Machine learning applications in digital advertising performance optimization: A systematic literature review

RSCT Certification: κ=0.550 (pending) | RSN: 0.00/0.00/0.00 | Topics: AI Safety and Alignment

Analysis of "Machine learning applications in digital advertising performance optimization: A systematic literature review"

Core Contribution: This paper provides a comprehensive overview of how machine learning (ML) is transforming the rapidly growing digital advertising industry, which is projected to reach $798.7 billion by 2025. The key objective is to systematically analyze the scientific literature on the application of ML techniques in web advertising campaigns. The authors implement a rigorous PRISMA methodology to select 42 high-quality papers (scoring 13-15 out of 15) published between 2010 and 2025, representing the state-of-the-art in this domain.

The paper's central contribution is uncovering the dominant trends and critical gaps in the use of ML for digital advertising performance optimization. This includes identifying the predominant ML architectures (e.g., deep learning, attention models), the convergence towards click-through rate (CTR) as a universal optimization metric, the concentration of research in the e-commerce sector, and the fundamental challenge of data sparsity. By synthesizing these insights, the authors are able to derive actionable implications for future research and industry practice.

Technical Approach: The authors employed a systematic literature review methodology to identify, screen, and select high-quality papers on the topic of ML applications in digital advertising. They leveraged a comprehensive set of search terms across multiple databases to capture the relevant scientific literature. The paper selection process involved a rigorous five-dimensional quality assessment, with each paper scored between 0-3 points across criteria such as clarity, methodological rigor, and relevance. This allowed the authors to focus their analysis on the 42 most impactful publications (scoring 13-15 points).

In terms of the ML techniques covered, the review reveals a clear dominance of deep learning approaches, with deep neural networks (35.7%) and attention models (19.0%) being the most prevalent architectures. The authors also note the convergence towards CTR as the primary optimization metric, used in 95.2% of the selected papers. Interestingly, the majority of the research (61.9%) is concentrated in the e-commerce sector, with a significant contribution (14.3%) from the Alibaba ecosystem.

Key Results: The systematic review uncovers several key findings regarding the state of ML applications in digital advertising. First, the authors identify data sparsity as a fundamental limitation, with 59.5% of the papers highlighting this challenge. This suggests that developing robust ML techniques for handling sparse, high-dimensional advertising data is a critical area for future research.

Another important insight is the lack of attention to critical issues such as fairness (0% of papers), sustainability (0%), and robustness (0%). This indicates that the current body of research has primarily focused on improving the technical performance of ML models, while overlooking important societal and ethical considerations.

Significance and Limitations: The significance of this work lies in its comprehensive mapping of the landscape of ML applications in digital advertising. By systematically reviewing the state-of-the-art, the authors are able to identify both the dominant trends and the critical gaps in this rapidly evolving field. This knowledge can inform the research agenda and help guide the development of more impactful, responsible, and holistic ML solutions for the digital advertising industry.

A key limitation of this study is the inherent bias in the literature review process, as the authors rely on the existing published research. There may be valuable industry-driven innovations or unpublished work that are not captured in this analysis. Additionally, the review spans a relatively long time period (2010-2025), which may introduce heterogeneity in the included studies and limit the ability to draw precise temporal insights.

Through the RSCT Lens: The paper's approach aligns well with the core principles of Representation-Space Compatibility Theory (RSCT). By focusing on the systematic review of the scientific literature, the authors are effectively assessing the "representation quality" (R) of the existing ML applications in digital advertising. The comprehensive coverage of dominant techniques, optimization metrics, and industry sectors provides a clear understanding of the current state of the art.

However, the paper's RSCT compatibility score of κ=0.55 suggests that additional context is needed to fully integrate the findings into the existing knowledge base. The R=0.00/S=0.00/N=0.00 scores indicate that the paper does not provide a clear signal on the core research questions, the consistency of the findings, or the level of irrelevant or contradictory elements. This is likely due to the inherent limitations of a literature review, which aims to synthesize a broad range of studies rather than present a focused, coherent contribution.

To improve the paper's RSCT score and pass the κ-gate (≥0.7), the authors could consider several strategies. First, they could delve deeper into the specific ML techniques, their performance characteristics, and the underlying mechanisms driving the observed trends. This would enhance the "representation quality" (R) by providing a more granular understanding of the technical approaches. Second, they could analyze the stability (S) of the findings by examining the consistency of results across different contexts, such as industry sectors or advertising formats. Finally, they could identify and discuss the potential "noise" (N) factors, such as the biases in the literature or the limitations of the review methodology, to provide a more holistic assessment of the field.

By strengthening the RSCT dimensions, the authors can position this work as a valuable contribution that not only maps the current landscape but also offers a deeper, more contextual understanding of ML applications in digital advertising. This would further enhance the paper's ability to integrate with and build upon the existing knowledge in this rapidly evolving domain.

Paper Details

• Authors: Jesús Alfredo Apaza-Cáceres, Liz Alessandra Vilca Paja, Andre Pillihuaman Huachaca, Diego Alonso Pedraza Choqu

• Source: arXiv

• Published: 2026-03-05

This analysis was generated by the Swarm-It RSCT pipeline using Claude.