DSpace Community:

THE EFFECTIVENESS OF INFLUENCER MARKETING ON GEN Z CONSUMER PURCHASE DECISIONS

2026-10-14T10:09:46Z

Title: THE EFFECTIVENESS OF INFLUENCER MARKETING ON GEN Z CONSUMER PURCHASE DECISIONS Authors: DIXIT, YASHOVARDHAN; Agrawal, Saurabh (supervisor) Abstract: The contemporary marketing landscape has witnessed a paradigm shift from traditional advertising models to digital and social media-driven marketing strategies. Among the most prominent of these shifts is the rise of influencer marketing, a practice wherein brands collaborate with individuals who possess significant social media followings to promote products and services. This phenomenon has gained unprecedented traction, particularly among Generation Z (Gen Z) consumers aged 18 to 25, who represent the first truly digital native demographic cohort. This research project investigates the impact of influencer marketing on the consumer buying behavior of Gen Z in India. The study examines the extent to which influencer recommendations shape purchase decisions, the types of influencers most trusted by Gen Z consumers, and the social media platforms that generate the highest levels of purchase influencing engagement. By undertaking this investigation, the study seeks to bridge an existing research gap regarding the behavioral determinants of Gen Z purchasing patterns in the Indian market. Objectives of the Study • To examine the extent to which influencer marketing affects the purchase decisions of Gen Z consumers. • To identify the types of influencers (nano, micro, macro, celebrity) most trusted by Gen Z respondents. • To determine the social media platforms most effective for influencer-driven consumer engagement. • To analyze the product categories most susceptible to influencer-driven purchase behavior. • To assess the role of trust, authenticity, and content type in mediating the influencer-consumer relationship.

RETAIL FORECASTING AND STOCKOUT DETECTION USING MACHINE LEARNING: A UNIFIED COMPARATIVE FRAMEWORK

2026-10-14T10:05:10Z

Title: RETAIL FORECASTING AND STOCKOUT DETECTION USING MACHINE LEARNING: A UNIFIED COMPARATIVE FRAMEWORK Authors: SAXENA, RISHITA; Sharma, Yogesh (SUPERVISOR) Abstract: The retail industry faces a persistent and costly challenge in the form of stockouts, situations where product inventory is depleted before customer demand is met. These events not only result in direct revenue losses but also damage customer loyalty and disrupt supply chain efficiency. The advancement of machine learning technologies, there is a growing opportunity to develop intelligent systems capable of predicting demand patterns and detecting stockout risks before they occur. This study addresses that opportunity through a structured two phase comparative framework applied to sales data of retail. The primary objective of this study is to compare the performance of four machine learning models, Linear Regression, Random Forest, XGBoost, and Long Short Term Memory networks, across two distinct but interconnected phases. Phase 1 focuses on demand forecasting where models predict future sales values and are evaluated using Root Mean Squared Error and Mean Absolute Error. Phase 2 introduces a novel stockout classification layer where the same models are retrained as classifiers to predict whether a stockout event will occur, evaluated using F1 Score and AUC-ROC metrics. The study utilizes the Rossmann Store Sales dataset which contains historical daily sales records from 1115 retail stores across Germany spanning several years. The dataset includes key retail features such as promotional activity, state and school holidays, store type, assortment level, and competitor proximity. After merging and preprocessing the data, a binary stockout label was engineered by identifying days where a store's sales dropped below 50 percent of its historical average while remaining open, a threshold determined through systematic experimentation across multiple levels. To address the severe class imbalance of only 1.65 percent stockout cases in the dataset, the Synthetic Minority Oversampling Technique known as SMOTE was applied to balance the training data before classification model training. Experimental results from Phase 1 reveal that Random Forest achieved the best demand forecasting performance with an RMSE of 922.26 and MAE of 593.05, outperforming XGBoost which recorded an RMSE of 1200.56 and MAE of 859.80. Linear Regression performed the worst with an RMSE of 2805.82 confirming that retail sales data exhibits strong non linear patterns that simple models cannot capture. LSTM recorded an RMSE of 2520.57 suggesting that without proper sequential time series structuring per store, deep learning models do not automatically outperform traditional ensemble methods. In Phase 2, Random Forest again demonstrated the strongest classification performance with an F1 Score of 0.6338 and AUC-ROC of 0.9722, making it the most reliable model for detecting actual stockout events. XGBoost achieved the highest AUC-ROC of 0.9757 indicating superior ability to rank stockout risk across probability thresholds, however its lower F1 Score of 0.4897 suggests it is less precise in directly classifying individual stockout days. LSTM recorded an F1 Score of 0.2829 and Logistic Regression performed the weakest with an F1 Score of 0.1134.

BEYOND THE DOORSTEP: EXPLORING THE WORK CONDITIONS OF DELIVERY AGENTS IN QUICK COMMERCE

2026-10-14T10:01:44Z

Title: BEYOND THE DOORSTEP: EXPLORING THE WORK CONDITIONS OF DELIVERY AGENTS IN QUICK COMMERCE Authors: SHARMA, KARAN; Prasad, Renuka (SUPERVISOR) Abstract: The current study analyzes the multifaceted problems of delivery agents operating in the burgeoning quick commerce sector, in particular for the Delhi NCR market. As quick commerce continues to reshape consumer buying behavior and aspirations, it also places significant strains on delivery agents as the backbone of the service. The study adopts a mixed-methods research strategy involving qualitative and quantitative data to analyze the working conditions, occupational hazards, safety concerns, and socio-economic issues of these frontline workers in a holistic manner. The primary data were collected through standardized questionnaires completed by delivery agents operating on varied platforms in Delhi NCR, while secondary data were collected through published reports, industry reports, and research papers. The findings of the research attempt to underscore the human cost of convenience embedded in the quick commerce model and provide recommendations on how the working conditions and overall well-being of delivery agents can be improved.

SILENT SURVEILLANCE: HOW PERSONALIZED DIGITAL MARKETING INVADES CONSUMER PRIVACY IN INDIA

2026-10-14T09:57:40Z

Title: SILENT SURVEILLANCE: HOW PERSONALIZED DIGITAL MARKETING INVADES CONSUMER PRIVACY IN INDIA Authors: JAIN, MRIDUL; Singh, Yashdeep (SUPERVISOR) Abstract: India's digital penetration is growing faster than other markets around the world and has already reached scale, with almost one billion internet connections, and millions of users added every month. At the heart of this development is a business model based on the idea of 'free services for personal data'. Platforms use behavioural data to turn every use of the search, or every use of an app, or every map query into advertising dollars. Although there are many who do it willingly, most users are not aware of what information is being collected, who is responsible for it, or if they are actually being given the choice to consent. The gap is found in this Major Research Project "Silent Surveillance: How Personalized Digital Marketing Invades Consumer Privacy in India". A structured primary survey was conducted on a sample of 192 respondents, primarily young adults aged 18-30, who were interviewed to explore the levels of awareness and emotional responses to data tracking, actual privacy protection behaviors, and attitudes towards personalization in a transparent context. Descriptive analysis, Pearson correlation (H1 and H4), Chi-Square test of independence (H2), one-way ANOVA (H3), independent-samples t-test, and Cronbach's Alpha reliability analysis were used to analyse the data. Three composite indices have been created: Awareness Index, Privacy Concern Index and Personalization Acceptance Index. The study yielded some important results. First, 92.7% of respondents reported having seen a targeted ad for something they just talked about recently and then never searched out online – the “phone listening” phenomenon. Second, 74.5 per cent of respondents confessed to providing false personal information in applications at least sometimes to ensure privacy protection, which has not been extensively explored in the Indian research literature. Third, the results of the present study supported Hypothesis 1: awareness of data tracking is positively and significantly related to privacy concern (r = 0.322, p < 0.001). All three other hypotheses were not supported with each having its own implication: privacy policy neglect is a cross-sectional Page 5 | Delhi School of Management, DTU phenomenon across occupations, privacy concern is a cross-sectional phenomenon across age groups, and concern is not predictive of willingness to pay for privacy preserving products—a confirmation of the Privacy Paradox in the Indian context. Finally, the report proposes six recommendations for good marketing practices: contextual advertising, first-party data strategies, zero-party data collection, transparent consent architecture, Privacy by Design, and enforcement of the Digital Personal Data Protection Act 2023. The key idea is that privacy issue cannot be solved by market mechanisms, but needs structural regulatory measures.