NIMS - Dissertações de Mestrado em Marketing Analítico (Data-Driven Marketing)
URI permanente para esta coleção:
Navegar
Entradas recentes
- Graph-Based Recommender Systems for Value Creation in Horizon ConsortiaPublication . Martins, Ricardo Hurtado Coelho Parente; Pinheiro, Flávio Luís Portas; Pereira, João Pedro Lebre MagalhãesThis study addresses the problem of partner recommendation in research and innovation collaboration networks, focusing on consortium formation within the Horizon 2020 program. Collaboration in such contexts is a key mechanism for value creation, as organizations combine complementary resources, knowledge, and capabilities to generate innovation outcomes. However, identifying suitable partners remains a complex task, influenced by both network structure and organizational characteristics. To address this challenge, this research models collaboration as a link prediction problem and develops a comparative framework combining structural methods, graph neural networks (GNN), and subgraph-based approaches. Using publicly available data from the Horizon 2020 (CORDIS) database, multiple graph representations are constructed, including raw collaboration networks M1, community-based structures M2 and thematic (area-based) graphs M3. In addition, a set of organizational attributes, capturing dimensions such as participation intensity, funding capacity, thematic diversity, and international collaboration behavior, is incorporated to evaluate their contribution to recommendation performance. The study compares baseline structural models, LightGCN, GraphSAGE, and one BUDDY-style model under different experimental scenarios, using ranking-based evaluation metrics aligned with real-world recommender system settings. The results indicate that structural information plays a dominant role in predicting collaborations, with simple baseline methods outperforming more complex GNNbased approaches in several settings. Furthermore, the inclusion of organizational and marketing-related variables provides limited additional predictive value, suggesting that network topology is a primary driver of consortium formation. In contrast, structurefocused models such as BUDDY demonstrate strong performance by effectively capturing local structural patterns. These findings contribute to the literature by providing empirical evidence on the relative importance of structural and attributebased information in collaboration networks and offer practical insights for the design of recommendation systems supporting consortium formation in policy-driven environments.
- Selecting the Highest-Impact Product Finder: A Randomized Field Experiment with Behavioural DiagnosticsPublication . Cardona Rendón, Emmanuel; Lopes, Nuno Alexandre Moura PintoIn e-commerce, recommender systems are used to guide users toward suitable products and reduce potential choice complexity, which is especially relevant in highinvolvement categories. Emma Sleep United Kingdom online store uses a quiz-based guided recommendation tool referred to as the “Product Finder”. In the present study, a randomized field experiment was conducted to compare the current implementation, Provider A, with an alternative implementation, Provider B, which mostly differed in user experience and design. The experiment evaluates the two versions on conversion and revenue per product finder user and selects the outperforming provider according to these criteria. At the same time, it uses completion rate, recommendation clickthrough rate, and items viewed per user as complementary behavioural diagnostics to understand the tool’s funnel progression. The findings show that Provider B did not significantly improve conversion, contrary to the theoretical expectation, and the revenue analysis remained exploratory due to missing user-level distributions. However, the results show significant differences in the secondary steps of the funnel, evidencing meaningful behavioural differences. Provider B was associated with weaker upper and mid-funnel progression while presenting the highest click-topurchase rate among product page visitors. While the project could not identify a single superior implementation empirically according to the original criteria, it suggests a trade-off between broader funnel progression in the tool and stronger purchase efficiency at the last step. The study contributes to showing the value of evaluating live e-commerce recommender systems in both business outcomes and behavioural diagnostics. Additionally, from a managerial perspective, it supports the company’s provider decision and future iteration roadmap based on the empirical learnings and recognized strategic advantages.
- When Consumers Disconnect: Understanding Social Media Detox and Its Impact on Marketplace Responses and Purchase IntentionPublication . Alçada, Maria das Mercês Baptista de Sousa Baltazar; Neves, Joana Paisana Pires Costa dasThis study analyses the influence of social media quitting on consumer marketplace responses, together with their impact on purchase intention. While the literature has already explored key drivers, motivations, and psychological traits related to social media detox, little attention has been paid to this phenomenon from a consumer behavior perspective. Based on a sample of 173 individuals, results indicate that social media quitting negatively impacts brand trust and word-of-mouth. It also demonstrated the impact of market responses, such as the use of alternative channels and word-ofmouth, on purchase intention. Furthermore, the model shows a moderation effect of well-being and openness to experience throughout different relationships within the model. It also highlights the key role of word-of-mouth, given its full mediation on the relationship between social media quitting and purchase intention. Supported by the findings, this research presents guidance for maintaining consumer-brand relationships while in a social media disengagement period, which is becoming increasingly common.
- Behavioural segmentation of urban travellers: A data-driven approach to mobility profiling in the São Paulo Metropolitan RegionPublication . Ferreira, Wendy de Araujo; Lopes, Nuno Alexandre Moura PintoThis study applies a data-driven clustering approach to identify distinct urban mobility behaviour profiles among residents of the São Paulo Metropolitan Region. Drawing on individual-level trip data from the 2023 Origin-Destination Survey, a stratified household travel survey covering the full metropolitan area, the analysis constructed a dataset of 49,111 individuals who reported at least one trip on the survey day. Eight behavioural indicators were derived at the individual level, covering travel intensity, spatial reach, commuting structure, and transport mode use. After standardisation, kmeans clustering (k-means++ initialisation, n_init = 20, max_iter = 300, random_state = 42; scikit-learn 1.6.1, Python 3.12) was applied. The number of clusters was selected through combined elbow-method and silhouette analysis, with a four-cluster solution achieving a silhouette coefficient of 0.47. Stability analysis across five random seeds confirmed the robustness of the partition. The four profiles — Other-motorised mode users (29.3%), Car-dependent metropolitan commuters (15.2%), High-frequency public transport users (41.2%), and Active and multimodal travellers (14.3%) — differ substantially in travel time, distance, inter-municipal mobility, and modal composition. Sociodemographic analysis revealed statistically significant income differences across clusters (F = 363.45, df = 3, p < 0.001, η² = 0.033) and residential location differences (χ² = 1044.68, df = 3, p < 0.001, Cramér's V = 0.146). The findings suggest that mobility burden is unevenly distributed across the metropolitan population in ways that are systematically associated with income and residential location, contributing empirical evidence to debates on transport inequality and sustainable mobility in the Global South.
- The Role of Product Quality in Purchase Intention Across Augmented Reality and Traditional E-Commerce Shopping EnvironmentsPublication . Ribeiro, Maria Coelho; Barcellos, Márcia Dutra deAugmented Reality (AR) has increasingly transformed digital shopping experiences by enabling more immersive and interactive environments. Despite its growing adoption in online retail, limited research has compared how AR and Traditional E-commerce environments differently influence consumers’ purchase intention and the psychological mechanisms underlying these shopping experiences. This study investigates the relationships between Product Quality Perception (PQP), Trust (T), Immersive Experience (IE), Perceived Privacy Concerns (PPC), and Purchase Intention (PI) across AR and Traditional E-commerce scenarios. A quantitative research approach was adopted through an online survey, and the proposed relationships were analysed using Partial Least Squares Structural Equation Modelling (PLS-SEM) with SmartPLS. Additionally, a Multi-group Analysis (MGA) was conducted to compare the structural relationships between both AR and Traditional Ecommerce scenarios. The findings reveal that Product Quality Perception is a significant determinant of Purchase Intention in both scenarios. However, the mechanisms underlying this relationship differ across environments. In the AR scenario, Immersive Experience emerged as the strongest mediating factor influencing Purchase Intention, whereas both Trust and Immersive Experience significantly mediated Purchase Intention within the Traditional E-commerce scenario, with Trust emerging as the stronger mechanism. This study contributes to the literature on digital consumer behaviour by demonstrating how Trust and Immersive Experience influence consumers’ purchase intentions differently across AR and Traditional E-commerce environments. The findings also provide practical insights for retailers seeking to implement immersive technologies in online shopping environments while maintaining consumer trust and addressing privacy concerns.
- Social Media Self-Regulation: A Reflective-Impulsive Model ApproachPublication . Chorão, Inês Aires de Sá; Neves, Joana Paisana Pires Costa dasUsing the Reflective-Impulsive Model (RIM), this study analyses how reflective and impulsive processes influence individuals’ ability to regulate their social media use. While previous research has often examined these processes separately, this study looks at how they interact. Based on a sample of 229 social media users, the results show that reflective factors, such as cognitive dissonance and life satisfaction, increase individuals’ intention to reduce their use. In contrast, impulsive factors, particularly fear of missing out (FoMO) and fear of negative evaluation (FNE), increase compulsive use. The findings also show that stronger reduction intentions are associated with better self-regulation, while compulsive use (CU) makes it harder for individuals to control their behaviour. Digital wellbeing tools (DWT) do not directly improve selfregulation, but their effectiveness depends on habit strength. Overall, the study shows that both deliberate decisions and automatic behaviours play an important role in shaping social media use and provides insights for developing more effective digital wellbeing strategies.
- The Influence of Recommendation Algorithms in Consumer Behavioral IntentionsPublication . Amaral, Margarida Maria Vaz Rodrigues; Rohden, Simoni FernandaDigital platforms are increasingly using algorithms to personalize the content that users see and interact with in their daily online activities. These systems are based on large amounts of user data and are designed to improve the online experience by offering more relevant suggestions and making it easier for users to discover content. As personalization becomes more advanced and present in digital environments, it is important to understand how users react to it, especially in terms of privacy concerns, trust and their overall experience. This study examines how recommendation systems influence user’s behavioral intentions and well-being, and how trust and privacy concerns can shape the way users make decisions, based on privacy calculus theory. It also explores how users balance the benefits of receiving personalized content with the possible risks that are related with the collection and use of their personal data. By combining these different elements, the study aims to provide a more complete understanding of how people interact with recommendation systems in digital platforms.
- Hyper-Personalisation in Digital Finance: Understanding Disclosure Decisions in Data-Driven ServicesPublication . Moura, Helena Jerónimo Velho Cabral; Rohden, Simoni FernandaHyper-personalisation has emerged as a central capability in digital financial services, raising important questions about its impact on consumers’ willingness to disclose personal data. Grounded in the personalisation-privacy paradox and the privacy calculus framework, a quantitative experimental design was employed across two contexts: a routine home banking environment and a higher-stakes AI-driven investment platform. The findings indicate that hyper-personalisation does not exert a direct effect on willingness to disclose personal data. Instead, its influence is indirect and context dependent. In the banking context, hyper-personalisation increases both perceived benefits and perceived risks, generating opposing effects that offset each other at the behavioural level. In contrast, in the investment context, it does not significantly influence perceived risk, perceived benefits, or trust, even though these variables remain strong predictors of disclosure behaviour. This suggests that in higher-stakes environments, individuals rely less on interface cues and more on stable cognitive evaluations and pre-existing attitudes. Trust emerges as a stable antecedent rather than a short-term outcome of personalisation exposure, while data literacy plays a limited but directional role by strengthening the effect of perceived benefits. Overall, the findings support a conditional understanding of hyper-personalisation, demonstrating that its effectiveness depends on its ability to meaningfully shape users’ perceptions within specific decision contexts.
- AI-Driven Recommendation Systems and their Influence on Positive Consumer Product Choices: The Role of Ethical AI in Shaping Consumer Perceptions and Subjective Well-beingPublication . Timm, Nicol da Conceição Arone; Rohden, Simoni FernandaBased on the Stimulus-Organism-Response (SOR) framework, the study investigates how AI generated recommendations (stimulus) influence consumer perceptions, especially taste and health perception (organism) and how these, in turn, influence purchase interaction and subjective well-being (response). An experimental design approach was used with random participants assigned to one of two conditions: ethical AI recommendation condition or non-ethical non-AI recommendation. Data was gathered online that resulted in 109 valid responses. Mediation analyses were conducted to assess the indirect effects of taste and health perception and the results indicated that ethical AI recommendations significantly enhanced SWB but do not have a direct impact on purchase intention. Taste perception did not significantly mediate any relationship, suggesting that consumers rely more on cognitive cues than on anticipated sensory experiences. However, health perception significantly mediates the relationship between AI recommendation type and SWB emphasizing its importance as a psychological mechanism. The results suggest that ethical framing enhances consumer well-being, but additional strategies may be required to translate these effects into real purchasing behavior.
- Attached After Logging Off: How Social Media Reduction Transforms Consumer-Brand RelationshipsPublication . Lopes, Miguel Mesquita Seixas; Neves, Joana Paisana Pires Costa dasThe widespread use of social media platforms in modern society has intensified concerns regarding their psychological and behavioral effects, particularly in relation to emerging phenomena such as the Fear of Missing Out (FoMO). While FoMO has been shown to influence individual well-being, identity formation, and consumer-brand interactions, a growing trend of reduced social media usage raises new questions about how these dynamics evolve in contexts of digital disconnection. Addressing this gap, the present study investigates how social media discontinuance affects consumers’ emotional and behavioral relationships with brands, focusing on brand loyalty, consumer engagement, perceived value, and purchase intention. Additionally, FoMO is examined as an antecedent of discontinuance intention, and brand attachment as a moderating factor in these relationships. A quantitative survey was conducted using validated measurement scales, and the data was analyzed through Partial Least Squares Structural Equation Modelling (PLS-SEM). The findings reveal that social media discontinuance does not significantly reduce brand loyalty or perceived value, but has a positive effect on consumer engagement, suggesting more intentional and selective interactions with brands. Furthermore, brand loyalty, consumer engagement, and perceived value emerge as key drivers of purchase intention. The results also demonstrate that brand attachment plays a significant moderating role, with the effects of discontinuance varying depending on the strength of consumers’ emotional bonds with the brand. Overall, the study suggests that decreasing social media usage does not necessarily weaken consumer-brand relationships, but rather transforms how they are sustained, highlighting the importance of fostering deeper emotional connections beyond continuous digital exposure.