Marketing analytics has increasingly embraced multi-method approaches to capture the complexity of consumer behavior and organizational phenomena. Among the most prominent are Structural Equation Modeling (SEM), Qualitative Comparative Analysis (QCA), and Artificial Neural Network (ANN), three methodological approaches each grounded in a distinct logic of exploration/theory testing, explanation, and prediction. Despite their complementary strengths, researchers have lacked systematic guidance on how and why their combined application would be beneficial. This research addresses this gap by providing comprehensive guidelines for integrating SEM, QCA, and ANN in a single analytical project. In this article we summarize the theoretical foundations, assumptions, and practical considerations of each method. More specifically, using a synthetic dataset, we demonstrate how SEM is effective for identifying linear and global relationships in a proposed theoretical model. We then explain how QCA can be applied to uncover critical segment-specific “causal recipes” which often remain hidden in SEM. Finally, ANN searches for interactions as well as nonlinear relationships among predictors and which may lead to superior predictive accuracy. By combining these analytical approaches under a single unified analytical framework, this research enhances methodological rigor, mitigates risks of data overfitting and p-hacking, and supports identification of more nuanced insights into complex marketing phenomena. Recommendations for future research directions are also discussed.
The primary aim of this study is to examine the effects of social network support and network tie strength on a customer’s intention to using peer-to-peer digital payment. The study further posits that this relationship is conditioned by the mediator trust, and the moderator perceived cybersecurity. Four hypotheses were posited based on the social network theory, perceived risk theory, and trust commitment theory. This study used a quantitative research design to test the hypotheses. Data were collected from Prolific Academic online panel, yielding 262 responses. Data were analyzed using multi regression and Macro PROCESS, a moderated mediation Model 14. The study found that social network support has a significant effect on the use of P2P digital payments through both direct and indirect pathways. The effect of tie strength on P2P digital payment use is mediated by trust and moderated by perceived cybersecurity, with no significant direct effect observed. This finding demonstrates the crucial role of social networks and tie strength in influencing consumers to use P2P payment services, indicating that marketers can leverage these insights to design strategies that encourage peer recommendations, and ensure digital payment services emphasize trustworthiness and lessens the perceived risk of their offerings. The findings also highlight how interpersonal and platform-based mechanisms jointly reduce risk and shape user intention in digitally mediated transactions.
Traditionally marketers focus on brand equity metrics such as choice or purchase intention to assess the value of a brand for consumers. Studies of firm brand value typically employ a cross-firm level of analysis using data from a sample of firms to show that brands have greater financial value than unbranded or weakly branded products. However, this type of analysis only establishes the value of brands in general. An important marketing problem is to measure the value of a brand to an individual firm. This calls for an individual firm level of analysis. We propose a two-step method for this type of analysis, combining a consumer choice experiment with financial modeling to compare a firm’s cash flow with the brand to what it would be, counterfactually, without the brand. We provide an empirical proof-of-concept demonstration of this measurement methodology. The goal is treating brand as an asset, not an expense.
This study examines which university actions Generation Z views as most strongly associated with strengthening climate awareness and reducing climate nonchalance. Drawing on data from a structured online questionnaire completed by 1,050 Generation Z respondents aged 18–25 from the university communities of nine Latin American institutions, the analysis uses the CHAID segmentation technique to identify differentiated profiles of climate awareness and climate nonchalance. These segments are statistically validated through ANOVA and Chi-square tests, confirming significant variations in how young people perceive institutional actions related to climate change. The findings show that integrating specific climate‑change content into university curricula, promoting applied research, and encouraging co‑creation processes between universities, Generation Z, and territorial stakeholders are perceived as the university actions most strongly tied to climate engagement. Although the study focuses on Latin American institutions, the results highlight the relevance of comparing these patterns with European contexts in future research and suggest incorporating additional psychological variables—such as motivation, subjective norms, or satisfaction—to refine educational and environmental policies. By applying CHAID to analyse affective responses toward climate change, the study offers a nuanced, context‑specific view that avoids broad generalisations and provides actionable insights for designing targeted university‑based climate mitigation initiatives.
Firms use RFM analysis to identify valuable customers, particularly loyal customers. In practice and in prior studies, RFM-based scores are often treated as static indicators. However, when observed periodically, these scores can be analyzed dynamically, allowing firms to track changes in customer loyalty and design retention strategies. Although prior studies have modeled customer segments as states in Markov models, they have mainly focused on predicting future segment distributions and have not sufficiently examined how long customers remain in each state before reaching churn. To address this gap, this study applies an absorbing Markov chain framework to dynamic RFM-based customer segments. We define High-Active, Active, Needs-Attention, Low-Active, and Inactive as transient states and Churn as an absorbing state. Using POS transaction data from a retailer operating both physical stores and an online store, we compute the expected steps to absorption and the expected number of visits to transient states before absorption. We also compare online-only, offline-only, and multichannel customers. The results show that the expected steps to absorption and state-visit patterns before absorption into churn vary across channel-based customer groups. These indicators provide decision-support information for CRM monitoring, retention intensity, and channel-specific marketing interventions.
Virtual influencers have emerged as a novel force in influencer marketing, challenging human influencers and reshaping social media dynamics. Despite growing academic interest, research remains fragmented across multiple theoretical perspectives, with inconclusive and often contradictory findings regarding their effectiveness relative to human influencers. Following PRISMA 2020 guidelines, we systematically reviewed 72 empirical studies initially identified from 473 Scopus-indexed publications published between 2018 and 2025. This review synthesizes and analyzes the emerging body of knowledge by organizing findings around themes and sub-themes, including research focus, benefits, risks, and operational mechanisms. The results highlight that virtual influencers represent a strategic asset for brand engagement; however, their effectiveness depends on the dynamic interplay among anthropomorphism, emotional expressiveness, interactivity, transparency practices, and brand-audience congruence. By integrating these findings, this review develops a structured framework of operational mechanisms and boundary conditions for leveraging virtual influencers in the evolving landscape of influencer marketing. In doing so, this study seeks to advance the theoretical understanding of virtual influencers and provide actionable guidance for future advertising and marketing research.
The adoption of augmented reality (AR) shopping is transforming the consumer shopping experience, but it is unclear how exactly it affects consumers' purchase intention, and whether providing either an informational or a risk-based cue would facilitate or impede the adoption of this technology. This study adopts a hybrid method of PLS-SEM, artificial neural network (ANN) and necessary condition analysis (NCA) to investigate the impact of seven antecedents of AR purchase intention: diagnosticity, information quality, memorable experience, hedonic motivation, perceived risk, social approval, and trust. The results demonstrate that the variables of memorable experience, hedonic motivation, trust and social approval significantly explain purchase intention, while diagnosticity, information quality and perceived risk do not significantly explain purchase intention despite having a negative direction. When nonlinearity is assumed, the four significant predictors are as expected again included in the ANN analysis. In NCA results, all seven constructs are found to be medium necessity effects in the absence of sufficiency effects in PLS-SEM. The study contributes theoretically by offering a more balanced understanding of the positive and negative dimensions of AR adoption, while methodologically demonstrating the value of combining symmetric and asymmetric analytical techniques to explain complex consumer behavior in immersive digital environments. Practical implications for the design of more transparent, user-centered, and cognitively sustainable AR retail experiences are also discussed. The results indicate that the primary motivators towards AR purchase intention are the relational and experiential mechanisms.
Null results are central to accumulating knowledge in marketing, yet conventional null hypothesis significance testing (NHST) renders such results ambiguous. A nonsignificant p-value does not distinguish between data insufficiency and data that meaningfully support the absence of an effect. In this brief note we argue that Bayesian inference provides a useful framework for interpreting null findings in marketing analytics. By allowing probability distributions on unknown parameters and then comparing the predictive adequacy of null and alternative models, Bayesian methods allow researchers to distinguish between the evidence of absence and the absence of evidence. This note outlines how Bayes factors, posterior intervals, and interval null hypotheses can make null findings more informative whether those findings derive from laboratory experiments, field experiments or secondary data studies. We note that publication bias in favor of significant results is not just an issue for editors but is rather a problem across the full field of marketing since it impedes accumulation of knowledge. In fact, when null results are analyzed and reported in ways that quantify evidential strength, they become easier to interpret, easier to compare across studies, and more useful to both academic theory and managerial decision-making.
Understanding how omnichannel customer experiences influence repurchase intention is increasingly important as retailers seek to retain consumers across integrated physical and digital touchpoints. This issue is particularly relevant in fast-moving consumer goods (FMCG) markets, where purchasing decisions are often habitual, convenience-driven, and characterized by relatively low involvement. This study investigates how omnichannel customer experience shapes repurchase intention using the Stimulus–Organism–Response framework and a dual-method approach integrating linear and configurational analyses. Although prior research has examined the influence of omnichannel retailing on consumer behaviour, less is known about how multiple psychological mechanisms combine and interact to generate repurchase intention in FMCG settings, where repeat purchasing is frequently assumed to be driven primarily by habit and functional considerations. Data were collected from 209 consumers in the Netherlands and analysed using partial least squares structural equation modelling (PLS-SEM) and fuzzy-set qualitative comparative analysis (fsQCA). The findings indicate that omnichannel customer experience strengthens value co-creation, customer loyalty, and customer engagement. Customer engagement emerges as the strongest predictor of repurchase intention, followed by customer loyalty, whereas value co-creation does not exhibit a significant net effect. However, fsQCA reveals multiple sufficient pathways to repurchase intention, demonstrating that different combinations of omnichannel experience and organismic responses can generate favourable behavioural outcomes. The study extends the S–O–R framework by showing that organismic responses do not contribute uniformly to repurchase intention but operate through different experiential and relational configurations. In particular, engagement and loyalty provide alternative and reinforcing pathways, while high value co-creation is not required when other experiential or relational conditions are present.
This study examines how trust-related signals are expressed and how they relate to continued engagement, as expressed in reviews, on digital lending platforms, using user-generated content from mobile app marketplaces. Arabic-language reviews from Google Play and the Apple App Store were collected, cleaned, and analyzed using a multi-stage analytics pipeline that combined exploratory text mining, sentiment analysis, topic modeling, construct mapping, and confirmatory modeling. Drawing on an extended UTAUT2 framework, the study operationalizes review-level indicators for performance expectancy, effort expectancy, hedonic motivation, price value, perceived risk, Sharia compliance, emotional assurance, user trust, and continued engagement. The findings show that performance expectancy and effort expectancy are important positive correlates of trust, while perceived risk is negatively associated with trust. The results further show that Sharia compliance and emotional assurance appear as context-specific trust-related signals in digital lending environments. User trust is positively associated with continued engagement expressed in reviews. However, the explanatory power of the engagement models is modest, suggesting that engagement-related expressions in app reviews are influenced by additional factors not captured by the present framework. The study contributes to marketing analytics and fintech research by demonstrating how app review analytics can transform unstructured user feedback into theory-grounded insights into trust and engagement. It also offers practical implications for digital lenders seeking to improve transparency, reassurance, service design, and governance communication in high-stakes financial service settings.
This paper examines whether consumer-oriented linguistic features that make brand slogans effective in the marketplace also influence their competitive position on the trademark register. The study analyses 688 global brand slogans using computational linguistic methods to capture processing fluency, lexical distinctiveness, lexical diversity, and emotional valence, and links these features to a USPTO-based measure of competitive trademark crowding operationalised as the inverse of the number of similar registered trademarks. The results reveal an apparent disconnect between consumer-oriented linguistic heuristics and trademark registry outcomes. These preliminary findings suggest that the linguistic construction of a slogan has a negligible impact on its competitive position in the trademark register, and that the factors driving legal protectability lie largely outside the domain of linguistic design. The results challenge the assumption that linguistically effective slogans automatically translate into protectable brand assets, and underscore the need for closer coordination between branding and legal teams when developing verbal brand identity.
Consider a paradox at the heart of marketing strategy: managers devote substantial effort to determining what products should be bundled and how those bundles should be priced, yet comparatively little attention is paid to whether the message describing the bundle reflects the same value logic through which customers evaluate it. Product bundling has long been recognized as a powerful mechanism for increasing perceived value, reducing consumer search costs, encouraging trial, and enhancing firm profitability (Stremersch Tellis 2002; Venkatesh Mahajan 1993). Yet even carefully designed bundles often underperform when the rationale communicated to customers fails to align with the value they seek from the offering. This research proposes “communication bundling” as a theoretical construct distinct from integrated marketing communications and multichannel strategy, and introduces a synchronization matrix that quantifies product-communication alignment as the mechanism linking Quality Function Deployment (product configuration) with Elaboration Likelihood Model principles (platform-specific communication). We derive four testable propositions and evaluate them through a quasi-experimental field study of 31 companies across four industries in China over 6 months. Synchronized strategies improved ROI by 42.3
This study examines how service robot features, namely anthropomorphism, autonomy, and intelligence, influence consumers’ willingness to pay more through perceived warmth, perceived competence, and social comfort. Using survey data from consumers with prior hotel service robot experience, the study applies PLS-SEM and combined importance-performance map analysis to identify the key drivers of willingness to pay more. The findings show that perceived intelligence is the strongest predictor of both warmth and competence, while anthropomorphism also positively influences both perceptions. Autonomy has no significant effect. Warmth and competence enhance social comfort, which subsequently increases willingness to pay more. Identity threat does not significantly moderate these relationships. By examining hotel service robots as a form of disruptive innovation in service delivery, this study contributes to marketing analytics by modelling the linkages between robot design features, consumer social evaluations, and price-premium outcomes.
This study examines how YouTube advertising influences brand trust, brand loyalty, and brand stickiness, while also assessing the mediating role of brand trust and the moderating role of advertising stickiness. Integrating advertising value dimensions, Stimulus–Organism–Response (S–O–R) logic, and relationship marketing, the study develops a coherent framework for explaining how YouTube advertising evaluations are translated into downstream relational outcomes. A comparative quantitative design was employed across three service industries; telecommunications, banking, and airways, yielding 428, 430, and 419 usable questionnaires, respectively. The findings show that entertainment, informativeness, irritation, and credibility shape brand trust, which in turn influences brand loyalty and brand stickiness. Credibility emerges as the most consistent antecedent of brand trust across sectors, while the effects of other advertising dimensions vary by industry. The moderating role of advertising stickiness is supported only for the relationship between brand trust and brand loyalty in the telecommunications sector, indicating a limited and context specific boundary effect. The study contributes by offering an integrated explanation of how YouTube advertising evaluations influence relational outcomes, clarifying brand trust as the central mediating mechanism, distinguishing brand loyalty from brand stickiness, and positioning advertising stickiness as a conditional boundary factor. Managerially, the findings suggest that brands should prioritise credibility and informativeness in YouTube advertising and align promotional claims closely with service delivery to strengthen trust and its downstream effects.
Video ads remain one of the most important ad formats. As multichannel video ad campaigns increasingly span TV, social media, and YouTube, advertisers need insights into specific content features associated with higher recall and attitudes for each channel. Research linking ad content to effectiveness has, however, largely focused on single channels, limiting insight into which granular content features generalize across channels and which are channel-specific. Building on limited-capacity processing and processing fluency, we argue that creative dimensions operate differently across channels because TV, online video, and social video create different attention constraints and fluency demands, leading to different granular content element effects per channel. We analyze these effects empirically using 3495 TV, online, and social media ads. Using hybrid algorithmic and human content coding, we extract 161 common content features and estimate their contributions to recall and attitudes with Bayesian Additive Regression Tree models. The findings identify top elements per channel and reveal substantial differences across channels and objectives, showing that multichannel video creative requires adaptation not only by channel but also by campaign objective. We further show how the approach can support scalable, feature-level creative diagnostics for ad pretesting.
As online marketplaces expand, product-safety incidents have become a growing concern for consumers, marketplace operators, and regulators. This study examines consumers’ intentions to conduct pre-purchase safety checks, such as verifying product recalls, safety marks, and seller credibility, by integrating consumers’ safety-relevant knowledge, product-level perceived risk, platform-level perceived risk, self-efficacy, and risk preference. Survey data were collected from 1,215 Japanese online marketplace users who had purchased both electrical appliances and children’s toys in the past year. Furthermore, structural equation models were estimated separately by product category. The results show that safety-check intention was most strongly associated with product-level perceived risk. It was also positively associated with self-efficacy, safety-relevant knowledge, and platform-level perceived risk, while risk preference was negatively associated with intention. Safety-relevant knowledge was associated with intention both directly and indirectly, mainly through perceived risk and self-efficacy, and unexpectedly, with higher perceived risks. Furthermore, prior trouble experiences and longer online shopping experience were associated with greater knowledge, whereas learning experiences were primarily associated with stronger self-efficacy. The findings support a two-layer risk framework and suggest analytics-based, low-friction safety interventions, including safety mark displays, recall prompts, and targeted interface support, for marketplace operators and regulators.
The right assortment can make or break a retailer’s business. Assortment optimization is no longer a simple matter of stocking shelves with popular items. The retailer aims to satisfy the demand of the broadest possible customer base while keeping the cost low—by narrowing the product range. In this paper, we present an algorithm for efficient product range reduction. In addition, we propose several metrics to measure the mutual substitutability of pairs of products. We define two products to be substitutable if a customer is willing to take one product instead of the other. The proposed substitution methodology results are then utilized to reduce the overall assortment using a diminishing-returns-based range reduction approach. In an empirical study of a retail chain, we demonstrate our methodology’s properties and verify its usability. Our heuristic approach provides swift, computationally efficient, and meaningful insights into the market structure. Therefore, it can support in the assortment optimization process.
With the growing importance of social networks and digitalisation, influencer marketing is becoming a key tool for reaching young consumers in particular, gradually replacing traditional one-way communication channels. Despite the abundance of empirical studies in this area, the problem of fragmentation of knowledge and reliance on isolated constructs or linear models that do not reflect the sequence of the decision-making process persists. Moreover, purchase intention is often studied in the literature only as a unidimensional category. The aim of this article is to comprehensively explain the impact of key determinants of influencer marketing—credibility, perceived influence and engagement—on purchase intention. The primary research focused on young adults as a key segment affected by digitalization and the main consumers of influencer marketing. Based on the principles of the AIDA model, the study deconstructs purchase intention in more detail into three specific phases: (i) interest in the product, (ii) search for more information and (iii) desire for the product. From a methodological perspective, the article fills an existing knowledge gap focused on multistage purchase intentions by using empirical research and an ordinal regression model supplemented by machine learning models and SHAP-based interpretation. This approach allows for a more precise understanding of the transitions between different levels of behavioural intentions while also assessing the predictive relevance of the analysed determinants and the consistency of findings across inferential and predictive approaches.
In many studies, it is important to reduce not only the variable mode of a data matrix but also the individual mode, because characterizing entities in the first mode can be equally desirable. In this manuscript, we propose a new two-mode component model, introduced as Disjoint-TMC, together with an algorithm for computing disjoint orthogonal components in both modes. The method combines an Alternating Least Squares (ALS) technique with a heuristic procedure to enforce disjointness in the loading matrices. As a result, two disjoint orthogonal loading matrices are obtained, along with a core structure that captures the most relevant interactions between reduced modes. A distinctive feature of the Disjoint-TMC model is that it enables grouping of entities in the first mode-such as firms, consumers, suppliers, products, brands, time periods or regions in marketing applications-and allows a formal characterization of these groups. While this approach produces a loss of fit compared to unconstrained decompositions, our computational experiments demonstrate clear benefits in interpretability. Applications in marketing analytics are included to demonstrate the benefits of the proposed model, and a Python library is provided to facilitate its use and extension by other researchers. This work provides both methodological innovation and practical tools for interpretable two-mode analysis.
