Design, implement and evaluate personalized video ads for mobile shoppers

Thesis Abstract

The rapid growth of mobile commerce and the ubiquity of video-enabled devices have intensified consumer exposure to personalized advertising, yet evidence on the causal impact of personalized video ads (PVAs) for mobile shoppers remains fragmented across contexts, formats, and user segments. This study investigates how PVA design, implementation, and evaluation influence engagement, perceived relevance, and conversion outcomes in mobile retail, addressing the problem of suboptimal targeting and ad fatigue that undermines return on advertising spend. The aim is to develop a validated framework for designing PVAs that optimally balance personalization and user experience to maximize purchase intention and actual sales. Specific objectives are (1) to identify design determinants of effective PVAs (content personalization, pacing, length, and contextual cues) for mobile shoppers; (2) to implement a scalable PVA system integrating user-level data (demographics, browsing history, and cart activity) with creative variants using a factorial design; (3) to evaluate the impact of PVAs on engagement metrics (view-through rate, click-through rate, dwell time) and business outcomes (add-to-cart rates, completed purchases) in a controlled field experiment; (4) to examine mediating roles of perceived relevance and ad annoyance and moderating roles of device type and purchase typology; and (5) to propose a theory-informed, practitioner-ready model for PVA deployment. A mixed-methods research design is employed, combining quantitative and qualitative approaches to ensure both causal inference and contextual understanding. The population comprises mobile shoppers aged 18–55 visiting three typical e-commerce platforms over a six-month period. A sample of 1,200 users is recruited through randomized assignment to treatment and control groups, with 400 users per platform and 100 users per experimental condition in a 2x2x2 factorial design manipulating personalization depth (classic vs. deep), pacing (fast vs. slow), and creative variant (standard vs. value-prop emphasis). Data collection instruments include (i) platform analytics extracting view-through rate (VTR), click-through rate (CTR), dwell time, add-to-cart, and purchase data; (ii) a post-exposure survey measuring perceived relevance, ad annoyance, brand attitude, and purchase intention using validated scales; and (iii) semi-structured interviews with 24 participants to explore experiential nuances and cognitive processing. Instrument validity and reliability are established through confirmatory factor analysis, Cronbach’s alpha, and test-retest checks. Ethical considerations address informed consent, data privacy, anonymization, and compliance with platform terms of service. Data analysis proceeds in stages. First, descriptive statistics summarize engagement and conversion patterns across conditions. Second, regression analyses (ordinary least squares and logistic regression) quantify the effects of design factors on VTR, CTR, and purchase likelihood, controlling for covariates such as age, gender, and device type. Third, structural equation modeling tests the hypothesized mediation effects of perceived relevance and ad annoyance, while multi-group SEM examines moderation by device and purchase typology. Fourth, ANOVA and post-hoc comparisons identify interaction effects among personalization depth, pacing, and creative variant. Fifth, qualitative data from interviews are analyzed thematically using an inductive-deductive approach to triangulate quantitative findings and uncover deeper cognitive and emotional responses to PVAs. The study draws on the Uses and Gratifications Theory and the Elaboration Likelihood Model to frame consumer processing of personalized video content, and integrates the Information Processing Theory to explain attention dynamics. A conceptual model is proposed to synthesize theoretical constructs with empirical results. Expected findings indicate that deeper personalization combined with moderate pacing and value-prop–focused creative variants yields the highest engagement and conversion, with perceived relevance mediating the relationship and ad annoyance moderating effects by device type (mobile phone vs. tablet) and purchase typology (impulse vs. planned buyers). The study contributes to knowledge by advancing a robust, empirically validated framework for designing PVAs in mobile contexts, clarifying the interaction between personalization depth, pacing, and creative emphasis, and offering measurable guidelines for practitioners on balancing user experience with advertising effectiveness. Policy implications include recommendations for platform- and brand-level guidelines on data-driven personalization boundaries and opt-in consent. The main conclusion posits that thoughtfully engineered PVAs can substantially enhance mobile shopper conversion without exacerbating ad fatigue, provided that personalization is contextually appropriate, pacing respects user attention, and creative narratives align with explicit consumer goals. Practical recommendations for marketers include adopting modular PVA templates, deploying A/B and factorial experiments for ongoing optimization, and integrating post-exposure feedback loops to continuously refine personalization algorithms. Suggestions for further research include exploring cross-cultural differences in PVA reception, long-term brand equity effects, and the role of multimodal (audio-visual) cues in mobile advertising ecosystems.