Visual communication is central to marketing. With the help of convolutional neural networks (CNNs) marketing has labeled large image datasets to understand visual impact. However, CNNs focus on local cues (e.g., smiles). They can miss marketing-relevant meanings shaped by context and configuration (e.g., joyful vs. sar- castic smiles). Recent advances like transformer-based vision models (TVMs) apply text-analytical concepts to image data. Vision language models (VLMs) such as GPT- 5 or Phi-4 jointly represent images and text. These richer linguistic representations might succeed in classifications where CNNs fall short. Unlike CNNs, pretrained VLMs require no additional training, even for new image-related tasks. However, it remains unclear how accurate they classify marketing-relevant labels. Which of these paradigms and classification models should marketing rely on? Is any single model best suited for all applications? Drawing on prior marketing publications, we identify 18 datasets covering what and who appears in images, and how images are perceived. VLMs such as GPT-5 and Phi-4 achieve state-of-the-art accuracy across a wide range of image- related tasks without requiring task-specific fine-tuning. However, they should not be trusted blindly. They can result in unexpectedly high error rates for some tasks. A multi-paradigm ensemble of TVMs and VLMs can overcome these challenges. We conclude with recommendations when to test which models.

Generative artificial intelligence (AI) is poised to transform how brands communicate with consumers. Recent research demonstrates AI's benefits in producing text, but marketing research has not yet explored how marketers can leverage AI to create visual advertising. Despite their impressive capabilities, "off-the-shelf" generative AI models are not aligned with marketing objectives, raising the question of whether it is possible to fine-tune generative AI directly on conventional advertising objectives (e.g., evoking attention, driving interest). In this research, the authors train an open-source generative AI model on marketing mindset metrics and show that the resulting visual content can match and even exceed conventionally produced advertising content in associated performance metrics. The results demonstrate that generative AI can be fine-tuned on multiple communication objectives simultaneously and adapted to specific audiences. In addition to highlighting generative AI's potential in marketing, this article explores the limitations of aligning visual generative AI with marketing objectives.

This study presents the first field investigation of the sales impact of cause-related marketing promotions (CMPs) in retail settings. Whereas prior work primarily studies CMPs in simplified experimental settings, actual fast-moving consumer goods markets are considerably more complex; ergo, consumers are unlikely to consider and evaluate all brands and CMPs in detail. In this analysis based on 63 CMPs across 20 categories, the authors therefore investigate the short-term sales impact of CMPs as a function of the brand and category context in which they are executed. On average, CMPs run 11 weeks and donate 3.2% of product price, resulting in an average sales lift of 4.9% per week. The findings suggest that a necessary precondition for CMP success is that consumers notice it at the point of sale and hence have considered the CMP brand for reasons other than the CMP itself. Accordingly, the sales impact of CMP can more than double when the category assortment is smaller, the category has less price dispersion, the brand is a category leader, or the brand is priced below the category average. Brands operating in less favorable market conditions can still achieve above-average CMP impact by combining CMPs with price promotions to ensure consumer consideration.

Sentiment is fundamental to human communication. Countless marketing applications mine opinions from social media communication, news articles, customer feedback, or corporate communication. Various sentiment analysis methods are available and new ones have recently been proposed. Lexicons can relate individual words and expressions to sentiment scores. In contrast, machine learning methods are more complex to interpret, but promise higher accuracy, i.e., fewer false classifications. We propose an empirical framework and quantify these trade-offs for different types of research questions, data characteristics, and analytical resources to enable informed method decisions contingent on the application context. Based on a meta-analysis of 272 datasets and 12 million sentiment-labeled text documents, we find that the recently proposed transfer learning models indeed perform best, but can perform worse than popular leaderboard benchmarks suggest. We quantify the accuracy-interpretability trade-off, showing that, compared to widely established lexicons, transfer learning models on average classify more than 20 percentage points more documents correctly. To form realistic performance expectations, additional context variables, most importantly the desired number of sentiment classes and the text length, should be taken into account. We provide a pre-trained sentiment analysis model (called SiEBERT) with open-source scripts that can be applied as easily as an off-the-shelf lexicon.

User attention in social media has long shifted from text to images. These images often contain brand logos in various contexts. The rapidly proliferating amount of such brand imagery makes it challenging for marketing research and practice to keep track of how brands appear online. The most intuitive data collection approach is to search for brand tags (e.g., #heineken) and study all related images. However, less than 20% of branded user-generated content contains a brand tag, leaving the remaining 80% unconsidered. We use deep convolutional neural networks to analyze more than 200,000 brand image posts with and without brand tags, covering more than 150 brands. Pursuing a multi-method approach, we consistently find that brand tags are not assigned randomly by users. Specifically, users are more likely to add brand tags to images with higher logo presence and image quality. Moreover, brands have differential tag potential such that brands with higher levels of user interest feature a higher share of brand-tagged images. Conversely, objects on the image that compete with brand visibility result in fewer brand tags. These systematic differences also affect viewer response, with tagged real-world images obtaining higher purchase-intention ratings in a controlled lab setting. Overall, we provide converging evidence that analyzing only brand-tagged content paints a biased picture of visual brand presence and discuss how marketing research can address these sample selection issues.

Cause-related marketing (CM), which links corporate donations to consumer purchases, has ongoing momentum in marketing. As the magnitude and direction of consumers’ response to CM are inconclusive, this meta-analysis synthesizes evidence on main and moderator effects from 237 studies. On average, the authors find a moderate effect for attitudinal response (d = .458) and a weak effect for behavioral response (d = .283; both ps < .001), both with high underlying heterogeneity. A multivariate meta-regression on CM moderators grounded along four conceptual pillars—transparency, signals of sincerity, purchase context, and consumers’ emotional attachment to CM—shows that attitudinal effects hinge mostly on emotional attachment. Suboptimal execution and poor communication of the donation appeal in particular can even have detrimental effects on attitudes. In addition, various moderators from other pillars play a relevant role. For behavioral outcomes, both emotional attachment and signals of sincerity are equally important. The visual prominence of the donation is the most relevant individual moderator, with only a few others related to the two pillars following at some distance. Therefore, CM requires different priorities depending on corporate objectives. This research further compares the effects of CM with those of discounts and other corporate social responsibility marketing instruments, simulates practical examples, and provides avenues for further research.

In service marketing, customers typically pay more when they use more. Based on this principle, various non-linear pricing plans or flat-rate tariffs attempt to lure customers into higher use and higher-revenue contracts. An emerging marketing practice we term precommitment-based pricing turns these principles around and asks customers to pay extra when they use the service too little. For example, a local fitness club offers customers a discount when they reach a minimum training frequency, and those who fall short pay a premium. This form of pricing aligns directly with customer objectives and assists them in achieving their goals. In this research, we assess which type of precommitment-based pricing is best suited to pay off for marketers and customers alike. We study whether refunds for high use (prepaid) or premium payments for low use (postpaid) are more effective and find empirically that these different types of payment have a differential impact on service adoption, goal attainment, and retention. Five empirical studies in three service domains demonstrate that prepaid contracts attract more customers, but postpaid contracts increase goal achievement and, thus, loyalty. We test boundary conditions and discuss practical implications on how to implement precommitment-based pricing.

Smartphones have made it nearly effortless to share images of branded experiences. This research classifies social media brand imagery and studies user response. Aside from packshots (standalone product images), two types of brand-related selfie images appear online: consumer selfies (featuring brands and consumers’ faces) and an emerging phenomenon the authors term “brand selfies” (invisible consumers holding a branded product). The authors use convolutional neural networks to identify these archetypes and train language models to infer social media response to more than a quarter-million brand-image posts (185 brands on Twitter and Instagram). They find that consumer-selfie images receive more sender engagement (i.e., likes and comments), whereas brand selfies result in more brand engagement, expressed by purchase intentions. These results cast doubt on whether conventional social media metrics are appropriate indicators of brand engagement. Results for display ads are consistent with this observation, with higher click-through rates for brand selfies than for consumer selfies. A controlled lab experiment suggests that self-reference is driving the differential response to selfie images. Collectively, these results demonstrate how (interpretable) machine learning helps extract marketing-relevant information from unstructured multimedia content and that selfie images are a matter of perspective in terms of actual brand engagement.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

Video content has become a major component of total internet traffic. Growing bandwidth and computational power conspire with an increasing number of video editing tools, smartphones, and online platforms that have facilitated video production, distribution, and consumption by businesses and consumers alike. This makes video content relevant across business research disciplines. However, analyzing videos can be a cumbersome manual task. Automated techniques are scattered across technical publications and are often not directly accessible to business researchers. This article synthesizes the current state of the art and provides a consolidated tool to efficiently extract 109 video-based variables, requiring no programming knowledge. The variables include structural video characteristics such as colorfulness as well as advanced content-related features such as scene cuts or human face detection. The authors discuss the research potential of video mining, the types of video features of likely interest, and illustrate application using a practical example.

For many consumer goods, the visual appearance is a vital determinant of market success. Although there is an emerging literature on how objective design characteristics drive consumer preferences, this literature has not yet taken into account that product design happens in the context of a brand’s equity. This research addresses the question of how to leverage brand equity when designing the visual appearance of a product. Specifically, it investigates the role of two key strategic visual design decisions: brand typicality (similarity within the brand’s range) and segment typicality (similarity to the competitive set). Drawing on fluency theory, the authors argue that high-equity brands benefit more from brand typicality but less from segment typicality than low-equity brands. Using data from the U.S. car market tracking market shares of 456 car models of 39 major brands operating in seven market segments across 13 years, the study provides empirical evidence for this conjecture and, thereby, implications for strategic product design and visual design theory.

Brand managers increasingly appeal to altruistic consumer motives by emphasizing corporate social responsibility (CSR) in marketing communication. However, little empirical evidence describes how CSR converts into market demand. This study investigates the dimensions of CSR that conspire with dimensions of brand equity to drive brand performance, by combining firm-level data on CSR, customer-based brand equity, market demand, and covariates covering 256 companies over a 15-year period. The findings show that internal CSR concerns relate negatively to brand associations as well as brand performance. Moreover, the demand effect of CSR can be fully explained by brand equity, underscoring the relevance of branding for CSR management. In particular, brand esteem and familiarity mediate the link between ethicality and market outcomes. Specifically, coping with CSR concerns appears to be important. In our data, perceived brand differentiation is unaffected by low responsibility, but ethical concerns undermine customers’ quality expectations and brand identification, suggesting managing CSR crisis has better prospects when the associated brand equity dimensions are addressed.

Information about ethical strengths and weaknesses of individual products (e.g., cause-related marketing, corporate social responsibility records) is increasingly available in the marketplace. However, market shares of ethical brands are still low, even though prior research indicates that ethical attributes influence brand choice. This research broadens the perspective of prior research by investigating the role of ethical attributes during earlier stages of the decision funnel, namely, consideration set formation. Four empirical studies demonstrate that brands' ethical strengths exert less impact on the consideration than on the choice stage. Specifically, brands that are not otherwise part of consideration sets benefit less from ethical strengths in larger assortments where consideration set formation plays a more important role. By investigating the screening rules that consumers apply, this study determines that while ethical strengths are subordinate, ethical weaknesses evoke asymmetric effects, such that misconduct exerts a stronger effect during screening processes than benefits do. To increase effects of ethical strengths on screening, firms can enhance the emotional intensity of ethical attributes and make them more salient and subjectively important.

Online social media drive the growth of unstructured text data. Many marketing applications require structuring this data at scales non-accessible to human coding, e.g., to detect communication shifts in sentiment or other researcher-defined content categories. Several methods have been proposed to automatically classify unstructured text. This paper compares the performance of ten such approaches (five lexicon-based, five machine learning algorithms) across 41 social media datasets covering major social media platforms, various sample sizes, and languages. So far, marketing research relies predominantly on support vector machines (SVM) and Linguistic Inquiry and Word Count (LIWC). Across all tasks we study, either random forest (RF) or naive Bayes (NB) performs best in terms of correctly uncovering human intuition. In particular, RF exhibits consistently high performance for three-class sentiment, NB for small samples sizes. SVM never outperform the remaining methods. All lexicon-based approaches, LIWC in particular, perform poorly compared with machine learning. In some applications, accuracies only slightly exceed chance. Since additional considerations of text classification choice are also in favor of NB and RF, our results suggest that marketing research can benefit from considering these alternatives.

This article proposes a framework for studying how a brand, firm, or individual can use networking activities to manage a social network and drive its success. Using data from ego networks of music artists, the article models how artists can enhance their social networking presence and stimulate relationships between fans to achieve long-term benefits in terms of music plays. The authors use a Bayesian modeling framework to model the heterogeneous and dynamic impact of networking activities on network structure and on music popularity, while relying on instrumental variables from another independent online social network to handle potential endogeneity. The results imply that artists can shape network structure via marketing activities and thereby achieve a long-term impact on success that far exceeds the direct and short-term impact in magnitude. Specifically, improving the density of ego networks enables long-term effects beyond those that stem from growth in network size.

Product designers continuously emphasize how important yet difficult it is to create new products with designs similar enough for brand recognition, yet dissimilar enough for a unique positioning. Advancing the understanding of perceived design similarity would greatly help to systematically manage this balance. Design literature typically assesses how much a new design deviates from the status quo based on consumers’ holistic similarity perceptions. For management practice, however, tests of numerous design alternatives with consumers throughout the NPD process are time‐consuming, costly, and threaten confidentiality. In addition, simple surveys often do not suffice to understand the role of individual design characteristics in the holistic perception of a product design. This article introduces an objective measurement approach of design similarity to overcome these drawbacks. The results of an empirical study demonstrate that consumers’ perception of design similarity can be reproduced by means of this approach. In addition, this approach makes measurement of design similarity more efficient, uncovers the design characteristics that drive holistic similarity perceptions, and enables managers to predict the positioning of alternative product designs in the perceptual space minimizing the need for consumer surveys.

Many consumer markets are becoming increasingly complex. In particular, as variety of products and product features proliferate, choosing is becoming a time consuming task for consumers. In response, consumers apply simplifying screening heuristics. However, established choice-based conjoint methods are challenged when consumers follow such decision rules. In particular, individual-level estimates of hierarchical Bayesian techniques (CBC-HB) cannot fully account for all possible heuristics applied by consumers. This research conducts a comprehensive comparison of four different indirect approaches for eliciting decision heuristics. We simulate a sample of heterogeneous synthetic respondents and measure whether these methods are capable of (1) inferring decision heuristics correctly, (2) adequately predicting consideration sets, and (3) improving choice predictions compared to CBC-HB. Our results show that noncompensatory inference provides additional information on decision heuristics as well as consideration sets, which can be valuable for advertising purposes as well as competitive strategy. Methods for inferring heuristics should be selected based on the complexity of products and markets. Across all scenarios, there exists at least one method which significantly improves predictive accuracy of CBC-HB.

Zusammenfassung Immer wieder ist im Marketing die Frage nach jenen Facetten eines Produkts zu beantworten, die dessen Attraktivität in den Augen der K unden ausmachen. Zumeist zielen Produktmanager darauf ab, ihren E rzeugnissen eine einzigartige Position im Markt zu verleihen, die sie von den E rzeugnissen der Wettbewerber differenziert. Erkenntnisse aus der Wahrnehmungs- und K ognitionspsychologie suggerieren die besondere Attraktivität des mittleren Produkts bzw. von G ütern, die geringfügig von diesen Prototypen abweichen. Anhand von E xperimenten im Automobilmarkt lässt sich die R elevanz dieses psychologischen Befundes auch für Produkte konstatieren. Hieraus ergeben sich I mplikationen für die segmentspezifische Produktgestaltung und das Variantenmanagement.