Problems Faced by Digital Marketing Agencies Due to AI Analytics
Jaligama, Sunitha (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025061623099
https://urn.fi/URN:NBN:fi:amk-2025061623099
Tiivistelmä
Artificial Intelligence (AI) has revolutionized digital marketing practices greatly by automating, optimizing, and predicting performance reporting. The growing dependence on AI-driven insights has, however, brought about a serious challenge: the mismatch between automated performance labels and campaign results. This research aimed to investigate and assess the occurrence and consequences of such "insight mismatches" in digital marketing contexts.
The most fundamental problem solved was the inefficiency and lack of transparency in AI-powered campaign analysis, creating strategic threats for agencies and eroding confidence in data-driven decision-making. The problem is significant because it has the power to lead to the misallocation of resources, inefficient campaigns, and client dissatisfaction. The study compared AI-allocated labels (e.g., "High," "Moderate," "Low") against actual performance results based on dimensions such as click-through rate (CTR), cost-per-click (CPC), bounce rate, and conversions through a structured dataset of 1,000 synthetic but realistic marketing campaigns with Python-based data analytics.
The study scope was limited to general campaign and audience segment types, ruling out reverse engineering of internal AI algorithms. Systemic trends were interpreted using statistical visualizations like mismatch matrices and correlation heatmaps. The analysis depicted a mismatch rate of 25.9% with the highest discrepancies in engagement-oriented campaigns and audience segments such as Fashion and Healthcare.
The findings of the thesis indicate that AI insight systems, though effective, are not entirely trustworthy as independent decision-making tools. The findings support hybrid intelligence models, integrating automated systems with human monitoring, and suggest post-AI auditing processes to improve decision accountability. The suggested methodology provides an actionable model for marketing agencies to authenticate AIderived insights and enhance campaign integrity.
The most fundamental problem solved was the inefficiency and lack of transparency in AI-powered campaign analysis, creating strategic threats for agencies and eroding confidence in data-driven decision-making. The problem is significant because it has the power to lead to the misallocation of resources, inefficient campaigns, and client dissatisfaction. The study compared AI-allocated labels (e.g., "High," "Moderate," "Low") against actual performance results based on dimensions such as click-through rate (CTR), cost-per-click (CPC), bounce rate, and conversions through a structured dataset of 1,000 synthetic but realistic marketing campaigns with Python-based data analytics.
The study scope was limited to general campaign and audience segment types, ruling out reverse engineering of internal AI algorithms. Systemic trends were interpreted using statistical visualizations like mismatch matrices and correlation heatmaps. The analysis depicted a mismatch rate of 25.9% with the highest discrepancies in engagement-oriented campaigns and audience segments such as Fashion and Healthcare.
The findings of the thesis indicate that AI insight systems, though effective, are not entirely trustworthy as independent decision-making tools. The findings support hybrid intelligence models, integrating automated systems with human monitoring, and suggest post-AI auditing processes to improve decision accountability. The suggested methodology provides an actionable model for marketing agencies to authenticate AIderived insights and enhance campaign integrity.