Success drivers of co-branding: A meta-analysis

While considerable research attention has been given to co-branding (brand alliance), empirical evidence of the success drivers remains fragmented with inconclusive findings. This meta-analysis aims to synthesize the existing research and provide a comprehensive and generalisable set of findings. It integrates data of 197 effect sizes from 37 independent studies reported in 27 articles. The findings reveal that the relationship between the partner brands has a significantly larger impact on the success of co-branding than the individual brand characteristics, and brand image fit is a relatively more important driver than product category fit and brand equity. Moderator analysis indicates that the relative importance of the relationship between brands is generalisable to the type of industry, business and co-branding strategy. This paper advances theoretical understanding in three ways: (a) it increases generalisability of existing studies by investigating the impact of theoretical, contextual, and method-related moderators on the effect sizes, (b) it brings a consensus to the equivocal findings on the importance of success drivers and (c) it identifies the knowledge gaps, and presents a future research agenda. In so doing, the paper guides practitioners by highlighting which factors to be considered and prioritised when forming a brand alliance.


| LITER ATURE RE VIE W
The scope of this meta-analysis covers the co-branding strategy where two brands intentionally come together to form one joint product or service in order to enhance the combined assets of the brands. Based on the integration level of the two brands, there are two types of co-branding covered in the literature: ingredient branding and composite branding strategies (Helmig et al., 2008).
Ingredient branding, which is also known as vertical co-branding, entails the integration of a branded product (i.e. ingredient) within another brand as a component (Desai et al., 2014;Helmig et al., 2008;Radighieri et al., 2014)

. Notable marketplace examples of this include
Dell with Intel inside, Apple Watches with Hermes leather straps, and McFlurry ice cream with pieces of Oreo cookie. The main motivation behind the use of this strategy is to create differentiation via the ingredient's attributes and hence, enhance brand equity (Desai and Keller, 2002). In horizontal co-branding, a composite multibranded product or service is formed by more than one producer at the same step in the value chain, for example, Nike + (product of Apple and Nike; Helmig et al., 2008;Naidoo & Hollebeek, 2016).
Horizontal co-branding is based on the premise that a co-branded product or service inherits the tangible and intangible attributes of both partnering brands .
Prior researchers have measured the success of co-branding mostly by the attitude and behaviour intention towards co-branding.
Building on the theory of planned behaviour, both attitude and behavioural intentions are accepted as proxies for actual consumer behaviour (Ajzen & Fishbein, 1977). Attitude towards the co-branding entails the judgment and evaluation of the consumers' general feelings about the co-branded product (Garcia et al., 2017;James, 2006;Roswinanto, 2015;Senechal et al., 2014;Simonin & Ruth, 1998) whereas behavioural intention refers to the consumers' intention to purchase the co-branded product (e.g. Helmig et al., 2007;Mazodier & Merunka, 2014;Moon & Sprott, 2016;Rodrigue & Biswas, 2004); intention to recommend it to others (Ho et al., 2017) and willingness to pay (e.g. Rodrigue & Biswas, 2004). Since marketing literature investigates brand attitude and behavioural intention as the success parameters for co-branding practice (e.g. Garcia et al., 2017;Helmig et al., 2008;Simonin & Ruth, 1998), this meta-analysis studies both of these variables as the dependent variables.
Prior studies have not tested the comprehensive set of drivers and have rather focused on specific factors, neglecting the others.
For example, some studies have investigated only the relationship between 'fit among partner brands' and the 'evaluation of co-branding' (e.g. Yu et al., 2017), while others have analysed only the relationship between 'partner brand characteristics' and the 'evaluation of TA B L E 1 Literature review of co-branding success drivers
For example, Simonin and Ruth (1998) and Ahn et al. (2009) stated that brand fit had the highest impact on attitude and behavioural intentions (i.e. purchase intention), whereas Helmig et al. (2007) found that product fit was more influential on consumer evaluation. Senechal et al. (2014) showed that both product and brand fit were important antecedents of attitude towards co-branding, whereas pre-attitudes towards brands were less influential on cobranding evaluation. On the other hand, Singh et al. (2014) reported that neither product nor brand fit had an impact on consumer evaluation of co-branding, which contrasts with the findings of Simonin and Ruth (1998), Senechal et al. (2014), and Helmig et al. (2007). Ahn et al. (2020) demonstrated the significant influence of sensory fit on co-branding evaluation. Rodrigue and Biswas (2004) did not find any support for the influence of pre-attitudes on the positive attitude towards co-branding. Baumgarth (2004) and Lafferty et al. (2004) reported that co-branding evaluation was positively influenced by brand attitude and brand fit but did not find any support for the influence of product fit. On the other hand, Bouten et al. (2011) showed that product fit, brand fit, and pre-attitude towards brands all had a significant positive impact on the evaluation of co-branding, but that brand familiarity did not. While co-branding has been extensively studied in the last 20 years, there has been no research aggregating all the relevant studies on success drivers (Dalman & Puranam, 2017;Helmig et al., 2008).

| OVERVIE W OF THE CON CEP TUAL FR AME WORK
As a response to brands being perceived as the most valuable assets of companies pursuing commercial success (Blackett & Russell, 2000;Keranen et al., 2012), there has been extensive attention to branding studies in the literature. Despite the increasing academic interest in branding strategies, the extant literature remains fragmented and inconclusive at certain areas, which call for further meta-analysis or literature review. In order to advance the expanding branding literature, several review and meta-analysis studies have been performed on various branding topics such as corporate branding (Melewar et al., 2012), nation branding (Hao et al., 2019), place branding in the tourism industry (Gertner, 2011), branding in B2B (Keranen et al., 2012), branding strategy in the alliance of mass and luxury brands (Kumar et al., 2020), brand personality (Eisend & Stokburger-Sauer, 2013), celebrity endorsement of brands (Knoll & Matthes, 2017), brand extension (Völckner & Sattler, 2006) and co-branding (Besharat & Langan, 2014). Co-branding among other branding strategies has emerged as an attractive strategy that can enhance companies' existing brand equities. The conceptual paper of Besharat and Langan (2014) has summarised the equivocal findings in the literature, defining the areas of agreement, disagreement and the boundary conditions and has called for forming a consensus in the co-branding domain. This meta-analysis aims to find a resolution on the debate over the success factors of co-branding. Similar to the meta-analysis of Völckner and Sattler (2006) on brand extension strategy, the conceptual framework of this meta-analysis aggregates the success drivers that previous studies have identified as relevant and depicts the relationship between these success drivers and the consumer evaluation of co-branding, while investigating the potential differences under various moderators (Figure 1). The conceptual framework is built on the study of Helmig et al. (2008). Helmig et al., 2008 offer a conceptual review of co-branding literature, while highlighting the need for a meta-analysis that can synthesize the existing inconclusive findings and empirically test the relative importance of these success factors. Helmig et al. (2008) suggest studying the impact of brand characteristics, the relationship between brands and consumer-related variables on the economic outcome of co-branding.
This meta-analysis differs from the proposal of Helmig et al. (2008) in two ways: (a) it takes a consumer perspective and studies the impact of these antecedents on consumer evaluation of co-branding (attitude and behavioural intentions), (b) it extends it by identifying the effect of theoretical, contextual and method-related moderators.
The literature review reveals twelve drivers of co-branding success that were proven to be significant in at least one empirical study. Charlton (1996) does not recommend meta-analysis to be used for testing hypotheses, and states that the purpose of a metaanalysis is to obtain a more precise estimate of an effect, which is found in already existing hypotheses-testing studies. Therefore, this meta-analysis aims to synthesise the exiting research and bring a consensus on the inconclusive findings, while showing the dispersion of effect sizes under various moderators. Therefore, the study variables will be presented and defined without specific hypotheses, as done in previous systematic meta-analytical reviews that aimed to offer a consensus on the unsettled debate in the literature (e.g. Rosario et al., 2016). The potentially relevant success drivers that have been examined in at least one study in literature will be explained in detail in the following sections under three main categories: (a) brand characteristics; (b) relational characteristics between brands; and (c) consumer-related characteristic variables (Helmig et al., 2008;Völckner & Sattler, 2006).

| Brand characteristics
In the conceptual framework, brand characteristics refer to consumer perceived brand attitudes, perceived quality, brand equity, brand familiarity and brand trust for partner brands that form the co-branding (e.g. Baumgarth, 2004;Helmig et al., 2008;James, 2005;Lafferty et al., 2004;Senechal et al., 2014;Simonin & Ruth, 1998). These brand characteristics have been found to be positively correlated with the success of brand extension and co-branding (Helmig et al., 2008;Völckner & Sattler, 2006).

PAYDAS TURAN
More specifically, brand attitude, which is the positive or negative judgment or feeling held about a brand, influences consumer brand choice (Haugtvedt et al., 2008). Another assessment criterion of a product is its perceived quality, which is its superiority to the other alternatives (Dickinson & Heath, 2006;Keller, 1993).
Similarly, another factor that affects consumer information processing and brand evaluation is brand familiarity, which is the accumulated knowledge about a product (Lafferty et al., 2004;Naidoo & Hollebeek, 2016). Because of the extensive brand-related experience and knowledge, brand familiarity produces trusted cues related to the brands and positively affects purchase intention (Naidoo & Hollebeek, 2016). Moreover, brand trust, which reveals consumers' confidence in the brand to perform its promised function, has been shown to have a positive effect on consumers' purchase intention (Chaudhuri & Holbrook, 2001;Naidoo & Hollebeek, 2016). Brand equity, the value that is driven by consumer perception of the brand, is also studied as one of the success drivers in branding strategies since higher brand equity may positively affect the consumers' decision to buy certain brands over others (Arnett et al., 2010). Brand equity reveals the overall value of a brand, which is mainly a function of consumers' trust and confidence (perceptions) in the brand to deliver the expected performance as well as consumers' willingness to favour (behavioural responses) the brand over competing alternatives (Dutta and Pullig, 2011;Koschate-Fischer et al., 2019;Ma et al., 2018). Existing brand literature conceptualises the consumer-based brand equity as a composite of many consumer perception aspects, such as brand association, brand image, perceived quality, brand familiarity and brand awareness (Aaker, 1991;Keller, 1993), and also incorporates consumer behaviour aspects, such as preferences, loyalty, and purchase intention (Aaker, 1991;Agarwal and Rao, 1996).
The researchers in co-branding literature have extensively investigated the impact of one or more aspect of brand equity of partnering brands on co-branding evaluation (Ma et al., 2018;Yu et al., 2020).
Researchers have used information integration theory (Anderson, 1981) to show that the overall co-branding evaluation can be formed and modified by integrating the prior attitudes and beliefs about the partner brands and that these prior brand attitudes are positively related to the attitude towards the co-branding (Rodrigue & Biswas, 2004;Simonin & Ruth, 1998). Therefore, the individual brand characteristics of partner brands, which cover all the associations, attitudes, and beliefs about the brands, are to some extent related to the evaluation of co-branding.

| Relational characteristics between brands
'Relationship between brands' in the conceptual framework refers to the relationship characteristics between the partner brands, such as brand image fit and product category fit. Extensive research on brand alliance has focussed on the concept of fit between two brands, because the fit is an important factor in the evaluation of the alliance (Ahn et al., 2009;Bouten et al., 2011;Lafferty et al., 2004;Moon & Sprott, 2016). The fit concept has two dimensions: brand fit and product fit. Both brand fit and product fit refer to the compatibility of the two partner brands in co-branding (Ahn et al., 2020;Dickinson & Heath, 2006). Brand fit is defined as the congruence of consumer perceptions/associations about the brands, whereas product fit captures the consumer perception of the similarity and compatibility between two product categories (Ahn et al., 2020;Bouten et al., 2011;Ma et al., 2018;Senechal et al., 2014;Simonin & Ruth, 1998). While some researchers have demonstrated that the better the products fit, the easier it is for consumers to integrate and transfer their favourable attitudes to the co-branding (Bouten et al., 2011;Helmig et al., 2007), some empirical studies report that consumer evaluation of co-branding is positively influenced by brand image fit in particular (Baumgarth, 2004;Simonin & Ruth, 1998). Ahn et al. (2020) have examined the effectiveness of perceived sensory fit which have not been explored before. Sensory fit refers to the congruence of colour, shape or size of the partner products in a co-branding (Ahn et al., 2020). Congruity theory suggests that consumers seek to sustain F I G U R E 1 The conceptual framework for co-branding success drivers PAYDAS TURAN balanced and consistent associations among partners, and choosing the right partner is, therefore, one of the determinants of a successful co-branding; this notion is borne out by the literature (Ahn et al., 2009;Singh et al., 2014). The perceived fit between partner brands plays an important role in the evaluation of co-branding because it reduces the likelihood of a negative outcome from purchasing the co-branded product or service (Ashton & Scott, 2011). Therefore, this meta-analysis studies relationship between brands as a success factor for the evaluation of co-branding.

| Consumer-related characteristics
'Consumer-related variables' in the conceptual framework refer to consumer-specific characteristics, such as brand consciousness, brand involvement, self-congruity, variety seeking, and need for uniqueness.
For example, the brand involvement construct is found to be positively related to the consumer evaluation of co-branding because involvement with the partner brands triggers the appeal of co-branding (Mazodier & Merunka, 2014). Another consumer-related factor that affects the evaluation of co-branding is the brand consciousness of the consumer: brand-conscious consumers can identify brand names and they seek out well-known brands that signal a reduction in risk (signalling theory; Shim & Gehrt, 1996;Spence, 1973). Moreover, selfcongruity, which denotes the perceived fit between the brand and the consumers' actual and ideal selves (Aaker, 1999), has been studied as one of the consumer-related variables that are positively related to attitude towards the co-branding  and purchase intention (Mazodier & Merunka, 2014). Another examined consumerrelated variable in the literature is dialectical self, which refers to the degree of cognitive tendency to tolerate the inconsistencies in one's self-concept . Prior research in branding has investigated variety seeking and the need for uniqueness as consumerrelated variables (Helmig et al., 2007;Mazodier & Merunka, 2014).
Variety seeking consumers and consumers with a need for uniqueness derives utility from the change itself (Givon, 1984), and these consumers may favour co-branding since they prefer to switch first to alternatives for specific brands rather than switching to completely different brands (Helmig et al., 2007;Mazodier & Merunka, 2014). Based on prior research, this paper posits that consumer-related variables could affect the evaluation of co-branding.

| Moderators
The objective of having moderators in the conceptual framework is to identify how the relationship between the antecedents and the co-branding outcome differs under the influence of potential moderators. The substantive and methodological characteristics of studies may cause differences in the reported effects (Bijmolt & Pieters, 2001). Identifying such boundary conditions is particularly relevant when prior literature has produced inconsistent findings (Gonzalez-Mule & Aguinis, 2018). The moderators in this meta-analysis capture the theoretical, contextual, and research method-related differences among the studies, as is common practice in meta-analysis papers in marketing (e.g. Arts et al., 2011;Verbeke et al., 2011).

| Theoretical moderators
Horizontal versus vertical co-branding strategy A co-branding strategy can be applied horizontally or vertically, depending on how the products are integrated (Helmig et al., 2008).
Integration in co-branding refers to the level of the partner brands' connection in form and function (Newmeyer et al., 2014). In studies where the brands are highly intertwined in a form (vertical cobranding), the relational characteristics between the brands may be more important in the evaluation of the co-branding than they are in horizontal co-branding because a mismatch between highly integrated brands may increase the risks of negative outcomes for the co-branding (Ashton & Scott, 2011). Given that the level of integration may have different effects on consumers' evaluation of cobranding (Newmeyer et al., 2014), this meta-analysis examines the moderating role of the type of co-branding strategy.

| Contextual moderators
A co-branding strategy is based on the premise that brands collaborate to generate positive consumer evaluations, including the attitude and behavioural intention towards the co-branding (Besharat & Langan, 2014). The services sector has dynamics that are different from those of the goods sector, and the same applies to B2B versus B2C businesses (Verbeke et al., 2011). Analysing context-related moderators enables researchers to obtain empirical generalisations and assess how these contextual moderators vary the effect sizes (Rosario et al., 2016). As the types of industry and business have been shown to moderate the effect sizes in other marketing meta-analyses (e.g. Hogreve et al., 2017;Palmatier et al., 2006), this paper investigates whether the relative importance of success factors may vary depending on the business and industry settings.

Services versus non-services
Services and non-services (goods) have unique characteristics, which result in differences in the consumer evaluation processes (Zeithaml et al., 2006). In the case of services, the offer is mostly intangible and it is more difficult for consumers to evaluate its quality before experiencing it. On the other hand, in non-service goods, which are usually high in 'search' qualities rather than 'experience' qualities, consumers have the opportunity to be informed about the individual brands prior to the moment of sale (Verbeke et al., 2011). Signalling theory (Spence, 1973) explains the role that brands play in transmitting credible signals of product quality in the field of marketing (Helm & Ozergin, 2015;McCarthy & Norris, 1999;Rao et al., 1999;Spence, 1973). Consumers benefit from quality cues signalled by brands when they are evaluating attributes of services (Erdem & Swait, 1998;Zeithaml et al., 1985). Brand characteristics may play a more important role in the context of services due to the need to reduce the uncertainty that accompanies an intangible offer (Naidoo & Hollebeek, 2016). Since service and non-service industries have different characteristics that change the consumers' evaluation process, it is relevant to find out how industry type moderates the impact of the success drivers.

B2C versus B2B
The B2B and B2C markets have different dynamics (Hogreve et al., 2017;Verbeke et al., 2011). This paper explores whether the correlation between the drivers and the level of success varies depending on the business setting. Business customers in B2B settings are involved in decision-making processes that are more complex and rational than those of the end-consumer (Manning et al., 2010;Verbeke et al., 2011). Based on information integration theory (Anderson, 1981), people integrate attitudes and beliefs about partner brands to form an evaluation of the co-branding (Rodrigue & Biswas, 2004). The complexity of the decision process for business customers may mean that consumer-related variables might moderate the assessment process differently in B2B versus B2C markets.

Fictitious versus real brands
In some studies, the co-branding product or service is 'real', which means it exists in the market, whereas other studies analyse the success of co-branding with 'fictitious' brands. Both methods have their merits. Researchers use fictitious brands to avoid any brand associations that could bias the participants' evaluations due to previous experience with real brands (Bleijerveld et al., 2015;Geylani et al., 2008;Rao et al., 1999). In contrast, real brands are used so that genuine brand perception change can be activated by exposure to the manipulation (Brady et al., 2008). The evaluation of co-branding may vary based on whether the partner brands are real or fictitious, and it may be that this research method-related moderator varies the effect size.

Sample types
Scholars in social science research have discussed the use of a participant sample entirely made up of students and questioned its representation power of the heterogeneous population (Ashraf & Merunka, 2016;Kardes et al., 2007). Some researchers claim that studies conducted with a student sample may produce larger effect sizes because there is less error variance in measurement due to the homogeneity of student samples (Orsingher et al., 2009;Peterson, 2001). Therefore, the moderating effect of the sample on the effect sizes is explored in this meta-analysis.

Publication date
The year of publication date is often a moderator in meta-analyses performed in the marketing literature (e.g. Bijmolt et al., 2005). The marketing discipline has experienced tremendous change over the last two decades, triggered by changes in the nature of consumers and markets (Kumar, 2005). During this time, the vast number of marketing messages being constantly transmitted means that attracting consumer attention to a new product and service is becoming increasingly challenging (Teixeira, 2014). Therefore, this meta-analysis codes the publication year for all studies as 'before' and 'after' the year 2000 to observe whether the magnitude of the effect size varies over time (Chang & Taylor, 2016).

| ME THOD
It is important to specify the type of this review paper since each review paper has different purposes and approaches that require them to be evaluated differently. These are structured review focusing on used methods, theories and constructs (e.g. Gilal et al., 2019;Mishra et al., 2020); framework based reviews (e.g. Paul & Benito, 2018); narrative reviews with a framework for setting future research agenda (e.g. Dabić et al., 2020;Kumar et al., 2020) and systematic meta-analytical review to aggregate results from different studies to identify the disagreements, and show the effect sizes that differ in various contexts (e.g. Barari et al., 2020;Rosario et al., 2016).
Among all these review methods, this research adopted a systematic meta-analytical review to meet the objective of synthesising existing previous researches on success factors of co-branding to resolve inconsistencies and create a generalised accumulative knowledge on success drivers of co-branding (Paul & Criado, 2020).

| Data collection
Data collection consists of an elaborate search followed by a screening process with exclusion and inclusion criteria. To identify a population of relevant studies on the evaluation of co-branding, an elaborate search strategy was conducted following the guidelines of Paul and Criado (2020). Qualitative research on co-branding has started to take place at the beginning of the 1990s (e.g. Norris, 1992;Rao & Ruekert, 1994), and the first quantitative research on co-branding (e.g. Park et al., 1996;Shocker, 1995) did not begin before 1995 (Helmig et al., 2008;Singh et al., 2016). Therefore, the search aimed to retrieve all the relevant empirical research on cobranding success drivers during the period 1995-2020. As depicted in Figure Figure 2. After these two steps, the non-academic studies, qualitative papers, and the studies that investigated the benefits of the strategy, structural dynamics of forming a co-branding and performance of it from firms' perspective, and the studies that mentioned co-branding but analysed other branding strategies were excluded.
Before applying the inclusion criteria, the remaining 121 fulltext papers were read and coded according to the coding manual (Appendix A1). A more detailed assessment of this subsample enabled identifying the papers that are relevant to the conceptual framework of this meta-analysis that aimed to study the success drivers of co-branding. The inclusion criteria were that the article (a) measured consumer evaluation of co-branding, and (b) reported one or more empirical success driver (e.g. brand equity, brand image fit, product category fit) as the antecedents. Despite the existence of a broad range of studies on co-branding in the literature, this systematic meta-analysis has a selective approach and aims to synthesise only the success factors of co-branding to resolve inconsistencies and create a generalised accumulative knowledge on antecedents of co-branding success.
Systematic meta-analytical reviews, unlike narrative reviews, always have an eligibility criterion that consists of inclusion and exclusion steps that are determined before the search is implemented.

PAYDAS TURAN
It is common to start with a large pool of studies and end up with a much smaller set of studies after the eligibility criterion is applied along with the author's knowledge and judgement (Borenstein et al., 2009;Paul & Criado, 2020).
At the end of this inclusion process, the final sample of this metaanalysis consisted of 37 independent studies from 27 articles that delivered a total of 197 effect sizes (N = 10,862). The detailed data collection process with exclusion and inclusion criteria is shown in Figure 2.

| Data coding
All necessary study characteristics and the effect sizes were coded based on a coding manual (Lipsey & Wilson, 2001; see Appendix A1).
The studies were coded in four steps: (a) searching for the antecedents and the outcome relevant to the conceptual framework; (b) searching for the effect sizes and study descriptors; (c) entering all data into an Excel sheet; and (dd) classifying and grouping variables (see Appendix A2). To check for coder reliability, two coders coded four of the articles simultaneously. Having an overall agreement rate of over 90% in these studies implied no significant discrepancies (Verbeke et al., 2011). Coded papers are presented at the end of the paper in Section 8. As in previous meta-analyses, items were merged when the differences between them were not meaningful to meta-analysis purposes such as brand 'fit' and brand 'match up' (Verbeke et al., 2011).
Another reason to merge one or more predictors into a subvariable is to increase the power of the effect sizes and to yield a more precise estimate of the effect than would be possible with individual studies examining the limited number of variables (Borenstein et al., 2009). In the context of co-branding, there are limited studies examining some of the variables such as brand equity, trust, perceived quality, brand familiarity and trust. Since brand equity, as defined in the literature, is a composite of these variables, they are merged and coded under 'brand equity' to be able to yield a more precise estimate of the effect of brand equity on the success of co-branding. Comparing effects in subgroups can have very low power if there are so few studies examining a specific variable such as brand involvement, variety seeking, self-congruity (Borenstein et al., 2009). Therefore, the variables that are examined in less than five independent studies are dropped and not studied as sub-group variables in this meta-analysis. These variables are only analysed under main categories such as brand characteristics, relational characteristics and consumer-related variables (e.g. sensory fit, dialectical self).

| Effect-size metric
The studies included in the analysis measured the relationships between antecedents and the outcome variable by means of correlations. Thus, the correlation was used as the effect-size metric, which is consistent with other meta-analyses in marketing (Arts et al., 2011;Kirca et al., 2005;Palmatier et al., 2006;Rosario et al., 2016;Verbeke et al., 2011). The effect sizes of studies that used regressions or reported standardised beta coefficients (e.g. Bouten et al., 2011;Garcia et al., 2017) were converted into correlations with a formula suggested by Peterson and Brown (2005).

| Random-effects analysis
The random-effects model was adopted (Hunter & Schmidt, 2004). This model assumes that the studies in the meta-analysis are selected from a universe of studies meeting the inclusion criteria and that the effect sizes are a random selection of the effect sizes in the universe (Borenstein et al., 2009). The researcher accumulated data from a series of studies that had been conducted by other researchers, and it would be unlikely that all these studies were functionally equivalent. Because the subjects and interventions in these studies would have differed in ways to change the impact on the results, the researcher should not assume a fixed effect size (Borenstein et al., 2009;Lipsey & Wilson, 2001).
Therefore, the researcher adopted a random-effects model for this meta-analysis.
Comprehensive Meta-Analysis Software (CMA) was used to report the key statistics, such as the summary effect, confidence intervals, and measures of heterogeneity (Borenstein et al., 2009).
To correct for sampling errors, studies' sample sizes were used as weights, and 95% confidence intervals were constructed to determine the significance of the mean correlations (Borenstein et al., 2009). CMA enabled working with multiple independent variables and two dependent variables.

| Level of analysis
The individual effect sizes were used as the unit of analysis. This was justified by a Q statistics test for heterogeneity, which indicated significant heterogeneity of correlations (Verbeke et al., 2011

| Heterogeneity
The researcher analysed the dispersion of effects by calculating the

| Meta-regression analysis
To test the model simultaneously with all the covariates, a metaregression was performed using the CMA software (Borenstein et al., 2009). First, the dummy coded categorical variables were incorporated as the covariates and then the impact of the covariates with a sufficient number of studies accepted by CMA was assessed (Borenstein et al., 2009).

| RE SULTS
This meta-analysis aggregates the existing research by calculating the average correlations for two success drivers of co-branding evaluation: 'positive attitude' and 'behavioural intention towards co-branding'. Table 2  The results for the first outcome variable, attitude towards cobranding, show that both brand characteristics and the relationship between brands are positively correlated with the attitude towards co-branding. The bivariate data in Table 2 indicates significant antecedents and outcome associations: the correlation for both brand characteristics (r characteristic = 0.257, p < .001) and relationship between brands (r relationship = 0.411, p < .001) is positive and significant. When compared to pairwise, the correlation of the relationship between brands is significantly larger than the correlation of brand characteristics (p < .05). By aggregating all the studies, this paper offers a robust analysis of all the factors and concludes that the importance of the relationship between brands is significantly higher than that of individual brand characteristics.
For the next outcome variable, behavioural intention, the bivariate data indicates the following significant associations between the success drivers and behavioural intention: brand characteristics (r characteristic = 0.379, p < .001); relationship between brands (r relationship = 0.465, p < .001); consumer-related variables (r consumer = 0.199, p < .001), and attitude towards co-branding (r = 0.606, p < .001). A pairwise comparison proves that the correlation of consumer related variables and attitude towards co-branding is significantly less than the correlation of attitude towards cobranding with both brand characteristics (p < .05) and relationship between brands (p < .001).
Bivariate analysis depicted in Table 3 shows that success drivers have significant (p < .001) effects on attitude towards co-branding, in the following order of magnitude: brand fit (r = 0.471), product fit (r = 0.342), and brand equity (r = 0.257), where brand equity is a composite variable of brand attitude, perceived quality, brand familiarity and brand trust.
The heterogeneity test reveals that Q values for all the relationships presented in Table 2 are significant (p < .001), indicating that the true effect size is not the same in all the studies and hence, there is heterogeneity. I-squared values, being all over 70% also confirm that there is heterogeneity and indicate that a moderator analysis is worthwhile to explore the dispersion of true effect sizes, such as has been performed in other meta-analyses (e.g. Barari et al., 2020;Bijmolt & Pieters, 2001;Borenstein et al., 2009).

| Results of moderator analyses
The studies were classified based on a moderator variable to compare the effect sizes.  There is a moderate correlation of relationship between brands and attitude towards co-branding in both the vertical and the horizontal co-branding strategy, whereas the correlation of brand characteristics and attitude towards co-branding is low in both.

| Service versus non-service
The results for the industry setting, which is a contextual moderator, reveal that the effect of the relationship between brands on attitude towards co-branding is significantly stronger for the non-service sector (p < .05).

TA B L E 2
Meta-analysis results for positive attitude and behavioural intention towards co-branding Abbreviations: ATTC, attitude towards co-branding; CI, confidence interval with lower (LL) and upper limit (UL); fail-safe number attenuated at 0.05; I 2 , I 2 statistic; k, number of effect sizes; Q, Q statistic; r, sample size-weighted mean correlation coefficient; SE, standard error. a p < .05. b p < .001.

| B2C versus non-B2C
The results for the contextual moderator, business setting, show that the correlation of success drivers and the attitude towards co-branding does not vary depending on the business setting.
In both business settings, there is a moderate correlation of the relationship between brands and attitude towards co-branding, (r relationship = 0.410 and r relationship = 0.419 in B2C and non-B2C business settings, respectively). Moreover, in both business settings, the relationship between brands has a higher correlation than brand characteristics.

| Student sample versus non-student sample
The sample that was used in the studies varies the effect sizes.
The correlation of brand characteristics and attitude towards cobranding is significantly stronger (p < .001) in studies with a student sample than in those conducted with a non-student sample.

| Year of study: after 2000 versus before 2000
The relationship between brands shows a stronger effect after 2000 than before 2000 (p < .001).

| Published versus non-published studies
The results show that the effect sizes do not vary significantly in published and non-published studies. This is an indicator of the absence of publication bias in this meta-analysis. Table 5 provides the results of the meta-regression, testing the influence of all moderators together on the effect size for attitude towards co-branding. The aim of performing a meta-regression is to investigate the effects of multiple factors simultaneously. However, the use of meta-regression with multiple covariates is not recommended when the number of studies for each covariate is small (< 10) because of the lack of power to explain variable relationships (Lipsey & Wilson, 2001). The meta-regression therefore was performed for the dependent variable, attitude towards co-branding (Borenstein et al., 2009). Table 5, the composition of the branding strategy (horizontal vs. vertical) and type of sector (non-service vs. service)

As shown in
were the covariates that were found to be significant (p < .05) in TA B L E 4 Meta-analysis results for theoretical, context-related and method-related moderators Abbreviations: ATTC, attitude towards co-branding; BC, brand characteristics; CI, confidence interval with lower (LL) and upper limit (UL); failsafe number attenuated at 0.05.; k, number of effect sizes; N, combined sample size; n.a., not available; r, sample size-weighted mean correlation coefficient; RC, Relational characteristics between brands. a p < .05.

PAYDAS TURAN
identifying the dispersion of effect sizes in individual brand characteristics and relational characteristics between brands. The model is able to explain 27% of the variance in true effects. Since these are categorical variables, the coefficients give the mean differences in effect sizes for studies on horizontal co-branding vs. vertical co-branding, and studies conducted in non-service businesses vs. service businesses, respectively. In both models, the correlation of brand characteristics and attitude towards co-branding and correlation of relationship between brands and attitude towards co-branding is significantly stronger for the horizontal co-branding practices and also for the non-service sector (p < .05).

| Publication bias test results
In order to minimize publication bias, which occurs when the pub- to be a reason for concern (Borenstein et al., 2009). Therefore, the author concludes that there is no publication bias. This is consistent with the moderator analysis that indicated that the effect sizes do not vary significantly in published and non-published studies.

| Theoretical contribution
Although some researchers have quantitatively studied the success factors that drive the positive evaluation of co-branding, there have been recent calls to aggregate the factors and examine their relative importance (Helmig et al., 2008;Newmeyer et al., 2014). This paper responds to these calls by providing the first comprehensive metaanalysis of co-branding success factors.

PAYDAS TURAN
First, this research provides a robust analysis of all the studied success factors that lead to the positive evaluation of co-branding and brings consensus to the fragmented literature of co-branding. By aggregating all the success factors that have been studied in the literature, this meta-analysis has been able to assign the subgroup variables to one of three categories: brand characteristics, the relationship between brands and consumer-related variables.
By identifying each category's magnitude of importance, the meta-analysis provides robust testing of the conceptualisation of these success factors. The relative importance of these factors is shown in Table 6. The meta-analysis shows that it is the relationship between partner brands rather than the individual characteristics of the brands that have a greater impact on the success of co-branding. While previous literature identifies separate fit factors, such as brand fit and product fit, and separate individual brand characteristics, such as brand attitude, perceived product quality, brand equity and brand familiarity, this paper draws attention to the distinguishing aspects of co-branding, namely the individual characteristics of brands and the relationship between the brands. The dyadic relationship between brands plays a key role in the evaluation of co-branding. As brand collaboration becomes an increasingly popular strategy in the fiercely competitive marketplace, choosing the correct partner becomes one of the critical determinants of the strategy's success (Ahn et al., 2009;Phelps, 2016;Singh et al., 2014). Therefore, brands must focus their attention on finding the right partners for them, the right partner in this case is one with whom they have a high degree of fit.
Moreover, the findings of this research address and align prior equivocal findings on particularly the importance of brand image fit and product category fit to the evaluation of co-branding. Brand image fit is more important than product fit to a positive attitude and behavioural intention towards co-branding. In this meta-analysis, bivariate analysis for the sub-group variables shows that the sequence of success drivers in terms of the magnitude of effect on attitude towards co-branding is as follows: brand fit, product fit and brand equity (composite of brand attitude, perceived quality, trust and familiarity). This meta-analysis thereby answers the call to examine the relative importance of the co-branding success factors. The results find consensus from equivocal findings on the importance of brand fit and product fit on the co-branding evaluation. Brand fit is a success factor that is relatively more important to a positive attitude and behavioural intention towards co-branding when compared to the product fit. Therefore, this paper finds concludes the congruence of perceptions and associations of consumers about the partner brands is more important than the degree of similarity and compatibility between the partner product categories.
Secondly, by aggregating the relevant empirical studies, this meta-analysis moves the discussion away from individual contextbased studies towards one that is more generalisable. It addresses the inconsistency in prior findings by investigating the impacts of theoretical, contextual, and method-related moderators on the effect sizes. The findings of the meta-analysis reveal that the relationship between partner brands has a significantly larger impact on the success of co-branding than that exerted by the individual brand characteristics of partner brands, and this is generalisable to every type of co-branding strategy, business, and industry type.

| Theoretical moderators
Both vertical (e.g. Baumgarth, 2004;Moon & Sprott, 2016;Simonin & Ruth, 1998) and horizontal co-branding (e.g. Ashton & Scott, 2011;Naidoo & Hollebeek, 2016;Roswinanto, 2015) have been investigated in the co-branding literature (although the type of the branding strategy was not specified in some cases). The moderator analysis reveals that the correlation between success factors and attitude towards co-branding does not vary with the co-branding strategy adopted. In both strategies, the correlation of the relationship between brands, which consists of brand fit and product fit, is significantly higher than the individual brand characteristics of the partner brands. This meta-analysis proves that the findings are generalisable to both co-branding strategies, despite the differing levels of integration between brands.

| Contextual moderators
This meta-analysis studies type of industry and business type as contextual moderators. The effect of the relationship between brands on attitude towards co-branding is significantly stronger for the nonservice sector (p < .05). This is presumably because, in services, consumers take the internal coordination between partner brands for granted since relational coordination is a mutual interaction process between the service providers who are integrated with the task and it is positively related to internal service quality and the customer outcome perceptions (Gittell, 2002). suggestions (Shadish & Haddock, 2009), there was no significant difference in the effect sizes when the studies were conducted with real versus fictitious brands. Finally, the correlation of brand characteristics and behavioural intention towards co-branding is significantly stronger (p < .05) in studies with a student sample versus studies with a nonstudent sample. This is consistent with prior meta-analyses that show that the effect sizes derived from student samples often differ from the effect sizes in non-student samples (Peterson, 2001;Roschk et al., 2017). This finding might encourage researchers to not overly rely on students as study participants and to take the necessary steps to ensure that the study can be generalised when a student sample is used.

| Method-related moderators
Thirdly, while this meta-analysis elevates the discussion from one that is concerned with many individual studies towards one that gives an overview of a broader stream of research (Borenstein et al., 2009), it also identifies the gaps in the co-branding literature and presents an agenda for future studies in the field, which will be discussed in detail in Section 7. The results of this meta-analysis highlight the paucity of empirical studies in various industries, such as services and non-B2C, while also pointing out the lack of attention that has been given to some variables, such as consumer-related variables and actual purchasing behaviour.

| Managerial contribution
Paul Parisi, the former President of PayPal Canada, claims that companies collaborate and combine their strengths to deliver greater consumer experiences (Parisi, 2017). In practice, co-branding successes and failures abound. In order to achieve the requisite degree of positive attitude and behavioural intentions in consumers, it is essential that managerial decision makers can identify the success drivers of co-branding. This meta-analysis with its robust results can guide marketing managers who are considering adopting a cobranding strategy by highlighting which factors must be considered and prioritised. The managerial contribution of this research is twofold: it highlights the importance of the relationship between brands over individual brand characteristics in the evaluation of cobranding, and it recommends that managers acquire understanding about the type of industry and business in which they operate before they embark on a co-branding strategy.
First, this meta-analysis contributes by informing managers about the importance of fit between brands; this is vital information if managers are to correctly prioritise their search criteria for the right partners.
The findings show that the correlation between the relationship between brands and co-branding evaluation is significantly more important than the correlation of brand characteristics. This indicates that a brand with relatively high brand equity should not necessarily use its resources to target a potential partner for its specific brand, such as high brand equity; instead, it needs to ensure that there is an adequate fit between the partners. Co-branding, by its nature, requires some interdependence between partner brands (Blackett & Boad, 1999;Newmeyer et al., 2014). In order to be successful, managers need to manage the co-branding as if it were a single brand, and to be able to do that, they must ensure that the brands fit well with each other. The managerial findings of this meta-analysis, summarised in Table 6, call for deeper consideration of the relationship between partner brands in terms of brand fit and product fit. The sub-category analysis reveals that brands' image fit is more important than the fit between product categories. Specifically, brand image fit has the highest correlation with co-branding evaluation out of all the variables. The potential partner might have high brand equity, but if the two brands' images do not fit, there is little chance that the consumers will positively evaluate the cobranding and it is unlikely to become a commercial success. Comparing the relative importance of brand image fit versus product category fit, brand image fit needs to be assigned a higher weighting in the partner selection and decision-making process. If the partner brands succeed in creating a seamless logic out of the combined offer, consumers can perceive the benefits of the offer that the brands aim to deliver, and hence, will positively evaluate the co-branding.
Secondly, this meta-analysis contributes by recommending managers to take into account the impact of the type of industry and business in which they operate. The findings indicate that the correlation of relational characteristics between brands and attitude towards co-branding is significantly higher in the non-service industries than in service industries. In the context of service industries versus nonservice industries, consumers benefit significantly more from brand characteristics that act as quality signalling cues in forming positive towards the co-branding (Rao & Ruekert, 1994). For that reason, managers who are adopting a co-branding strategy in services must pay close attention to the brand characteristics of the partner brands. Managers in the service sector should also ensure that they associate their brand with a partner brand from which they can derive incremental benefit.
Moreover, this meta-analysis shows that the type of business does not vary the results, and that in both B2C and non-B2C contexts, the relationship between brands has a higher correlation than brand characteristics. Therefore, the finding that the relationship between brands is the co-branding priority is generalisable to every business type.
Having analysed the theoretical, contextual, and method-related moderators, this meta-analysis concludes that the relationship between brands has a higher correlation than individual brand characteristics on the success of co-branding. This result is generalisable to different types of co-branding strategies (vertical vs. horizontal), businesses (B2C vs. B2B), and industries (service vs. non-service).
By knowing the magnitude and direction of the relationship of the success drivers of co-branding, marketing managers can implement effective co-branding strategies.

| Limitations
The aim of this paper is to aggregate studies on co-branding in order to analyse and consolidate knowledge of the strategy's success drivers. However, it has limitations that are endemic to meta-analyses.
First, meta-analysis papers are often criticised for compiling findings from studies of different quality (Leonidou et al., 2002). This paper includes, to the best of the author's knowledge, all the relevant studies that provide effect sizes. In order to be inclusive and avoid publication bias, this paper includes relevant papers that study co-branding success drivers, whether they are published and unpublished (e.g. doctoral dissertations) as it is the norm in systematic meta-analyses that aim to reach a consensus on the equivocal findings in the literature (e.g. Rosario et al., 2016;Schmidt & Bijmolt, 2020).
Second, although the review of the literature was systematically and comprehensively carried out, making every effort to cover all the literature on co-branding success factors, meta-analysis studies inevitably run the risk of overlooking relevant articles. To the best of the author's knowledge, no such paper has been overlooked in this study. That being said, the literature review uncovered a vast number of co-branding studies, some of which adopt the perspective of the firm rather than the consumer (e.g. Gammoh, 2006), or do not address the relevant conceptual issues such as the papers about crisis strategies in a co-branding context (e.g. Singh & Crisafulli, 2020). Consistent with the implementation of eligibility criteria in other meta-analyses (e.g. Frigerio et al., 2020;Rana & Paul, 2020), the studies that do not meet the criteria are excluded from the metaanalysis. Based on the eligibility criteria presented in Section 4.1, this meta-analysis is performed with studies examining the antecedents of co-branding success as depicted in the conceptual framework in Section 3.
Third, another literature-imposed limitation of a meta-analysis is that some variables or moderators cannot be analysed or are left under-represented due to an insufficient number (or even complete absence) of papers investigating those variables (Knoll & Matthes, 2017;Rana & Paul, 2020). In the case of the literature on co-branding success factors, there is an insufficient number of papers on some of the variables such as variety seeking (Mazodier & Merunka, 2014), brand involvement (Moon & Sprott, 2016), dialectical self , retail channel (e.g. Yu et al., 2020) and sensory fit (e.g. Ahn et al., 2020). These variables were merged together under the relevant main categories in the meta-analysis (e.g. the sensory fit is grouped under 'relational characteristics between brands'). Any variable that is under-represented in the literature (being examined in less than five studies) and could not be grouped under the main categories (e.g. brand characteristics, relational characteristics and consumer-related characteristics) is excluded from the meta-analysis in order to avoid the mean effect size lacking statistical power (Borenstein et al., 2009). Due to the underrepresentation of some industries and business types in the literature of co-branding (i.e. service industries and non-B2C businesses) and the limited number of studies conducted for some variables (i.e. consumer-related characteristics) the specified moderators leave some variance unexplained. However, the conceptual framework and the number of moderators examined can be expanded as more studies accumulate in the literature. Despite these limitations, this paper contributes to a better understanding of the success drivers of co-branding.

| FUTURE RE S E ARCH D IREC TI ON S
In addition to yielding valuable information about the success drivers of co-branding, this meta-analysis identifies gaps in the literature and exposes several avenues for further research as depicted in Table 7. Such knowledge gaps become visible to researchers when they find they cannot analyse some of the variables or moderators due to the limited number of studies that have the relevant data set in the literature (Frigerio et al., 2020;Knoll & Matthes, 2017;Rana & Paul, 2020). Therefore, this meta-analysis encourages further research to focus on the following issues: (a) understudied antecedents; (b) additional outcome variables; (c) additional moderators; (d) methods other than those used in prior research; and (e) the underlying psychological mechanisms.
First, this paper calls for further research to study the additional or understudied antecedents that may affect the consumer evaluation of co-brandings, such as the marketing context and consumerspecific variables. Future research might investigate marketing support and retailer acceptance as the marketing context variables; these were studied in the meta-analysis of Völckner and Sattler (2006) and found to have a direct effect on brand extension. The level of marketing support, such as the advertising budget or the firm's marketing competence, might play an important role in consumers' evaluation of co-branding.
Similar to the effect of marketing support, retailer acceptance might be as important to co-branding success as it is to brand extension. If retailer acceptance of co-branding is high, the retailer is likely to give more exposure to the co-branding, increasing its visibility.
Given that the effect size of the marketing context on the success of a brand extension has been found to be high (Völckner & Sattler, 2006), it is worthwhile studying this as a potential antecedent in future studies on co-branding.
Next, this meta-analysis on the success factors of co-branding reveals that consumer-related variables have not attracted enough attention in the studies measuring attitude towards co-branding.
It has been shown that consumer-related characteristics might be important explanatory variables in branding studies (Barone et al., 2000;Czellar, 2003). Consumer innovativeness and risk acceptance level are two of the consumer-related characteristics that, although neglected in the co-branding literature, have been examined in other branding studies (Klink & Daniel, 2001;Nijssen, 1999). Therefore, it would be interesting to incorporate these consumer-related variables into the studies of co-branding, where two brands are brought together on purpose to create additional value (Blackett & Boad, 1999 if studied in a more comprehensive meta-analysis (Simonin & Ruth, 1998). This meta-analysis takes a consumer perspective, whereas a company perspective could equally as well have been taken. Helmig et al., and and's (2008) framework suggest studying the economic outcome of co-branding as the dependent variable, which would require incorporation of the sales profit and return on investment. The models covering the firm's perspective could be very informative for both academics and practitioners (Helmig et al., 2008). By focusing on these understudied outcome variables, future research could expand this meta-analysis to derive a more holistic understanding of co-branding.
Third, researchers are encouraged to conduct further research on understudied moderators, both theoretical and contextual. This meta-analysis reveals that one of the theoretical moderators, vertical co-branding or ingredient branding, has not attracted as much attention in the literature as horizontal co-branding. Further empirical research into vertical co-branding (ingredient branding) would contribute to the overall knowledge about co-branding strategies (Moon & Sprott, 2016). This meta-analysis reveals a paucity of studies exploring some contextual moderators that have been included in the meta-analysis, such as the industry and business type. It shows that prior research has mainly concentrated on fastmoving consumer goods (FMCG) in B2C businesses, and therefore this meta-analysis includes comparatively more studies from FMCG in B2C settings. However, the dynamics of the service and nonservice markets are different, as are the decision-making processes of customers in B2C and non-B2C businesses (Hogreve et al., 2017;Manning et al., 2010;Verbeke et al., 2011). Future research could focus on various service industries and non-B2C businesses to get a better understanding of co-branding success factors in various contexts. Another research context that has not received any attention is that of the online alliance. Given the growing number of digital alliances in practice (e.g. Spotify and Starbucks), the online context would benefit from academic attention (Jayawardhena et al., 2007;Singh et al., 2016). Further research into co-branding in an online context could help inform managers who are navigating digital co-branding, which will enable them to gain a competitive advantage. These understudied moderators are symptomatic of the generalisability issue of the findings in the current literature.
Moreover, it is possible that some of the success drivers jointly and synergistically influence consumers' evaluation of co-branding.

Research agenda
Antecedents How does the marketing context (e.g. advertising, price) affect the evaluation of co-branding? What impact do retailers exert on the success of co-branding? How do the innovativeness and risk acceptance of consumers affect the evaluation of co-branding? How does the sensory fit between brands impact the evaluation of co-branding relative to the product in various industries?
Outcome variables How do the attitude and behavioural intention towards co-branding relate to its actual sales and profit? How do associations transfer between brands? How do the partner brands elevate the positive spillover effects? How do the partner brands mitigate the risks of negative spillover effects?

Moderators
How do the different characteristics of services and goods impact consumer evaluation of co-branding? What are the success drivers in the evaluation of vertical co-branding? What impact do the differing characteristics of B2C and B2B have on the consumer evaluation of co-branding? What are the success drivers in the evaluation of digital co-branding? How generalisable are the findings in the current literature?

Method
What other methods, other than surveys, could be used to measure actual consumer behaviour rather than intentions? What other methods, other than surveys, could be used to explore the underlying mechanisms in consumers' perception of fit between the brands? What other methods could be used to investigate change in attitudes and behavioural intention towards a co-branding over the life cycle of the co-branding? How generalisable are the findings of studies conducted with a student sample in co-branding literature?

Underlying psychological processes
What are the underlying psychological mechanisms for perceiving 'brands fit'? What are the underlying cognitive processes behind the transfer of associations between the brands in a co-branding? How do consumers relate the different positioning of the individual brands to the success of the newly launched cobranding offering?

PAYDAS TURAN
insights on the nature of co-branding success and supplement the attitude and intention studies that currently dominate the literature. It is accepted that an individual's behaviour intentions do not always translate into real behaviours. Therefore, collecting actual field data would contribute to the understanding of co-branding success and offer practical implications (Dalman & Puranam, 2017;Nisbett & Wilson, 1977). Although the use of cross-sectional data is understandable due to limited resources, it does not allow for a change in attitude towards co-branding over different stages of its life cycle to be taken into account. Research that is more exploratory and qualitative could be used to shed light on the cognitive processes that consumers go through when exposed to co-branding.
The use of qualitative research in the field of co-branding could help explain the underlying psychological processes in the evaluation of fit between the brands in co-branding and the possible transfer of association between brands. Another method-related issue to be addressed in future research is the selection of the sample. Some researchers claim that studies conducted with a student sample may produce larger effect sizes because there is less error variance in measurement due to the homogeneity of student samples (Orsingher et al., 2009;Peterson, 2001). Indeed, the findings of this meta-analysis show that the correlation between brand characteristics and evaluation of co-branding is significantly stronger in studies with a student sample than in studies with a non-student sample.
This finding should encourage researchers to conduct studies with a non-student sample to ensure proper generalizability. By drawing attention to these method-related issues, this meta-analysis offers some further avenues for future research.
The fifth item identified for future research addresses the characteristic of meta-analysis itself and shows how future research could benefit from and expand it further. The scope of the metaanalysis papers is limited to the relevant studies conducted in the literature. Meta-analysis studies aim to synthesise prior studies, which might be state of the art in their field. Although meta-analysis studies are often limited in how they explain the underlying psychological processes, they can highlight the gaps in the literature and call for further studies examining these concepts for depth understanding (Rana & Paul, 2020). This meta-analysis highlights the importance of brand image fit in the evaluation of consumers by calculating the average effect size. However, despite the importance of fit between the partner brands to the success of co-branding, there is no universally accepted definition or conceptualisation of fit (Völckner & Sattler, 2006). While most researcher identifies fit as the brand image fit and product category fit, Ahn et al. (2020) examine the effectiveness of sensory fit, that is under-investigated so far in the literature. This meta-analysis encourages further research to examine why consumers perceive some brands as being a good fit with each other while others are seen as less so. Therefore, this metaanalysis offers guidance for how future research might reveal the underlying processes in consumer attitudes and behaviours towards co-branding. Hence, this meta-analysis contributes to the literature by outlining directions for further research, which would advance the understanding of this promising branding strategy and help maximise its benefits while mitigating its risks.

| ARTICLE S IN CLUDED IN THE ME TA-ANALYS IS
Coded articles are included in Table 8.

ACK N OWLED G EM ENTS
The author is grateful to Professor Tammo H.A. Bijmolt for the invaluable feedback on the first draft of this manuscript. The author is thankful for the support of Dr Floh, who coded four of the articles included in this meta-analysis simultaneously with the author. The author thanks three anonymous reviewers for their helpful comments on this manuscript.

CO N FLI C T O F I NTE R E S T
The author declares that there is no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.   Simonin and Ruth (1998) Brand equity Prior attitudes towards brands Rodrigue and Biswas (2004) Brand equity Perceived value Bleijerveld et al. (2015) Brand equity Brand-specific attitude Ahn et al. (2009) Brand equity Consumer-based brand equity Arnett et al. (2010) Continues PAYDAS TURAN

Brand-specific variables in metaanalysis model Variable name in the original study Source
Brand equity Brand familiarity Bouten et al. (2011);Lafferty et al. (2004); Naidoo and Hollebeek (2016) Brand equity Perceived quality towards original brand; positioning perception James (2006); Singh et al. (2014) Brand Equity Brand trust Ma et al. (2018); Naidoo and Hollebeek (2016) A PPE N D I X A3

PU B LI C ATI O N B I A S TE S T S
To minimise publication bias, several approaches were adopted at various stages of data collection and analysis.