Effectiveness of engagement initiatives across engagement platforms: A meta-analysis

Engagement implies a subject who is engaged and an object with which this subject engages, as initially presented in organizational behavior research pertaining to employee engagement (Kahn, 1990) and informed by social identity theory (Ashforth & Mael, 1989). In marketing contexts, the engaged subject refers to existing or prospective customers, and the engagement object might be a brand or firm (Hollebeek & Macky, 2019). Marketing scholars also apply various labels to engagement (e.g., customer brand engagement, digital customer engagement; van Doorn et al., 2010). In turn, different conceptualizations of CE are available (Clark et al., 2020; Web Appendix B). In general, this variety can be summarized in two schools of thought: a behavioral conceptualization that focuses on behavioral CE exclusively (Kumar et al., 2019) or a multidimensional conceptualization that spans behavioral and psychological CE components (Clark et al., 2020). But even within these schools of thought, authors emphasize different substantive content (e.g., which dimensions comprise CE) and structures (e.g., relationships among CE dimensions).

That is, behavioral conceptualizations tend to suggest that recommendation, influence, and feedback behaviors drive firm-related outcomes and performance (Pansari & Kumar, 2017; van Doorn et al., 2010). However, scholars who adopt a value-based perspective broaden this scope to include transactional (e.g., customer purchases) in addition to nontransactional (e.g., word-of-mouth) contributions (Beckers et al., 2018). In line with the narrow conceptualization, we argue that only behaviors that go beyond transactions represent CE (van Doorn et al., 2010), so the goal should be to stimulate nontransactional, behavioral CE. Following this reasoning, in considering behavioral CE, we focus on nontransactional behaviors; transactional behaviors represent outcomes.

The multidimensional view of CE instead has been informed by organizational research (Rich et al., 2010), in which employees’ work engagement comprises cognitive, emotional, and behavioral CE (Hollebeek et al., 2019). As we noted previously, in consumer settings, CE refers to investments of resources, such as cognitive (e.g., knowledge), emotional (e.g., enthusiasm), and behavioral (e.g., skills) resources in specific interactions with a firm or brand. Engaged customers invest cognitive resources first, followed by emotional and behavioral investments (Hollebeek et al., 2014; Oliver, 1999). Attitudinal theories also provide relevant predictions of how cognitive and emotional constructs can influence behavior (Fishbein & Ajzen, 1975). Thus, we anticipate that cognitive CE relates to emotional CE, which influence behavioral CE.

Prior studies have traced the evolution of marketing, from transactional to relationship to CE marketing (Harmeling et al., 2017; Verhoef et al., 2010). That is, initial firm relationships with customers prioritized transactions (e.g., purchases), but firms’ focus on such transactional value creation shifted, prompting efforts that seek to establish and maintain customer relationships (Palmatier & Steinhoff, 2019; Pansari & Kumar, 2018). Customers appreciate connections with firms, which can occur through various platforms, as well as interacting with other customers (Pansari and Kumar 2018). Beckers et al. (2018) predict deeper relationships with engaged customers, who interact with the former at various points, beyond the point of purchase. During such interactions, customers form opinions about and become attached to the firm and its brands (Fuchs et al., 2010). Some engagement studies refer to the service-dominant (S-D) logic (Vargo & Lusch, 2004); this stream of literature emphasizes value co-creation and customers’ nontransactional contributions (Brodie et al., 2011; Payne et al., 2008). Even as the interrelationship of CE and the S-D logic has been established (Brodie et al., 2011), only recently have scholars begun developing integrated, conceptual frameworks (Hollebeek et al., 2019).

Regarding the antecedents of CE, van Doorn et al. (2010) propose three categories: customer (e.g., identity, consumption goals), firm (e.g., brand characteristics, firm reputation), and context (e.g., competitive factors). Harmeling et al. (2017) focus on firm strategies and consolidate insights from different literature streams to define and classify CE-specific marketing strategies, based on whether the strategies (1) deliberately motivate customers to invest resources and engage with the firm (i.e., CE marketing) or (2) organically evoke CE, without deliberate effort by the firm (i.e., traditional marketing strategies). In addition, they clarify the two types of engagement strategies that we cited previously. Due to their promising outcomes, task-based initiatives were the focus of many early studies. They assign structured tasks to customers (e.g., write a review, refer a friend), which customers must invest mental and physical effort to complete. In Fuchs and Schreier’s (2011) study of how task-based initiatives influence participation in new product development, they identify customers’ sense of psychological ownership, stemming from the effort they invest, as a source of more favorable evaluations of the firm (see also Van Dyne & Pierce, 2004). In contrast, experiential initiatives often feature playful events, rather than work-like tasks, that can induce a sense of self-transformation and encourage customers to incorporate the brand into their self-concept (Markus & Kunda, 1986). These initiatives strengthen psychological and emotional connections. For example, Cova and Pace (2006) show how Ferrero’s experiential events, such as Nutella Parties, prompted customers’ identification with the “my Nutella Community.” However, relatively fewer studies have examined this type of strategy.

We also acknowledge that traditional marketing strategies might organically stimulate Harmeling et al. (2017) stress the importance of product performance and brand associations, with the prediction that when customers have excellent experiences with an offering’s performance or the brand, they want to support the firm (e.g., through word-of-mouth), beyond purchasing (Verleye et al., 2014; Wallace et al., 2014).

Thus, in our proposed framework, we include task-based and experiential engagement strategies, together with traditional marketing strategies, so that we can compare their effectiveness. In some ways, our framework follows Harmeling et al.’s (2017), but we note several key differences. Their framework features behavioral CE only, it ignores platform differences, and excludes marketing outcomes. Moreover, we analyze the influence of a larger set of antecedents (Web Appendix C). In line with suggestions that CE can influence marketing outcomes, including firm performance (Kumar & Pansari, 2016) and customers’ behavioral intentions (Santini et al., 2020), we include these considerations in our framework.

Platforms enable and facilitate interactions between two or more parties, such as customers and firms (Rangaswamy et al., 2020; Wichmann et al., 2022). They function to sell goods and services, support socializing, and exchange information (Bonina et al., 2021). Platform theory distinguishes physical and digital platforms (Breidbach & Brodie, 2016; Wirtz et al. 2019). The Consumer Electronic Show is a physical platform that allows exhibitors and potential customers to interact directly, whereas Amazon or eBay provide digital platforms that facilitate online seller–buyer interactions (Breidbach & Brodie, 2016). Wirtz et al. (2019) also differentiate types of digital platforms, such as search (e.g., Bing), communication (e.g., WhatsApp), social media (e.g., Facebook), and matching (e.g., TaskRabbit) platforms. In their literature review, Rangaswamy et al. (2020) stress that despite the many studies of platforms in general, few of them elucidate the role of marketing (e.g., Bhargava & Rubel, 2019; Lamberton & Rose, 2012; Rosario et al., 2016). Thus, we need more research that takes a marketing perspective on platforms and tests their characteristics.

In customer engagement literature, studies often refer to “engagement platforms” that enable customer–firm or customer–customer interactions (Blasco-Arcas et al., 2016; Breidbach & Brodie, 2016). Digital engagement platforms tend to include review sections, brand development co-creation spaces, and social network elements (Blasco-Arcas et al., 2016). They allow customers to engage with a focal object (e.g., brand) during the purchase process, so firms in turn can actively drive customer engagement through their CE marketing efforts, such as encouraging customers to create content and exchange with others (Wichmann et al., 2022). Yet we find few studies examining the influence of platform characteristics on the effectiveness of engagement strategies, possibly because most research in this domain features a single sample on one platform and thus cannot undertake a comparative assessment. Santini et al. (2020) compare CE effects across blogs, Facebook, and Twitter, but they do not address engagement strategies or platform characteristics.

Finally, platform theory also suggests that some platforms are better suited to facilitate interactions than others. Specifically, prior literature identifies four main platform characteristics, which we accordingly include as moderators in our framework. That is, firms might leverage platforms with varying interaction intensity (continuous vs. spot) and richness (rich vs. lean) (Kaplan & Haenlein, 2010; Sawhney et al., 2005). They might also use digital or physical platforms to access customers, and they can consider whether the firm or customers initiate interactions the platforms they use (Beckers et al., 2018; Meire et al., 2019).

The influences of task-based versus experiential initiatives on CE likely move through two distinct mechanisms: psychological ownership and self-transformation.² First, because task-based initiatives require customers to invest some mental effort to perform a specific task (e.g., idea competition), they may induce a sense of psychological ownership (Harmeling et al., 2017), defined as “possessive feelings that some [engagement] object is ‘MINE’” (Van Dyne & Pierce, 2004, p. 440). This sense of ownership also evokes a feeling of responsibility toward the engagement object (Morewedge et al., 2021). Therefore, customers feel motivated to live up to their responsibility by supporting the firm, that is, by investing their cognitive, emotional, and behavioral resources in interactions with it.

Second, experiential initiatives, which tend to be multisensory, highly participatory, and shared (Harmeling et al., 2017), can lead to self-transformation and shifted customer beliefs and attitudes (Schouten et al., 2007), such that customers incorporate the brand that has provided the experience into their self-concept (Markus & Kunda, 1986). Once a firm or brand is incorporated into customers’ self-perceptions, they likely support it by investing cognitive, emotional, and behavioral resources in interactions (Harmeling et al., 2017).

Prior research indicates that it is easier to induce psychological ownership feelings through task-based initiatives than self-transformation through experiential initiatives (Harmeling et al., 2017). Self-transformation processes are complex and difficult to trigger because a customer’s self is relatively stable and resistant to change (Markus & Kunda, 1986). Accordingly, we anticipate that task-based initiatives have stronger effects on all three CE dimensions than experiential initiatives do.

Product performance (i.e., product-related experiences with the core offering) and brand associations (i.e., brand-related experiences with the core offering) can stimulate CE. Perceptions of the firm’s core offering, developed over time, often get stored in customers’ minds (Anderson & Bower, 2014; Harmeling et al., 2017; Keller, 1993), and when these positive experiences are recalled, customers are more likely to engage with the firm. Such product and brand knowledge may trigger cognitions, emotions, and behaviors (Anderson & Bower, 2014), signaling their positive links to the three dimensions of CE (Dessart et al., 2016). That is, due to their past positive experiences, customers should be motivated to invest cognitive, emotional, and behavioral resources into interactions with the firm or brand.

Finally, we predict differential effectiveness of CE versus traditional marketing. If customers can identify an underlying motive for firms’ strategies, they find the marketing efforts less credible and convincing (Roehm & Brady, 2007). The CE initiatives deliberately exist to push customers’ resource contributions, but traditional marketing strategies prioritize positive product and brand experiences, with engagement as an organic outcome (Harmeling et al., 2017). Thus, traditional strategies should be more credible and effective for driving CE.

The separate CE dimensions likely evoke different marketing outcomes, which might include behavioral intentions toward a firm (e.g., intention to repurchase; van Doorn et al., 2010) or behavioral outcomes (e.g., actual repurchases; Beckers et al., 2018). Engaged customers invest their cognitive, emotional, and behavioral resources into specific interactions and perceive their efforts to support the firm as rewarding, such that they should be eager to remain in a relationship with the firm. We also expect dependencies between intentions and behavior, as suggested by attitudinal theories (Fishbein & Ajzen, 1975) and prior engagement literature (Hollebeek et al., 2014).

Platform characteristics arguably might inform both the formation of CE and its translation into marketing outcomes. We propose that all platform characteristics likely moderate the translation of CE into marketing outcomes, but only two of them matter for the formation of CE. Furthermore, though the effectiveness of task-based initiatives depends on customers’ understanding of the assigned task, experiential initiatives are more effective when customers experience unusual, emotionally intense, interactive events (Harmeling et al., 2017). As we discuss next, prior literature implies that the intensity and richness of platform interactions moderate the effects of CE initiatives (Kaplan & Haenlein, 2010).

In turn, we predict stronger relationships between CE dimensions and marketing outcomes in spot compared with continuous interactions. Santini et al. (2020) explain that the strength of CE effects on marketing outcomes depends on the personal relevance of the engagement object to customers. When personal relevance is greater, customers likely consider their investments in interacting with the engagement object when they reflect on their future behavioral intentions or behaviors (Santini et al., 2020). The engagement object is more salient in spot than in continuous interactions. That is, as Johnston and Lane (2021, p. 2) explain, spot interactions occur “over short periods of time when an organization seeks to affect or connect with its stakeholders around a focal topic (e.g., issue or decision).” Because these spot interactions limited in time and specific to some particular issue or topic, they are more instrumental too (Johnston & Lane, 2018). For example, the Ice Bucket Challenge created vast awareness of the need for more research into ALS through intensive communication on social media within just three months. Customers likely reflect on the importance of this engagement object (Santini et al., 2020), so the effects of CE dimensions on marketing outcomes should be stronger for spot than for continuous interactions.

First, rich media might enhance communication effectiveness, because the greater amount and diversity of information transmitted improve understanding of the message (Vickery et al., 2004). Therefore, rich media should generally outperform lean media in terms of communication effectiveness. With respect to CE initiatives, both task-based and experiential initiatives could have stronger effects in rich interactions, in which companies communicate diverse information cues that help customers understand task-based initiatives, and if customers understand their assigned task, they likely develop feelings of ownership (Harmeling et al., 2017). Moreover, rich interactions may help customers incorporate the firm into their self-perceptions in response to experiential initiatives (Kaplan & Haenlein, 2010). Experiential events gain transformational capacities when they are unusual, emotionally intense, and interactive (Harmeling et al., 2017). Rich interactions offer more opportunities to communicate what makes an event unusual, emotionally intense, and interactive.

Second, rich media are not always better than lean media (Dennis & Kinney, 1998). According to the task–media fit hypothesis, unstructured tasks require rich media that can carry different types of messages, but structured tasks require lean media (Shirani et al., 1999). Task performance improves when there is a “match between information requirements of the task and a medium’s ability to convey information richness” (Suh, 1999, p. 297). If a medium is too rich for the task, it may distract customers and lead to inefficiency, because some of the rich information provided is not essential for effective communication (Suh, 1999). With respect to CE, task-based initiatives involve requests for structured tasks (e.g., retweet a hashtag) that guide customers’ voluntary resource contributions (Harmeling et al., 2017). To send such specific instructions, leaner media may be better suited. McGrath and Hollingshead (1993) stress, in their task–media fit framework, that leaner media (e.g., text-based) facilitate structured tasks, such as generating new product ideas and plans, better than richer media (e.g., virtual world). If customers get distracted by rich information that hinders their understanding of the assigned task, they are less likely to develop feelings of ownership. Thus, we acknowledge the argument that task-based initiatives may be less effective in rich interactions. We do not propose a rival hypothesis for experiential initiatives. According to Harmeling et al. (2017), experiential initiatives resemble unstructured play, and thus richer media still would be better suited according to the task–media fit hypothesis.

Finally, the richness of the platform interaction should influence the translation of CE into marketing outcomes, because rich interactions might disrupt the effects of CE dimensions on marketing outcomes. The diverse informational cues they transmit tend to increase the complexity of customers’ decision-making (Isen, 2001; Malhotra, 1984). In such a situation, customer’s own sense of engagement, becomes less important.

On the one hand, CE dimensions should display stronger effects on marketing outcomes on digital platforms, where CE is more relevant, due to the structural differences between digital and physical platforms (Lieberman & Schroeder, 2020). On digital platforms, customers can access information about the engagement object and enter into exchanges with others easily (Lieberman & Schroeder, 2020; Santini et al., 2020). For example, CE gained great popularity with the rise of social media, such as Twitter, Instagram, and Facebook, which facilitate the formation and maintenance of social networks (Lieberman & Schroeder, 2020). Through facilitated exchanges and information dissemination on digital platforms, customers learn about the importance of the engagement object (Santini et al., 2020), so CE should drive customers’ future marketing outcomes on digital platforms.

On the other hand, customers often rely on stereotypes when assessing computer-mediated communication (Jacobson, 1999) and perceive exchanges on digital platforms as shallow or lacking in depth (Taylor, 2022), as well as less meaningful than real-world exchanges. For example, after the peak of the COVID-19 pandemic, many firms encouraged employees to return to offices, citing the greater quality of real-world interactions compared with virtual interactions. Furthermore, Taylor and Kent (2014, p. 393) stress that “[p]osting comments on a social media site is no substitute for calling someone on the telephone, or meeting others, to discuss an issue.” Accordingly, exchanges about a focal issue (or CE effects) on a physical platform may be perceived as more meaningful than exchanges about the same issue on a digital platform.

The effects of product performance and brand associations, as traditional marketing efforts, also might vary with platform characteristics. Our focus explicitly is on CE initiatives, so we do not derive hypotheses for these traditional marketing strategies. Instead, we briefly explain a rationale for these parallel moderating effects. For both product performance and brand associations, we anticipate stronger effects of continuous versus spot interactions. Citing knowledge structures, Harmeling et al. (2017) note that repetition strengthens cognitive bonds, which improves information recall (Anderson & Bower, 2014). Repetition, as facilitated by continuous interactions, helps customers activate and consolidate memories (Pezzulo et al., 2014), such that they are more likely to recall past positive experiences, which can drive CE. In addition, product performance and brand associations both might exert weaker effects on CE dimensions in rich interactions, because their diverse informational cues distract customers (Dennis & Kinney, 1998), who then are less likely to pay attention to any specific cue that evokes their past positive experience with the firm. Because customers are less likely to recall past positive experiences in rich interactions, compared with lean ones, they are less likely to engage. We do not expect any differences for digital versus physical platforms or for different initiators of the interaction.

We include some methodological considerations in our framework, due to our need to account for the diversity of the studies we collected in our meta-analysis. First, we check for differences between customer surveys and other research designs. Surveys usually produce smaller effect sizes than experiments, because it is more difficult to eliminate potential confounds (Grewal et al., 2018). Second, the data collection approaches adopted by our source articles might involve single or multiple industries. With more diverse industries, the range of constructs of interest increases, which influences the magnitude of the effect sizes (Geyskens et al., 1998). Third, we consider the quality of the publication outlet. High-quality journals have more rigid mechanisms to control for factors that might inflate effect sizes (e.g., common method variance) (Hunter & Schmidt, 2004). Fourth, publication status serves as a moderator; we distinguish between published and unpublished contributions. Insignificant effects are less likely to be published in journals, due to publication bias (Hunter & Schmidt, 2004). Fifth, we include the study year, because customers’ greater experience over time with social media and other technology may foster their engagement (Hollebeek et al., 2014).

We searched for CE studies in several electronic databases (ABI/INFORM, ProQuest, Web of Science, EBSCO Information Services), using combinations of keywords such as “customer engagement,” “consumer engagement,” and “customer brand engagement.” We complemented these efforts with additional searches through Google Scholar, to identify conference proceedings and dissertations not published in journals. We also identified studies that cited van Doorn et al. (2010), as an influential contribution in this research domain, and reviewed the reference lists of all those studies, as well as those of other review articles, such as Santini et al. (2020) review. Finally, we searched for and included unpublished data sets.

Four criteria determine whether studies enter the meta-analysis. First, we required that they be empirical studies at the customer level and thus excluded conceptual articles (Hollebeek et al., 2019), qualitative studies (Hollebeek, 2013), and empirical studies that investigate engagement at the firm level (Gambetti et al., 2012). Second, the articles had to measure the CE construct; if they refer to CE but do not measure it, we exclude them. Third, eligible articles contain sufficient data to calculate effect size statistics (e.g., correlation coefficients, regression weights, t-values) for at least one relationship between CE and another construct. Fourth, we excluded studies of political engagement, civic engagement, or digital literacy, to ensure the study contexts involve firms and their products. We obtained 350 usable studies for the meta-analysis.

We used correlation coefficients as effect sizes because they are independent of the scale and reported in most studies. If this information was not available, we converted other statistical information into correlations (Peterson & Brown, 2005). In line with conventional standards for meta-analyses, we averaged the effect sizes of samples that reported more than one correlation for the same relationship, to avoid giving any sample too much weight in subsequent analyses (Palmatier et al., 2006). After averaging the effect sizes, the data set included 5,005 correlations, reported in 395 independent samples extracted from 350 studies. Of the 395 samples, 325 were published in journals and 70 in conference proceedings, dissertations, or unpublished works; 10 were published between 2005 and 2010, 101 between 2011 and 2015, and 284 between 2016 and 2020. The cumulative sample size was 434,233. In more detail, coders initially extracted 7,323 effect sizes, which were averaged into 5,005, of which 1,612 effect sizes involve constructs in Fig. 1, and the rest pertain to the antecedents discussed in the Web Appendix C.

Three coders extracted necessary information and calculated the effect sizes. They also classified the effects according to the construct definitions in Table 2 and examined the scale items to classify the effect sizes. They reached 97% agreement and discussed any disagreements jointly. The coders extracted information about sample sizes, construct reliabilities, study characteristics that reflect the contextual setting of the study, and method characteristics. With dummy codes, they gauge the four focal moderators: interaction intensity (1 = continuous; 0 = spot), richness (1 = rich; 0 = lean), platform type (1 = digital; 0 = physical), and initiator (1 = customer; 0 = firm). They also dummy-coded the control variables: data collection (1 = single industry; 0 = multiple), research design (1 = survey; 0 = other), and publication status (1 = published; 0 = unpublished). We extracted the study year from the articles and assessed publication outlet quality according to the ABS journal list, with ratings ranging from 1 (low quality) to 4 (high quality).

We used the widely adopted random effect approach suggested by Hunter and Schmidt (2004) to integrate effect sizes. First, we corrected the effect sizes for artifacts, including measurement error in the dependent or independent variable, then divided the correlations by the square root of the product of the respective reliabilities of the two constructs of interest. Second, we weighted the measurement error-corrected correlations by the sample size to correct for sampling errors. Third, with 95% confidence intervals for each sample size–weighted and artifact-adjusted correlation, we checked the power of our statistical tests (Muncer et al., 2003). We also calculated credibility intervals, which indicate the distribution of effect sizes (Hunter & Schmidt, 2004). Wide credibility intervals suggest variation in effect sizes and the need for moderator analyses to account for unexplained variance. Fourth, to assess the homogeneity of the effect size distribution, we used the χ² test of homogeneity (Hunter & Schmidt, 2004). Fifth, as a check for potential publication bias, we used Rosenthal’s (1979) fail-safe N (FSN), which indicates the number of studies with null results that would be required to lower a significant relationship to a barely significant level (p = 0.05). Rosenthal (1979) suggests a tolerance level, such that the results are robust when FSNs are greater than 5 ⋅ k + 10, where k equals the number of correlations. As another check for publication bias, we used funnel plots with effect sizes on one axis and sample sizes on the other; an asymmetric plot would indicate potential publication bias. Sixth, we calculated the shared variances between constructs and the binomial effect size display (BESD), to determine the practical relevance of the meta-analytic findings. A high BESD between CE and behavioral intentions indicates the likelihood of more favorable outcomes from one group (e.g., engaged customers) relative to a reference group (e.g., disengaged customers) (Grewal et al., 2018). Following these checks, we analyzed the data using a multilevel moderator analysis and structural equation modeling. Web Appendix D contains more information about the multivariate analyses.