How should co-branding in fashion be done? : investigating the role of fit and incongruity resolution in co-branding in the fashion industry

26-1-2016

Universiteit van Amsterdam

Master’s Thesis – Graduate School of Communication Persuasive Communication

Jara van der Ham 5960061

Abstract

Co-branding, a practice that involves the pairing of two brands to create a single, new product, seems to be increasingly popular in the fashion industry. Therefore, it is important for fashion brands to understand how co-branding should be done in the best way. This research focuses on the effect of product fit and brand fit in branding within the fashion industry, on consumers’ attitude towards the co-branding alliance, towards the brands involved, and the role of incongruity resolution herein. It was assumed that a higher product and brand fit would lead to more positive alliance attitudes, and more positive attitudes towards the involved brands, as mediated by the alliance attitude. Furthermore, a relatively higher importance of brand fit was presumed. However, when people are able to resolve the incongruities between two brands, a more positive effect was expected for a low product and brand fit. To test these assumptions, an online experiment (N = 244) was created in which participants evaluated a hypothetical co-branding alliance. The results indicate that while product fit had no effect on the attitudes towards the alliance and the involved brand, a high brand fit yielded a more positive alliance attitude than a low brand fit, and the effect of brand fit on the attitude towards the involved brands was fully mediated by alliance attitude. Partial evidence for the relatively higher

importance of brand fit, and the moderating effect of incongruity resolution was found. Limitations and suggestions for future research are provided.

Investigating the Role of Fit and Incongruity Resolution in Co-branding in Fashion

‘The Balmain x H&M release was complete chaos’1, ‘Everyone is freaking out about H&M’s new clothing line with Balmain’2_{, and ‘Release of Balmain x H&M}

creates chaos in London’3 _{are just a few of the headlines around the launch of H&M’s}

new designer collaboration in November 2015, which was all everyone could talk about for months. Not surprisingly, H&M is talking about their most successful designer collaboration yet (Bhasin, 2015). Meanwhile Alexa Chung is continuing to design garments for AG Jeans and Dutch department store HEMA launched their collaboration with young designer collective MOAM in December 2015. Co-branding within the fashion industry seems to be a common practice recently, but why are these collaborations so popular?

Co-branding is a practice where two brands join forces to create a new product (Besharat & Langan, 2014). This is seen as a good opportunity to enhance one’s brand image, tap into new markets and most importantly; enhance brand attitudes (Besharat & Langan, 2014). Co-branding is often seen in the FMCG4 industry (e.g. Milka & Oreo), and on technological products (e.g. Nike & Apple). But co-branding practices are becoming increasingly popular in the fashion industry as well, both within the fashion industry (e.g. H&M & Balmain, Uniqlo & Lemaire) and across different industries (e.g. LG & Prada, a crossover of the fashion and technological industry).

Thus, there are numerous ways for a fashion brand to form a co-branding alliance, but it is important for brands to know when positive attitudes towards a co-branding alliance are most likely to occur and on which basis a co-co-branding partner should be selected. One must also bear in mind the specificities of the fashion

industry, which is currently worth 99,7 billion dollars globally and has a value that has increased by 139% percent over the last decade (Millward Brown, 2015). Therefore, it is interesting to explore if classic co-branding theories can be applied to this particular field.

Moreover, however successful some co-branding alliances in the fashion industry might have been, there are also many collaborations that have failed (e.g. Karl Lagerfeld & La Redoute, DKNY & Veuve Cliquot and H&M & Lanvin) receiving disappointing sales and negative consumer reactions (Rollet, Hoffmann, Coste-Manière, Panchout, 2013; Besharat & Langan, 2014). With an overall co-branding failure rate of 80-90 percent, the development of specific co-branding guidelines is increasingly relevant.

Since the beginning of its existence, co-branding research has focused on the concept of between-partner fit. In general, researchers agree that just like in a brand extension, for a consumer to evaluate a co-branding alliance positively, there needs to be some degree of fit (Ahn & Sung, 2012). That is, the partners should be similar to each other on the level of either the product (i.e. product fit) that they bring to the collaboration (Dickinson & Barker, 2007), or the two brand images (i.e. ‘brand fit’) should be compatible (Ahn & Sung, 2012, Dickinson & Barker, 2007). In co-branding research, both types of fit have been deemed important, but empirical evidence of the importance of fit within fashion co-brands is scarce, and findings for the relative importance of brand and product fit are inconsistent (e.g. Baumgarth, 2004; Lafferty, Goldsmith & Hult, 2004; Ahn & Sung, 2012; Wang, Soesilo & Zhang, 2015).

Moreover, co-branded products in the fashion industry that have a high degree of product fit (e.g. H&M x Lanvin), or brand fit (e.g. DKNY & Veuve Cliquot) have not always been proven to be successful (Rollet et al, 2013). Thus, the mere idea that ‘there has to be a degree of fit’, seems too general for these types of alliances.

that don’t have a high degree of fit. If they do, this might result in more positive attitudes towards co-branding alliances that were harder to solve, and thus have a lower degree of fit, as people find enjoyment in resolving incongruities (Sjödin & Törn, 2006). Although research about evaluating a co-branding alliance in high versus low involvement situations (Walchli, 2007) exists, the concept of incongruity resolution has, to the best of my knowledge, not yet been integrated into research on co-branding within the fashion industry.

All in all, a positive attitude towards a co-branding alliance is an indicator of co-branding success and a successful co-branding strategy is believed to be able to enhance the attitude of the involved brands. However, previous research hasn’t fully agreed on the relative importance of product and brand fit and the role of incongruity resolution, and this may be dependent on the industry specifics and characteristics (i.e. a certain type of fit may be more important in the fashion industry than in other industries). Therefore the main research question in this study is:

RQ: What are the effects of product and brand fit in co-branding in the fashion industry, on the attitude towards the co-branding alliance, the attitude towards the involved brands and what role does incongruity resolution play?

The answer to this question will not only contribute to co-branding theories by finding out the relative importance of product fit, brand fit and incongruity resolution in a particular industry, but it will also be very useful to fashion brands that want to engage in a co-branding alliance, as it will make clear which characteristics brands should look for in a co-branding partner. After all, as co-branding is becoming increasingly popular, the dynamics behind this approach are of great concern to marketing scholars and industry. The aim of this research is to provide more clarity around the concept of co-branding and its possible consequences.

In the upcoming section, the theoretical foundation and the hypotheses are discussed, the following section contains an outline of the methods and design of the

conducted research, and lastly the found results and a broad conclusion and discussion of these results are provided.

Theoretical Framework

Co-branding

Co-branding is a term that has received many definitions and has operated under many different names in the past. In this research, co-branding is defined as “the combination of two brands to create a single, new product” (Leuthesser, Kohli & Suri, 2003; Besharat & Langan, 2014). Often a distinction can be made between the parent brand (i.e. mostly the ‘first’ brand in a collaboration, that initiates the co-branding and for instance sells the co-branded product) and the partner brand (i.e. the ‘second’ brand). Even though co-branding has existed for years, it’s popularity now seems to grow vastly with an increase in usage of 20 to 40% yearly (Helmig, Huber & Leeflang, 2007; Geylani, Inman & Hofstede, 2008; Ahn & Sung, 2012; Besharat & Langan, 2014). This is due to the fact that co-branding is an excellent opportunity for brands to tap into new markets, leverage their brand equity and to differentiate products in competitive environments (Helmig, Huber & Leeflang, 2007). However, the end goal of co-branding is often generating more positive attitudes towards the constituent brands. That is, it is expected that a positive attitude towards the branding alliance can spill over to the partnering brands. However, when co-branding is not done in the ‘right’ way, brands are often warned that this can lead to an impairment or dilution of the involved brands. Particularly, brands might lose some of their credibility, or unwillingly change some important brand associations by not pairing up with the right partner (Geylani, Inman & Hofstede, 2008).

Co-branding in the fashion industry

Kim, Ko, Lee, Mattila & Hoon Kim, 2014). The way consumers experience a brand is especially important regarding fashion brands, as these brands typically have a more symbolic and emotional meaning to consumers than for instance FMCGs (Kim, 2012; Liu, Li, Mizerski & Soh, 2012). Moreover, in order to fulfill consumer needs, for a fashion brand it is important to differentiate and offer a unique way of expressing oneself (Kim et al, 2014). This may be the reason that fashion brands are constantly trying to alter and improve their brand image by co-branding with other brands.

One popular way in which co-branding within the fashion industry exists, is through collaborations between luxury designers and mass-market fashion retailers (e.g. H&M x Balmain, Uniqlo x Lemaire, Target & Missoni), which is believed to enhance the image of the mass-market brand (Kim et al, 2014). Furthermore, fashion brands often form co-branding alliances with brands that position themselves outside of the fashion industry, such as technological products (e.g. Prada x LG). These kinds of co-branding alliances have been extensively researched from a managerial perspective (i.e. what value does co-branding have to the engaged partners in terms of extending the customer base, e.g. Bucklin & Sengupta, 1993; Wigley &

Provelengiou, 2011; Oeppen & Jamal, 2014), but it is still rather unclear what effect these alliances have on consumers (i.e. how do consumers perceive co-branding alliances).

Co-branding, fit and spillover effects

In order to understand the effects of co-branding on the consumer, the attitude that consumers hold towards the co-branding alliance can’t be overlooked. After all, this is the most immediate reaction that occurs in consumer’s minds after being exposed to a co-branding alliance, and it is this attitude that can eventually be responsible for purchase intentions and commercial success (MacKenzie, Lutz & Belch, 1986).

Researchers agree that for a co-branding alliance to be evaluated positively by consumers there needs to be a degree of fit. This proposition is largely building upon brand extension theories, which dictate that in order for a consumer to

understand the extension, and thus be able to evaluate it properly, it must fit with the parent brand (Aaker & Keller, 1990; Mao & Krishnan, 2006; Ahn & Sung, 2012). Often, people tend to like similarity because it provides reassurance and self-confirmation (Ahn & Sung, 2012). Moreover, a high degree of fit would enhance positive value and decrease the likelihood of negative outcomes for the consumer (Ashton & Scott, 2011).

Further findings of this kind were discovered regarding co-branding, which confirm the importance of fit. A perceived match-up between two brands within different product categories leads to a more positive attitude towards the co-branding alliance (Ahn, Kim & Forney, 2010) and particularly in the fashion industry, where it is found that a co-branding alliance needs to be coherent, indicating that there needs to be a fit between the partner brands (Rollet et al, 2014). However, this idea of ‘fit’ should not be seen as a unitary concept and can be further divided into two

dimensions; product fit and brand fit. These concepts will be further elaborated in the forthcoming sections.

Product fit and attitude towards the co-branding alliance

Product fit in co-branding describes how the product categories as suggested by the co-branding are related, regardless of the brands (Simonin & Ruth, 1998; Wang, Soesila & Zhang, 2015). It implies that the problem-solving capacity, or the attributes of each partner brand are related, which often evolves out of the

constituent brands producing products of the same product category (e.g. Target and Missoni have a high product fit, as they both bring products to the alliance that are in the same category (i.e. clothing), and are used in similar situations). In the past, this

these studies indicate that a high product fit results in more positive perceptions of the alliances (Simonin & Ruth, 1998; Ahn & Sung, 2012). This can be explained by using categorization theory (Fiske & Pavelchak, 1986; Aaker & Keller, 1990; Ahn & Sung, 2012), which explains that people tend to use existing schemas to categorize the world around them and thus, try to fit new information into existing categories. If the new information is consistent with the existing schema, affect that is stored regarding that category is used to evaluate the new information quickly, and without much effort (Fiske & Pavelchak, 1986). As such, brands are regarded as categories. If a high fit between alliance partners is perceived, this is consistent with people’s pre-existing schema and people use category-based processing, which means that the co-branding alliance is perceived in one and the same category. Consequently, the perception of one alliance partner will transfer directly to the other alliance partner, allowing people to efficiently evaluate the collaboration (Ahn & Sung, 2012). On the other hand, if there is a low fit, the alliance is not seen as fitting into one category and this enhances piecemeal processing of the separate attributes of both brands, which takes more time, lacks perceivable logic, and inhibits the transfer of affect. Also, this will cause frustration and make people question the validity of the co-branding alliance (Aaker & Keller, 1990; Ahn & Sung, 2012). Therefore, it is expected that:

H1: A high product fit in co-branding alliances in the fashion industry will lead to more positive attitudes of the co-branding alliance than a low product fit.

Product fit and spillover effects

Previous research does not only indicate that a high product fit leads to more positive evaluations of the alliance but has also focused on spillover effects (Simonin & Ruth, 1998; Ahn & Sung, 2012). A spillover effect occurs when evaluations that people have about the co-branding alliance transfer to the separate brands. A

positive attitude towards the alliance will thus lead to a more positive attitude towards the involved brands. For this, a high degree of fit is needed, because when brands have very few commonalities, attributes of the co-branded product are more easily discounted, and consumers’ attitude towards a brand evolves only slightly or not at all (Geylani, Inman & Hofstede, 2008). This effect of product fit on spillover effects has been replicated by some studies (Simonin & Ruth, 1998).

Furthermore, the co-brand will be seen as the integration of the best attributes of both brands, which highlights the positive aspects of these brands making them more salient, which will consequently result in a more positive attitude towards the separate brands. This happens in part because a positive attitude towards the alliance is already created and this attitude then spills over to the constituent brands (Leuthesser, Kohli & Suri, 2003). It is thus expected that:

H2a: A high product fit in co-branding alliances in the fashion industry will lead to a higher post-alliance attitude of the parent(A) and the partner brand(B), than a low product fit.

H2b: The effect of product fit on post-alliance attitudes will be mediated by the attitude towards the alliance.

Brand fit and attitude towards the co-branding alliance

An emerging body of studies has found that even when product fit is very low, co-branding alliances can still produce positive attitudes, because people still

perceive a degree of harmony or coherence between co-branding partners (Ahn, Kim & Forney, 2009). This may be a result of brand fit, which is a match between the symbolic meanings and brand concepts of the constituent brands. This includes corporate, product and symbolic images (Huili, Sin & Juan, 2012) and meanings uniquely related to the brand (Wang, Soesila & Zhang, 2015). In this way,

co-restaurants (Ashton & Scott, 2011), cars and microprocessors (Simonin & Ruth, 1998; Baumgarth, 2004) and between a charity and a commercial brand (Lafferty, Goldsmith & Hult, 2004; Dickinson & Barker, 2007) have proven to be successful. This can also explain the success of collaborations between fashion brands and brands from other product categories (e.g. Mercedez-Benz and Armani have a low product fit, but have a high brand fit because they share brand meanings of luxury and quality).

Apart from the aforementioned categorization theory, this can also be explained by a match-up effect. The match-up theory posits that a more positive effect will be attained in advertising if an endorser is coherent with the image of the endorsed brand (Erdogan, 1999), thus indicating that when people find a ‘match’ between two stimuli, they will evaluate the stimuli on the basis of stored affect in the pre-existing category schema (Ahn, Kim & Forney, 2010). Another possible

explanation is congruity theory, which states that people are looking for consistency in the world around them, which will affect an ill-fitting co-branding alliance in a negative way (Lafferty, Goldsmith & Hult, 2004). Consequently, it is expected that:

H3: A high brand fit in co-branding alliances in the fashion industry will lead to more positive attitudes of the co-branding alliance than a low brand fit.

Brand fit and spillover effects

Co-branding studies have also found positive effects of brand fit on the positive spillover of attitudes towards partnering brands (Simonin & Ruth, 1998; Dickinson & Barker, 2007; Wang, Soesilo & Zhang, 2015). When a brand forms an alliance with another brand that has a high brand fit, the co-brand will be perceived as one and the same category, combining the best of both brand images, thus making these aspects salient. This will in turn lead to a more positive post-alliance attitude towards the separate involved brands (Leuthesser, Kohli & Suri, 2003).

Again, this effect can be explained by a spillover of the positive attitude, which was previously generated by the alliance itself.

H4a: A high brand fit in co-branding alliances in the fashion industry will lead to a higher post-alliance attitude of the parent(A) and the partner brand(B), than a low brand fit.

H4b: The effect of brand fit on post-alliance attitudes will be mediated by the attitude towards the alliance.

The relative importance of product and brand fit

Previous research has found mixed results of the importance of product and brand fit, when comparing the two. While some research neglected the relative importance of both types of fit, other research has found that the match-up between the perceived brand images of two partnering brands wasn’t essential (Ahn, Kim & Forney, 2010), and that product fit is the leading form of fit that determines the evaluation of a co-branding alliance (Geuens & Pecheux, 2006). However, the majority of studies indicate brand fit to be more important than product fit in facilitating positive attitudes and spillover effects (Baumgarth, 2004; Lafferty, Goldsmith & Hult, 2004; Ahn & Sung, 2012; Wang, Soesilo & Zhang, 2015).

Also, the fashion industry is quite a specific one, in which symbolic meanings are especially important (Wigley & Provelengiou, 2011; Liu, Mizerski & Soh, 2012). Co-branding across different product categories may increase the added value of a co-branded product, as in this way, a fashion brand can benefit from the knowledge and expertise of a brand belonging to another product class (e.g. when LG

co-branded with Prada, they benefited from their knowledge of fashion design and style, which is a skill that LG doesn’t inherently posses). Moreover, the failure of certain co-branding alliances in the fashion industry in the past might be explained by the brand concepts of the constituent brands being too far apart (Rollet et al, 2013).

Altogether, this would suggest that brand fit is more important than product fit when co-branding alliances are formed with a fashion brand. In other words, a high brand fit will lead to more positive evaluations than a high product fit.

This would also suggest that:

H5: The effect of fit will be more pronounced for brand fit than for product fit, such that if brand fit is high, but product fit is low, attitudes towards the alliance and post-alliance attitudes towards the parent(A) and the partner brand(B) will be more positive than when product fit is high but brand fit is low.

Incongruity resolution

Although the idea of fit in co-branding alliances is severely stressed in

literature, a high degree of fit may not be ideal in all co-branding situations. In real life settings, a positive effect of co-branding without a logical fit occurs frequently (e.g. the H&M x Balmain collections involved pairing with a luxury brand with a brand image that some would say differs largely from that of H&M, but it was still their biggest co-branding success to date, Bhasin, 2015). In other words, brand fit may be of greater importance than product fit, but there are still cases of co-branding

alliances in the fashion industry, where brand fit was relatively low, but the co-branding alliance was a raging success nonetheless. Some research has indeed found no effects of product and brand fit on consumer evaluations of the co-brand at all (Singh, Kalafatis & Ledden, 2014), which may be due to incongruity resolution.

Incongruity, in this case, entails a discrepancy between the communications of a brand (through co-branding) and the pre-existing schemas regarding that brand in people’s memory (Sjödin & Törn, 2006). While studies of advertising and branding have mainly focused on the importance of congruity, studies that find inverse, and thus positive effects of incongruity are beginning to emerge (e.g. Dahlén, Rosengren,

Törn, & Öhman, 2008). This is due to the fact that an incongruent stimulus enhances processing, which in turn makes consumers appreciate the ad, and consequently the brand more. Especially when the incongruity can be resolved, that is, when a logical link between the new information and the pre-existing schema of the brand can be found, it provides an entertaining ‘puzzle’ for consumers, which brings positive affect as well (Dahlén, Rosengren, Törn, & Öhman, 2008). Following this explanation, in co-branding situations it can also be expected that when the incongruity between two partners can be resolved, this leads to a higher evaluation of the incongruent

element. An incongruent co-branding alliance, with a lower degree of fit, should generally be processed more extensively, and provide satisfaction. When the incongruity remains unresolved, a negative effect on evaluations of the incongruent element is to be expected (Sjödin & Törn, 2006). This may be an indication of a moderating effect of incongruity resolution on the attitude towards a co-branding alliance.

Previous research in the field of co-branding has indeed found greater positive effects of moderate fit than high fit (Walchli, 2007; Ahn & Sung, 2012). Particularly when consumers are highly involved and are expected to elaborate more on the incongruent co-branding alliance and are thus more likely to resolve it, a lower degree of fit leads to more positive attitudes towards the co-branded product than a high degree of fit. Namely, Walchli (2007) has found such results when researching co-branding of magazines. A similar pattern is expected for co-branding in the fashion industry:

H6: When the incongruity between co-branding partners is resolved, a higher product or brand fit leads to less favorable attitudes towards the alliance and the constituent brands than a low degree of brand or product fit

Method

Sample

To test the hypotheses, a convenience sample of 307 participants was recruited through Social Media, email and online forums. Of these participants, 61 were excluded from further analyses because they hadn’t completed the entire survey, and one participant was excluded for having indicated that he/she had participated in the pre-test. Thus, the results of 244 participants (34.4% male and 63.9% female) were further analyzed. The majority of the sample (75%) was Dutch. The age of participants ranged from 18 to 79 with a mean age of 27.65 (SD = 9.50). The experiment was conducted online so that participants could participate at their own convenience. As a reward, 10 Google Cardboard Virtual Reality headsets were raffled among all participants who had provided their email address at the end of the survey.

Design

To measure the effects of fit and incongruity resolutions on attitude towards the alliance and the constituent brands, a 2(product fit: low vs. high)x2(brand fit: low vs. high)x2(incongruity resolution: yes/no) mixed factorial design was conducted, with attitude towards the alliance, post-alliance attitude change towards brand A (the parent brand) and post-alliance attitude change towards brand B (the partner brand) as dependent variables. In doing so, 8 experimental conditions were created

Materials

Following the lead of numerous previous studies (e.g. Simonin and Ruth, 1998; Baumgarth, 2004; Helmig et al., 2007), this study conducted an experimental design with hypothetical co-branding alliances between well-known brands. This was done not only to increase external validity, but rather, using familiar brands was necessary for this study, as participants should have prior brand knowledge in order to assess the degree of fit (Simonin & Ruth, 1998), and an incongruity may only be

resolved by using existing brand knowledge. Also, to test for a change in attitudes towards a brand due to a spillover effect of the co-branding alliance, it was critical to be able to measure already existing brand attitudes.

Hence, hypothetical co-branding alliances were created using existing brands. These co-branding alliances were presented to respondents in the form of a news article on the website FashionUnited.com, which announced the launch of a particular co-branding alliance (see Appendix C). The news article allowed for the manipulations of brand fit, product fit and incongruity resolution.

Pre-tests

In order to assess which kind of brand pairs were regarded as having a low vs. high product and brand fit, a pre-test was conducted among 42 participants with a mean age of 30.38 (SD = 1.04). Of these respondents, 15 were male and 27 were female. A total of 8 brand pairs were evaluated, among which 4 included the fast fashion brand H&M and 4 included the jeans brand Levi’s, both mixed with several other brands (see Appendix B). For each brand pair people were asked to which degree they thought the brands concepts, images and meanings (i.e. brand fit) or product categories (i.e. product fit) were similar and/or compatible on a 7-point Likert scale.

The results indicated that for this study, the brand pairs involving Levi’s were most suitable, as it provided more useable brand pairs than H&M. It was preferred to use the same parent brand across conditions, to keep everything other than product and brand fit as constant as possible. In terms of product fit, a low fit was perceived when the brand (Levi’s) was paired with technological brands (Levi’s and Apple had a product fit of 2.00 (SD = .00) and Levi’s and Dell had a product fit of 1.67 (SD = 1.11)) and a high product fit was perceived when the brand was paired with other fashion brands (Levi’s and Diesel had a product fit of 6.25 (SD = .96) and Levi’s and Burberry had a product fit of 5.22 (SD = 1.86)). A paired samples T-test indicated that

these means differ significantly from one another (Levi’s x Apple vs. Levi’s x Diesel, t(3) = -8.88, p < .001; Levi’s x Apple vs. Levi’s x Burberry, t(8) = -2.79, p < .05; Levi’s x Dell vs. Levi’s x Diesel, t(14) = -9.57, p < .001; Levi’s x Dell vs. Levi’s x Burberry, t(6) = -7.84, p < .001).

In terms of brand fit, scores were lowest when Levi’s was paired with Dell (M = 2.27, SD = 1.53) or Burberry (M = 2.71, SD = 1.25) and a higher brand fit was perceived when Levi’s was paired with Diesel (M = 5.86, SD = 1.35) or Apple (M = 3.43, SD = 1.51). As indicated by a paired sample T-test, Levi’s x Dell and Levi’s x Burberry differed significantly from Levi’s x Diesel (t(14) = 7.12, p < .001; t(6) = -7.78, p < .001). However, the difference between Levi’s x Apple, Levi’s x Dell and Levi’s x Burberry was not significant.

In order to find a technological brand that brings about a higher brand fit when paired with Levi’s than Apple, a second pre-test was conducted among 8 participants with a mean age of 25.55 (SD = 5.92), consisting of all females. Participants

evaluated the brand fit of Levi’s with Canon, Olympus, Polaroid and Samsung. In this second pre-test, Polaroid was regarded as having the highest brand fit with Levi’s (M = 3.63, SD = 2.26), followed by pairings with Canon (M = 3.25, SD = 1.83), Samsung (M = 3.00 SD = 1.31) and Olympus (M = 2.00 SD = 1.07).

Manipulations of product fit and brand fit

Following the pre-test, the low product fit conditions were created by pairing Levi’s with Dell and Polaroid (technological brands), and the high product fit

conditions were created by pairing Levi’s with Diesel and Burberry (fashion brands). The low brand fit conditions were created by pairing Levi’s with Dell and Burberry, and the high brand fit conditions were created by pairing Levi’s with Diesel and Polaroid (as this brand had the highest brand fit scores in the second pre-test).

Incongruity resolution

The manipulation of incongruity resolution was twofold. Firstly, following the research of Walchli (2007), the incongruity resolution condition was created by priming involvement, as it is believed that when people are more involved, they resolve incongruities more quickly (Maoz & Tybout, 2002; Walchli, 2007). This was done by asking respondents to view the article that announced the co-branding with attention and instructing them to ‘really try to understand the co-branding alliance’. The no incongruity resolution condition was then created by priming low involvement, and asking respondents to view the article as if they would encounter it in real life and telling them that their ‘first and global impressions about it were what is important’.

Secondly, the articles in the incongruity resolution conditions contained subtle explanatory links (Pryor & Brody, 1998; Bridges, Keller & Sood, 2000). In other words, in order to manipulate participants in resolving a low fit, certain commonly held associations between the two partnering brands were made salient in the article. In the no incongruity resolution conditions, these attributes were replaced by

irrelevant and not commonly held associations. For instance, Levi’s and Polaroid were both described as pioneers, being experts on timeless products in the

incongruity resolution conditions, whereas in the no incongruity resolution conditions, Levi’s was just described as a jeans brand and Polaroid as an instant film and

camera producer (see Appendix C for other conditions). A schematic overview of all experimental conditions can be found in Table 1.

Table 1. Schematic outline of all experimental conditions

Condition 1 Condition 2 Condition 3 Condition 4

Product fit high low high low

Brand fit high high low low

Incongruity resolution no no no no

Manipulation Levi's x Diesel Levi's x Polaroid Levi's x Burberry Levi's x Dell American and Italian Jeans brand and instant jeans brand and jeans brand and jeans brand film producer luxury fashion house computer brand

Measurements

Attitude towards the co-branding alliance

To measure the attitude towards the co-branding alliance (alliance attitude), participants were asked to indicate how they would evaluate the co-branding alliance they had just seen by rating 5 bipolar items on a 7 point Likert scale;

negative/positive, unsatisfactory/satisfactory, unfavorable/favorable, bad/good, unlikeable/likeable (Appendix D) (Ahn & Sung, 2012). These items were averaged to calculate a mean score for alliance attitude (Cronbach’s α: .92, M = 4.29, SD = 2.01).

Post-alliance attitude change towards the constituent brands

To test if participant’s brand attitudes had changed after the alliance due to a spillover effect, attitudes towards the constituent brands were measured within- subjects, by using the same brand attitude scale before (pre-alliance) and after (post-alliance) exposure to the co-branding alliance (see Appendix D). After seeing a picture of the brand’s logo, participants were asked how they would evaluate these brands separately, using 6 bipolar items on a 7 point Likert scale; bad/good, unpleasant/pleasant, unfavorable/favorable, negative/positive, dislikeable/likeable, poor quality/high quality (Boerman, Reijmersdal & Neijens, 2012). After viewing the co-branding alliance, participants answered the same questions. All used scales were reliable (pre-alliance attitude brand A: Cronbach’s α: .87, M = 5.38, SD = 1.29; brand B: Cronbach’s α: .88, M = 4.57, SD = 1.83; post-alliance attitude brand A: Cronbach’s α: .89, M = 5.26, SD = 1.28; brand B: Cronbach’s α: .91, M = 4.67, SD = 1.72)

Condition 5 Condition 6 Condition 7 Condition 8

Product fit high low high low

Brand fit high high low low

Incongruity resolution yes yes yes yes

Manipulation Levi's x Diesel Levi's x Polaroid Levi's x Burberry Levi's x Dell

Both jeans Both oldest, experts Both classic, experts Both experts, encou- masters on timeless products on timeless fashion raging people to be

Control variables

To obtain a clear picture of the sample, respondents were asked about their gender, age and nationality at the end of the survey. Furthermore, respondents received questions about their familiarity with the involved brands and if they have experience using these brands. Previous research has indicated that these factors, as well as pre-existing brand attitudes, may influence people’s perceptions of co-branding alliances (e.g. Simonin & Ruth, 1998; Wang, Soesilo & Zhang, 2015). These variables could thus be taken into account as control variables.

Manipulation Check

To check if a low and high brand fit were regarded as such, each respondent was asked to rate how similar and/or incompatible the product categories of both brands are on a 7-point Likert scale. Similarly, to check if a high and low product fit were really regarded as such, respondents were asked to rate how similar and/or incompatible the brand concepts, brand images and brand meanings of the brands were on a 7-point Likert scale. These questions were asked before participants were exposed to the stimulus material, so that the experimental manipulations could not influence their perceptions of fit.

Lastly, to check if the incongruity between brands was really resolved more in the incongruity resolution conditions, respondents were asked if they saw the link between the brands, if the collaborations made sense, and if they understood why the brands chose to collaborate. In the incongruity resolution conditions, these questions were asked before the attitude questions, so that that participant’s attitudes would indeed be influenced by a perception of fit (i.e. incongruity

resolution). In the no incongruity resolution conditions, these questions were asked after the attitude measures, to ensure that if the mere posing of these questions would cause participants to think about a possible link between brands, this wouldn’t affect their attitudes. Respondents indicated how much they agreed with each

statement on a 7 point Likert scale (1: ‘Strongly Disagree’; 7: ‘Strongly Agree’). These scores were then averaged to calculate a mean score for incongruity resolution. A Principal Component Analysis revealed that all items loaded on one factor (EV = 2.46, R² = 81.89). Moreover, the scale was reliable (Cronbach’s α: .89,

M = 3.48, SD = 3.12).

Procedure

At the beginning of the survey, participants read an informed consent

(Appendix A) and additional information about the research, and had to accept this in order to be able to proceed. To prevent demand artifacts, participants were led to believe that the goal of the research was to find out “how people respond to new product ideas in the fashion industry when they see them for the first time.” After accepting the informed consent and reading the instructions, respondents were automatically and randomly assigned to 1 of 8 conditions.

Firstly, the logos of the two bands were shown and pre-alliance brand attitudes were measured. Directly afterwards, participants received questions to measure the perceived brand fit, product fit, brand familiarity and brand usage for both brands. Then, participants received instructions about the article they were about to see, explaining to them what co-branding entails and priming either low or high involvement. Next, participants were exposed to the manipulated news article containing the hypothetical co-branding alliance. After 20 seconds, participants were able to go to the next page. After being exposed to the article, the manipulation check for incongruity resolution took place and attitudes towards the alliance and post-alliance attitudes towards the brands were measured. In the last section of the research, respondents answered some demographic questions, they were asked if they had taken part in the pre-test and could write down what they thought the purpose of the research was. Finally, participants were informed that the material used was fake and were asked to leave their email address to be able to win a

Google Cardboard virtual reality headset. Participants were then thanked for their time and asked to click on the ‘next’ button to save their responses. The whole experiment took between 5 and 10 minutes.

Results

Control variables

To check if certain control variables differed across conditions, a Chi-Square test was conducted including the variables condition, age and gender. No significant associations were found, which means that age and gender were equally distributed across conditions. Furthermore, separate one-way ANOVAs were conducted to check for the equal distribution of pre-alliance attitudes, brand familiarity and brand usage experience. These indicated that the pre-alliance attitude towards brand B, F(7,236) = 2.46, p = .019, η2 = .068, familiarity of brand B, F(7,236) = 4.39, p < .001, η2 = .115, and usage experience of brand B, F(7,236) = 8.62, p < .001, η2 = .204, differed significantly across conditions. These variables were thus considered as covariates in all further analyses.

Manipulation checks

To test whether a low and high product fit, a low and high brand fit and the presence or absence of incongruity resolution were actually regarded as such, several independent T-tests were conducted as a manipulation check. Firstly, this revealed that participants in the high product fit conditions regarded a significantly higher product fit (M = 4.08, SD = 1.81) than participants in the low product fit conditions (M = 2.22, SD = 1.32), t(215.21) = -9.19, p < .001, 95% CI [-2.27, -1.47]. Secondly, participants in the high brand fit conditions perceived a significantly higher brand fit (M = 3.75, SD = 1.70) than participants in the low brand fit conditions (M = 2.73, SD = 1.37), t(236.90) = -5.17, p < .001, 95% CI [-1.41, -.63]. Lastly, participants

in the ‘incongruity resolution conditions’ scored significantly higher on the incongruity resolution scale that was used for the manipulation check (M = 3.85, SD = 1.56) than participants in the ‘no incongruity resolution conditions’ (M = 3.06, SD = 1.53), t(241) = -3.98, p < .001, 95% CI [-1.18, -.40]. The manipulations for product fit, brand fit and incongruity resolution were thus all considered successful.

Hypotheses testing Alliance attitude

To test the effect of product fit and brand fit on alliance attitude, a two-way ANCOVA was run with product fit and brand fit as independent variables, alliance attitude as dependent variable, and the above-mentioned covariates. In contrast to the expectations, there was no significant main effect of product fit on alliance attitude (all test results for alliance attitude can be found in Table 2). Participants in the high product fit conditions had a slightly higher alliance attitude (M = 4.41, SD = .11), than participants in the low product fit conditions (M = 4.15, SD = .10), but this difference was not significant. A high product fit did thus not lead to a more positive alliance attitude than a low product fit, which means that Hypothesis 1 has to be rejected.

The analysis did however reveal a significant main effect of brand fit. Alliance attitudes were significantly higher when participants were exposed to a co-branding alliance with a high brand fit (M = 4.56, SD = .10) than when they were exposed to a low brand fit (M = 4.00, SD = .11). This means that, in line with the expectations, a high brand fit in co-branding alliances in the fashion industry yielded more positive alliance attitudes then a low brand fit, which confirms Hypothesis 3.

Lastly, to see if the effect on alliance attitude is more pronounced for brand fit than for product fit, an interaction effect was sought after. Comparing the mean scores in the partial fit conditions suggests that brand fit is a greater determining factor in generating a positive alliance attitude. Participants in the high brand fit / low

product fit conditions have more positive alliance attitudes (M = 4.52, SD = .14) than those in the low brand fit / high product fit conditions (M = 4.22, SD = .16).

Nevertheless, in contrast with the expectations, no significant interaction effect of product fit and brand fit on alliance attitude was found. Hypothesis 5 can thus not be confirmed regarding alliance attitude.

Table 2. Degrees of freedom, Mean Squares, F, Statistical significance and Eta-Square from ANCOVA for alliance attitudes.

df Mean Square F p η2

product fit (low vs. high) 1, 237 3.80 3.00 .085 .012 brand fit (low vs. high) 1, 237 17.95 14.14 < .001 .056 product fit x brand fit 1, 237 1.54 1.22 .271 .005

Post-alliance attitude change towards the constituent brands

To analyze the effects of product and brand fit on post-alliance attitude change towards brand A (the parent brand) and brand B (the partner brand), a mixed two-way MANCOVA was conducted with product fit and brand fit as between-subject variables, the post-alliance attitude change towards brand A and brand B (consisting of pre-alliance and post-alliance attitudes) as within-subject variables and the

aforementioned covariates. The means are presented in table 3.

Table 3. Estimated Marginal Means(Standard Deviations) and difference scores of pre- and post-alliance attitude towards brand A and brand B for a low and high product fit and a low and high brand fit.

attitude Brand A attitude Brand B

pre-alliance post-alliance Diff pre-alliance post-alliance Diff product fit low (n =125 ) 5.17(.08) 4.62(.08) -.55 4.87(.07) 4.91(.07) .04

high (n = 118) 5.28(.08) 4.60(.08) -.68 4.88(.08) 5.00(.07) .12

brand fit low (n = 117) 5.14(.08) 4.43(.08) -.70 4.78(.08) 4.79(.07) .01 high (n = 126) 5.31(.08) 4.79(.08) -.52 4.97(.07) 5.12(.07) .15 Note: Diff: Difference scores = pre-alliance attitude - post-alliance attitude. A more positive score means a more positive attitude change.

Interestingly, the attitude towards brand A seems to decrease with low as well as high product fit and brand fit. Furthermore, no interaction effects of product fit or of

brand fit with post-alliance attitude change towards brand A were found (all test results can be found in Table 4). A high product or brand fit did thus not lead to a more positive attitude change towards brand A.

For brand B the attitude becomes slightly more positive after exposure to a high product fit than to a low product fit. A similar pattern arises when participants were exposed to a high brand fit in comparison to a low brand fit. However, in contrast to the expectations, no interaction effects of product fit or brand fit with post-alliance attitude change towards brand B were found. A high product or brand fit does thus not lead to a more positive attitude change towards brand A and brand B and Hypotheses 2a and 4a have to be rejected.

Table 4. Degrees of freedom, Mean Squares, F, Statistical significance and Eta-Square from ANCOVA for post-alliance attitude change towards brand A and brand B.

df Mean Square F p η2

changeAbA * product fit 1, 237 1.01 1.01 .315 .004 changeAbA * brand fit 1, 237 2.18 2.18 .141 .009 changeAbB * product fit 1, 237 .34 .97 .327 .004 changeAbB * brand fit 1, 237 1.25 3.51 .062 .015 changeAbA * product fit 1, 237 5.44 5.44 .021 .022 * brand fit

changeAbB * product fit 1, 237 .09 .24 .625 .001

* brand fit

Note: ChangeAbA and ChangeAbB represent the post-alliance attitude change towards brand A and brand B

To see if the effect is more pronounced for brand fit than for product fit, three-way interaction effects with the post-alliance attitude change towards brand A and brand B were sought after. Such an interaction was indeed found for the attitude change towards brand A, product fit and brand fit. The interaction effect is depicted in Figure 1.

Figure 1. Line graphs of interaction effect of product fit, brand fit and post-alliance attitude change towards brand A.

As can be seen from the figure, when product fit is low, the attitude towards brand A decreases less when brand fit is high (Mdifference = -.29), than when brand fit is low (Mdifference = -.80). A significant interaction effect of brand fit and post-alliance attitude change towards brand A, F(1, 121) = 7.57, p < .007, η2 = .059, proved that this difference is significant. Furthermore, when brand fit is high, the attitude towards brand A

decreases less when product fit is low (Mdifference = -.29) than when product fit is high (Mdifference = -.74) and this difference proved significant by a significant interaction effect of product fit and post-alliance attitude change towards brand A, F(1, 122) = 8.30, p < .005, η2 = .064. As for brand B, no such interaction effect was found, meaning that there was no difference in the effects of product fit or brand fit on the attitude change towards brand B.

All in all, the analysis reveals that attitudes decreased less when brand fit was high and product fit was low, and thus had a ‘more positive’ effect than when product fit was high and brand fit was low. These results indicate that brand fit is indeed a more important factor in bringing about a positive brand attitude, but only towards brand A (and not brand B and alliance attitudes), and only in helping the brand attitude to decrease less, which is not entirely in line with what was expected. Consequently, Hypothesis 5 can only be considered partially supported.

Mediation effects of alliance attitude

To test for possible mediation effects of alliance attitude, Hayes’ PROCESS model 4 was used. In all analyses the post-alliance attitude towards the brand was used as a dependent variable and the pre-alliance attitude towards this brand was controlled for by inserting it as a covariate, together with the covariates earlier described. Dummy coded variables of product fit and brand fit served as the independent variables.

To analyze the mediating effect of alliance attitude in the effect of product fit on post-alliance attitude towards brand A, firstly the direct effect of product fit on post-alliance attitude towards brand A was examined. This model was significant, F(5, 237) = 43.62, p < .001, R2 = .48, but only due to the pre-alliance attitude towards brand A being a significant predictor of post-alliance attitude, b = .66, t(237) = 13.29, p < .001. No direct effect of product fit on post-alliance attitude towards brand A was found, b = .03, t(237) = .29, p = .770. Following the Baron and Kenny (1986) method, since a direct effect wasn’t found, a mediating effect could no longer exist.

Consequently, against expectations, the effect of product fit on post-alliance attitudes towards brand A was not mediated by alliance attitude.

To analyze this mediation effect for brand B, the above steps were repeated. The direct effect of product fit on the post-alliance attitude towards brand B was first examined. This model was significant F(4, 238) = 67.23, p < .001, R2 = .53, but again, this was only because of the effect of the pre-alliance attitude towards brand B, b = .72, t(238) = 15.06, p < .001, and no direct effect of product fit on post-alliance attitude towards brand B was found, b = .12, t(238) = 1.23, p = .220. The effect of product fit on the post-alliance attitude towards brand B could thus also not be mediated by alliance attitude. As a result, Hypothesis 2b has to be rejected.

To analyze the mediating effect of alliance attitude in the effect of brand fit on post alliance attitudes, first the direct effect of brand fit on the post-alliance attitude towards brand A was tested. This model proved to be significant, F(5, 237) = 45.09, p

< .001, R2 = .49. Even after controlling for pre-alliance attitude towards brand A, b = .66, t(237) = 13.48, p < .000, brand fit was a significant predictor of post-alliance attitude towards brand A, b = .17, t(237) = 1.98, p = .049. Secondly, the effect of brand fit on alliance attitude was analyzed. This model was also significant, F(5, 237) = 13.64, p < .001, R2 = .22, and brand fit emerged as a significant predictor of

alliance attitude, b = .57, t(237) = 3.84, p < .001. Then, the combined effect of brand fit and alliance attitude on post-alliance attitude towards brand A was looked at. This model was also significant, F(6, 236) = 40.27, p < .001, R2 = .51. Further analysis revealed that alliance attitude was a significant predictor of post-alliance attitude towards brand A, b = .11, t(236) = 2.96, p = .003, and when controlling for alliance attitude, the effect of brand fit on post-alliance attitude towards brand A was no longer significant, b = .11, t(236) = 1.23, p = .219. Finally, looking at the bootstrapped unstandardized indirect effect, b = .06, 95% CI [.02, .15], proved that the indirect effect was significant. Just as expected, the effect of brand fit on post-alliance attitudes towards brand A was fully mediated by alliance attitude.

The effects for brand B were analyzed in the same manner. Firstly, the direct effect of brand fit on the post-alliance attitude towards brand B was tested. This model was significant, F(4, 238) = 70.24, p < .001, R2 = .54, and after controlling for the pre-alliance attitude towards brand B, b = .70, t(238) = 14.62, p < .001, brand fit was still a significant predictor of post-alliance attitude towards brand B, b = .26, t(238) = 2.68, p = .007. The model testing the effect of brand fit on alliance attitude also proved to be significant, F(4, 238) = 15.61, p < .001, R2 = .21, and brand fit once again emerged as a significant predictor of alliance attitude, b = .57, t(238) = 3.81, p < .001. Next, the model testing the combined effect of brand fit and alliance attitude on post-alliance attitudes towards brand B was also significant, F(5, 237) = 68.33, p < .001, R2 = .59. Alliance attitude was a significant predictor of the post-alliance attitude towards brand B, b = .21, t(237) = 5.33, p < .001, and when controlling for

no longer significant, b = .14, t(237) = 1.47, p = .142. Once again, looking at the bootstrapped unstandardized indirect effect, b = .12, 95% CI [.05, .22], proved that the indirect effect was significant. This means that, in line with the expectations, the effect of brand fit on post-alliance attitudes towards brand B was also fully mediated by alliance attitude. Hypothesis 4b can thus be confirmed.

Incongruity resolution

To verify if incongruity resolution did indeed have an impact on the effect of product fit and brand fit on alliance attitudes, a multi-way ANCOVA was run with product fit, brand fit and incongruity resolution as independent variables, alliance attitude as dependent variable and the aforementioned covariates.

In contrast with what was expected, no significant interaction effects were found for product fit and incongruity resolution, F(1, 232) = .07, p = .794, η2 = .000, nor for brand fit and incongruity resolution, F(1, 232) = .00, p = .957, η2 = .000. This means that, when a possible incongruity is resolved a high brand fit or a high product fit doesn’t lead to less favorable alliance attitude then a low product fit or brand fit.

To check if incongruity resolution moderated the effect of product fit and brand fit on the post-alliance attitude changes towards brand A and brand B, a mixed multi-way MANCOVA was conducted with product fit, brand fit and incongruity

resolution as independent, between-subject variables, the post-alliance attitude change towards brand A and brand B (consisting of pre-alliance and post-alliance attitudes) as within-subject variables and the aforementioned covariates.

Regarding brand fit, no significant interaction effects of incongruity resolution, brand fit, and post-alliance attitude change towards brand A, F(1, 233) = 1.35, p = .247, η2 = .006, and brand B, F(1, 233) = 3.65, p = .057, η2 = .015, were found. Incongruity resolution did thus not moderate the effect of brand fit on post-alliance attitude change towards brand A and brand B.

However, regarding product fit, a significant interaction effect was found of attitude change towards brand A, product fit and incongruity resolution F(1, 233) = 6.87, p = .009, η2 = .029. The means for pre- and post-alliance attitudes for brand A and B per low and high product fit can be found in Table 5.

Table 5. Estimated Marginal Means(SD) of pre- and post alliance attitudes for brand A and brand B per product fit (low vs. high), and Incongruity resolution (no/yes)

Attitude Brand A Attitude Brand B Incongruity product fit

Resolution pre-alliance post-alliance Diff.

pre-alliance post-alliance Diff. no low (n = 63) 5.21(.11) 4.50(.12) m.71 4.87(.10) 4.84(.10) m.03 high (n = 53) 5.20(.12) 4.72(.13) m.48 4.93(.11) 4.99(.10) .06 yes low (n = 62) 5.12(.11) 4.74(.12) m.38 4.87(.10) 4.99(.10) .12 high (n = 65) 5.34(.11) 4.52(.11) m.82 4.84(.10) 5.03(.09) .19 Note: Diff: difference score = post-alliance - pre-alliance attitudes. A more positive score equals a more positive attitude change

The interaction effect is depicted in Figure 2.

Figure 2. Line graphs of interaction effect of product fit, Incongruity Resolution, and post-alliance attitude change towards brand A.

As can be read from Figure 2 and Table 7, when the incongruity between two brands is resolved a low product fit leads to a more positive attitude change (i.e. less decrease) than a high product fit. A significant interaction effect of post-alliance attitude change towards brand A and product fit confirmed that this difference is significant, F(1, 121) = 6.87, p = .010, η2 = .054. This means that incongruity

so that a low product fit leads to a smaller decrease in attitude than a high product fit, when incongruity resolution is present.

For brand B, no such interaction with product fit and post-alliance attitude change was found, indicating that incongruity resolution does not moderate the effect of product fit or brand fit on the post-alliance attitude change towards brand B.

In conclusion, when incongruity resolution was present, a more positive effect was found for a low product fit, than for a high product fit on the attitude change towards brand A. This was in line with the expectations. However, incongruity resolution did not have any impact on the effect of product fit on brand B and the alliance attitude, and of brand fit on alliance attitude and the post-alliance attitudes towards brand A and B. Consequently, Hypothesis 6 can only be partially supported.

Discussion

This research aimed to expose the effects of product fit and brand fit in co-branding in the fashion industry on the attitude towards the alliance and the involved brands. The role of incongruity resolution was also investigated. From the results we can conclude that a high product fit does not lead to a higher alliance attitude than a low product fit, nor does it lead to a more positive attitude change towards the involved brands. Secondly, a high brand fit does lead to a more positive alliance attitude than a low brand fit, but a high brand fit does not directly lead to a more positive attitude change towards the constituent brands. However, the effect of brand fit on the post-alliance attitude towards the constituent brands is fully mediated by the attitude towards the alliance. Thirdly, brand fit is relatively more important than

product fit in bringing about a more positive attitude change, but only towards the parent brand. Lastly, when incongruity resolution is present, only a low product fit has a more positive effect than a high product fit on the attitude change, but only towards

the parent brand. As such, Hypotheses 3 and 4b are confirmed, Hypotheses 5 and 6 are partially confirmed, and Hypotheses 1, 2a, 2b and 4a are rejected.

Perhaps the most noticeable unexpected finding entails the absence of an effect of product fit, counter to the expectations and earlier co-branding research that has incorporated product fit (e.g. Simonin & Ruth, 1998; Geylani, Inman & Hofstede, 2008; Leuthesser, Kohli & Suri, 2003). This could have two possible reasons. Firstly, it could mean that in the particular case of co-branding in the fashion industry, product fit has lost its importance. Collaboration of fashion brands with non-fashion brands might have become so normal to people that it doesn’t affect their attitudes towards the alliance or towards the involved brands. Secondly, the absence of an effect of product fit could have something to do with the manipulations. A relatively new body of research has found that the relationship of product fit and attitude is non-monotonic rather than linear (e.g. Walchli, 2007; Ahn & Sung, 2012) so that co-branding alliances with a very high product fit are ‘too easy’ and alliances with a very low product fit are ‘too hard’ to process. Consequently, a moderate product fit would be most favorable. In this particular research, participants might have been frustrated by the pairing of a jeans brand with an electronic brand, and bored by the pairing of that jeans brand with another fashion brand. Adding the category of a moderate product fit might have produced the expected result. Furthermore, the absence of this third category could account for the fact that an interaction effect of product and brand fit on alliance attitude wasn’t found.

Secondly, the effect of brand fit on alliance attitude and on post-alliance attitude change, as mediated by alliance attitude, confirms the idea of brand fit being a very important factor in co-branding alliances in the fashion industry. Not only does a higher brand fit lead to more positive attitudes towards the alliance, this alliance attitude then seems to spill over to the attitudes towards the constituent brands. However, no direct effect was found when regarding attitude change as a repeated

case, one has to be wary in interpreting the results for attitude change, as pre- and post-alliance attitudes were measured with a relatively short time in between the measures, and no filler task or other distraction presented. Furthermore, a large proportion of the participants had guessed that the research was interested in investigating an attitude change of some sort, which might have influenced the results. In general, people prefer to be consistent in their answers (Guadagno & Cialdini, 2010), and this may have resulted in only a small attitude change, and hence no direct effect in the repeated measures analyses. However, when controlling for pre-alliance attitudes, brand fit did emerge as a significant predictor of

post-alliance attitudes.

Another unexpected finding was the decrease in attitude for the parent brand, after being exposed to the co-branding alliance, regardless of product and brand fit, whilst an increase of attitude was actually expected for both involved brands, dependent on seeing a high product or brand fit. This negative attitude change can be explained by the parent brand that was used in all conditions being a very strong brand that, for this sample, participants had more experience using (M = 4.63, SD = 1.62) than the partner brands (M = 3.55, SD = 1.89), scored higher on brand

familiarity (M = 5.32, SD = 1.24) than the partner brands (M = 4.00, SD = 1.71) and had a more positive pre-alliance attitude (M = 5.38, SD = .96) than the partner brands (M = 4.57, SD = 1.07). As such, the associations that people already held toward the partner brands transferred to the parent brand, making it less favorable. Previous studies indeed show that the associations of the parent brand can be deteriorated by co-branding, especially when partnering with more unfamiliar brands (Besharat & Langan, 2014). The other way around, for the partner brands, partnering with a brand that is stronger, more well-known and more well liked than the partner brands

themselves should leverage brand attitude and improve quality perceptions

(Besharat & Langan, 2014), which could be why in this research the attitude towards the partner brands wasn’t harmed.

The above might also explain why the effect of brand fit was more

pronounced than the effect of product fit, but only for the attitude change towards the parent brand. Seeing as how the parent brand was stronger than the partner brands, which could mean that held associations are considered more personal and relevant, incongruent (vs. congruent) stimuli might have a larger impact. As the associations for the partner brands were probably less strong, incongruent stimuli didn’t have a large effect and the effect of brand fit could not overrule the effect of product fit.

Finally, unexpectedly, incongruity resolution did not moderate the effect of product fit and brand fit on alliance attitudes, which means that when an incongruity between two brands was resolved, a higher brand or product fit didn’t lead to less favorable attitudes than a low brand or product fit. This also contradicts previous research (Sjödin & Törn, 2006; Walchli, 2007). However, the found effect can be explained as a consequence of the manipulations. Even though the manipulation for incongruity resolution was considered successful, the use of explanatory links might have been problematic in generating the expected effects. Because explanatory links were already provided to participants, this might have caused them to not look at the incongruity as if it was fully resolved by them, which in turn couldn’t lead to a sense of enjoyment. Without the enjoyment of incongruity resolution, it might not matter if the incongruity between two brands is resolved or not, the attitude towards the alliance remains the same.

In contrast, incongruity resolution did have an impact on the effect of product fit on the attitude change towards the parent brand. In this case, a low product fit might have caused an incongruity of such an extent, that an explanatory link was needed to make sense of the co-branding alliance. As such, incongruity resolution could actually contribute to producing a more positive attitude change. This probably only happened for the parent brand because, as explained earlier, incongruities have a larger impact on the parent brand, because it was a more familiar and more liked

personal. Consequently, incongruity resolution only helped product fit to bring about a more positive attitude change towards the parent brand.

While this study is partially in line with dictating co-branding theories, it might be useful to make some small adjustments concerning co-branding in the fashion industry. As this study found proof for the importance of brand fit, and didn’t for the importance of product fit, the relative importance of each type of fit might need revision in this particular domain. Also, the effect that co-branding has on the parent and the partner brand should be seen as two separates, as this study found different effect for each as a product of fit. Lastly, apparently incongruity resolution does not always have an impact on the effect of fit on alliance and brand attitudes. Following this research, the boundary conditions for such an effect can be further defined. Limitations and suggestions for future research

Despite this research being designed and conducted with the utmost of care, there are some considerable limitations of this study that have to be taken into account.

Firstly, as mentioned earlier, pre- and post- alliance attitudes were measured with a rather short time in between and no filler task or other distraction was

presented. This has caused people to be able to guess part of the purpose of the study, which could have biased the results in a way where participants were

deliberately trying to be as consistent as possible. Future research should take more care in measuring pre- and post-alliance attitudes, by allowing more time in between the measures and presenting participants with a distraction, so that the first measure will have less influence on the second. Alternatively, pre- and post- alliance attitudes could be measured as a between-subjects variable, by comparing attitudes in a pre-test by post-alliance attitudes in the actual experiment. Also, it might be interesting to measure the transfer of brand associations (i.e. ‘did the associations people hold towards one brand spill over to the brand it collaborated with?’), rather than an

attitude change, as the transfer of associations also forms one of the key reasons for a brand to collaborate with another brand.

Moreover, the use of familiar brands allowed for a high external validity, but also made the experiment and it’s manipulations less ‘clean’. With the use of familiar brands, participants had brought pre-existing brand attitudes, associations and beliefs to the study, which could influence the results in numerous ways. This also made it difficult to manipulate brand fit, as the idea of two brands matching or not is very subjective and strongly depends on the associations that the individual

consumer holds towards the brand. By using unfamiliar or fictional brands, in future research there can be more certainty that all other variables can be held constant. Extensive brand descriptions could then be made to give participants an idea of the brand and how it would possibly fit with another brand.

Additionally, the fact that incongruity resolution was manipulated rather than measured can form a thread to the enjoyment people feel when resolving an

incongruity themselves. As it is rather difficult to manipulate people into resolving an incongruity, without giving away the actual resolution, it might be inevitable that the effects of incongruity resolution diminish. In future research, it might be better to measure incongruity resolution. For example, participant’s Need for Cognition (i.e. a personal trait that can determine how much someone enjoys thinking and likes to solve ‘puzzles’, Cacioppo & Petty, 1982) can be used to determine if a possible incongruity will be resolved or not.

Lastly, the use of only one denim brand (i.e. Levi’s) as a parent brand means that the results of this study cannot necessarily be generalized to all other product categories. For instance, ‘denim’ might really be a niche, with different customers than other niches, and results might not be directly applicable to other branches of the fashion industry. It thus might also be interesting to look at the importance of fit and incongruity resolution in co-branding in the luxury, or fast fashion industry. Also,