<![CDATA[In the era of massive digital transformation, e-commerce platforms allocate substantial budgets across various digital advertising channels (e.g., Google, Meta, TikTok) to maximize customer acquisition. However, this channel fragmentation introduces a critical managerial challenge: Cross-Channel Advertising Cannibalization, where a single advertising campaign claims attribution credit for conversions that were actually driven by other channels or pre-existing organic demand. This study proposes a Machine Learning classification framework to detect and categorize e-commerce transactions into true incremental conversions or cannibalized conversions. Utilizing a comprehensive historical dataset extracted from 93 global e-commerce organizations over a continuous five-year period, we developed and compared several classification algorithms, including Random Forest and eXtreme Gradient Boosting (XGBoost) Classifiers. This model is designed to analyze multi-touch attribution overlap patterns, time-lag, and consumer interaction metrics. The empirical results demonstrate that the proposed classification algorithms can detect instances of budget cannibalization with high precision and recall, revealing that most Return on Ad Spend (ROAS) metrics reported by singular advertising dashboards are over-reported due to conversion overlaps. This research provides a strategic foundation for Chief Marketing Officers (CMOs) to mitigate budget inefficiencies and reallocate capital more accurately based on true incremental lift.]]>
