' willingness to pay premium prices (Keller, 1993) and the product's profits (Srivastava \& Shocker, 1991). In paid search advertising, because brand keywords signal product quality, they stimulate brand awareness and association. Thus, a keyword with a brand name is more likely than a non-brand keyword to induce a consumer response. However, Romaniuk (2006) found that for smaller share brands, an unprompted approach is less likely to elicit associations. Shopping online is a situation in which consumers spontaneously search keywords for their intended products. Consumers are less likely to associate products with small share brands, which in turn decreases their likelihood of searching for or buying them through brand keywords in e-commerce. Thus, we propose: H2: (a) Brand keywords generate higher product sales than non-brand keywords when the brand market share is large. However, (b) this effect is eliminated when the brand market share is small. The moderating role of consumer brand knowledge All brand-image associations are related to consumers' prior experience and knowledge. Consumer brand knowledge increases through buying, consuming, viewing the brand's advertising or through word of mouth (Romaniuk, 2006). Romaniuk et al. (2012) explained that brand knowledge is a key driver of brand-image associations (see also Bird et al., 1970; Romaniuk, 2006; Romaniuk \& Nenycz-Thiel, 2013). For example, a former user of a brand is more likely to make a brand association than someone who has never tried the brand (Romaniuk et al., 2012). Ku et al. (2019) also observed that brand familiarity increases recall and the association with that brand. A branded product often provides unique benefits and value to consumers. Thus, consumers with high brand knowledge are more likely to associate a brand with its products' unique attributes. Similarly, by assigning unique attributes to products, consumers associate with specific brands. When shopping on an e-commerce platform, consumers with high brand knowledge can search either by a specific brand name or a unique attribute of a brand. When the unique attribute matches the product, they are more likely to purchase it because the brand offers consumers a compelling reason to do so (Aaker \& Shansby, 1982; Keller, 1993; Ku et al., 2019). That is, for consumers with high brand knowledge, their intention to purchase a particular product will increase, compared to non-brand keywords. However, consumers with low brand knowledge tend to evaluate a product based on external clues, such as the brand (Dou et al., 2010; Narayanan \& Kalyanam, 2015). Therefore, they tend to search by brand name instead of by the unique attributes of a brand. In addition, compared with brand keywords, products associated with non-brand keywords may increase consumers' confidence in their decisions if they have low brand knowledge. Hence, we propose: H3: (a) Brand keywords generate higher product sales than non-brand keywords when consumer brand knowledge is low. However, (b) this effect is eliminated when the consumer brand knowledge is high. Hedonic/utilitarian keywords and product sales Hedonic keyword and utilitarian keyword defined Similar to 'brand keyword', we first define 'attribute keyword' as a keyword that contains a product's attribute information. Such attributes can be classified as hedonic or utilitarian based on the benefits and value the product provides (Batra \& Ahtola, 1991; Chitturi et al., 2007; 2008; Strahilevitz \& Myers, 1998). Utilitarian attributes have utilitarian benefits and value, whereas hedonic attributes have hedonic benefits and value (see Dhar \& Wertenbroch, 2013; Jones et al., 2006). For example, for products like shoes, 'round toe' (shape), 'retro design' (fashion trend) and 'net surface' (style) are hedonic attributes, whereas 'ventilated', 'keep warm' and 'antiskid' (related to function) are utilitarian attributes. For products like water purifiers, 'white' (colour), 'mini' (size) and 'wall hanging' (style) are hedonic attributes, whereas 'straight drink', 'reverse osmosis' and 'ultrafiltration' (related to function) are utilitarian attributes. Based on these two well-documented types of product attributes (hedonic and utilitarian), we classified attribute keywords as hedonic keywords and utilitarian keywords. A hedonic keyword mainly describes the aesthetic, experiential and enjoyment-related attributes of a product, whereas a utilitarian keyword mainly describes the functional, instrumental and practical attributes of a product. Keywords such as 'fashion round toe shoes' and 'retro design shoes' are hedonic keywords, whereas 'comfortable toe protection shoes' and 'antiskid shoes' are utilitarian keywords. Hedonic/utilitarian keywords and product sales Products are designed and produced to satisfy various consumer demands (Chitturi et al., 2007; 2008). Products can be classified as hedonic or utilitarian (Bridges \& Florsheim, 2008; Jones et al., 2006; Kempf, 1999; Woods, 1960). Previous studies have indicated that consumers expect different benefits from different types of products. They tend to seek function-related benefits from utilitarian products and experience-related benefits from hedonic products (Dhar \& Wertenbroch, 2013; Jones et al., 2006). When searching online, consumers often type in relevant keywords to describe what they want from a product (Chernev, 2006; Klein \& Melnyk, 2016; van Osselaer \& Janiszewski, 2012). That is, they instinctively input hedonic (utilitarian) keywords when seeking experience (function) related benefits from hedonic (utilitarian) products. Based on regulatory theory, Higgins (2000) pointed out that 'people experience a regulatory fit when they use goal pursuit means that fit their regulatory orientation, and this regulatory fit increases the value of what they are doing' (p. 1217). Regulatory fit makes individuals feel 'right' (Hamstra et al., 2013; Higgins, 2004), thereby enhancing their certainty about their initial goals and increasing their decision-making confidence (Hamstra et al., 2013; Higgins, 2000; Zheng et al., 2015). If the search results for hedonic keywords match the experience benefits that consumers seek, they can be confident in making decisions about hedonic products. Similarly, when consumers have expectations of the functional benefits of a utilitarian product, they focus on utilitarian attributes and input utilitarian keywords. Based on the regulatory fit (Hamstra et al., 2013; Higgins, 2000; 2004; Zheng et al., 2015), once the search results from utilitarian keywords fulfil consumers' function-related needs, the consumers can make their purchasing decisions with confidence. In paid search advertising, it can be assumed that searching keywords that their goals. Hence, we propose: H4: (a) For hedonic products, hedonic keywords generate higher product sales than utilitarian keywords, whereas (b) for utilitarian products, utilitarian keywords generate higher product sales than hedonic keywords. Overview of the studies Two secondary datasets were analysed and one lab experiment was conducted to test our hypotheses. First, we obtained two datasets from two online sellers on Taobao.com, China's largest consumer-to-consumer (C2 C) e-commerce platform. This platform offers sellers a chance to sell their products (including shoes, clothes and electronics) to individual consumers. It also offers keyword auction services to sellers. Sellers in this platform can create and bid for keywords related to their products. Based on a pre-test (see the Results section of Studies 1 and 2 below), we finally chose men's leisure shoes as the hedonic product and water purifiers as the utilitarian product for this study. Both sellers marketed products from multiple brands. The results of the analyses of the two secondary datasets supported H1,H2a,H2b,H4a and H4b. Second, because consumer brand knowledge is hard to measure using secondary data, we conducted a lab experiment using a real mobile phone brand in China to test H3a and H3b. Studies 1 and 2: secondary data analysis Within a three-month window, we downloaded two secondary datasets from two sellers on China Taobao.com. The first dataset was downloaded with the cooperation of a men's shoe seller 1 and, the second dataset was downloaded from a water purifier seller. The men's shoe seller sold several bands of men's leisure shoes, and the water purifier seller sold multiple brands of water purifiers. Overall, 10,966 records of shoe data and 53,701 records of water purifier data were obtained. Data coding Based on the definitions of 'hedonic keyword' and 'utilitarian keyword', four researchers coded the keywords according to the semantic features of Chinese. First, each Chinese semantic group consists of two or more single Chinese characters composed of two or more radicals with a particular semantic feature (Taft et al., 1999). For instance, '' (at leisure, free and having spare time) consists of the two characters '' and '', each of which comprises two semantic radicals. The character '' contains two radicals; the first, ' 1 ', means a man and the second, '', means wood. Therefore, '' refers to a man leaning against wood, feeling pleasant and comfortable. The other character, ' '', contains two radicals; the first, '', meaning a door and the second, '', meaning wood. Therefore, '' refers to a wooden door and to closing the door to sleep. Thus, the semantic unit and clues are the key factors in recognising Chinese words regardless of their radicals and characters. During the coding procedure, the coders needed to bear in mind that each result had to be a complete semantic unit; for example, '' (shoes) and ' ' (shoes + a nonsense syllable) had to be coded as the same unit because each one was a complete semantic unit. Second, hedonic products are different from utilitarian products in their sensory and functional characteristics (Woods, 1960). For instance, hedonic products are dependent on their sensory characteristics (and the visual features of any product, such as colour and design). To a large extent, shoes are appealing to consumers because of their sensory features, such as design, colour and type, whereas their appeal depends to a lesser extent on functional features such as their ability to protect the feet from injury. Conversely, water purifiers mainly attract consumers through their special or powerful functions, not through their design, colour or type. During the coding procedure, the coders referred to the category of hedonic/functional character as a basic coding framework. Third, the meaning of a lexical term can be distinguished according to the attributes of its semantic features, which may be defining or characteristic (Smith et al., 1974). In the keywords '' (a kind of leisure shoe), for example, the defining character '' (shoes) is an essential or defining aspect of the Chinese semantic group, and '' (leisure) indicates a non-essential or characteristic feature of the group. In the coding procedure, the coders considered the difference between defining and characteristic features. Fourth, after each Chinese semantic group was coded (see Table 3), the coders calculated and compared the number of hedonic and utilitarian groups for each keyword, then classified them as hedonic or utilitarian keywords. If the hedonic value outnumbered the utilitarian value, the keyword was classified as hedonic. Likewise, if the utilitarian value outnumbered the hedonic value, the keyword was classified as utilitarian. The keywords were classified as neutral if the numbers for each value were equal. This classification method was adopted from previous studies by Goh et al. (2013), Healey and Kassarjian (1983), and You et al. (2017) in other disciplines. Table 3. Coding results of Chinese semantic group in keywords for the men's shoes and water purifier data. N/A not applicable. Due to the confidentiality agreement, both sellers do not wish to disclose this information. We told the coders the name of the brand with the largest market share. The brand with the largest market share should be coded as 1 and the others as 0