The object of this study is the strategy of online retail marketing campaigns, particularly in the context of utilizing a modified ID3 decision tree algorithm to improve predictive effectiveness regarding consumer responses. It addresses challenges in audience segmentation, campaign evaluation, and market adaptation, while also tackling technical issues such as overfitting, prediction errors, and data imbalance. These challenges often hinder businesses from accurately identifying and targeting potential customers, leading to inefficient marketing strategies and resource allocation. The dataset was split into 80:20 and 70:30 ratios, and the model was tested across decision tree depths from max_depth 1 to max_depth 20. The highest accuracy occurred at max_depth 6, ensuring optimal computational efficiency. However, increasing tree depth led to declining accuracy and rising computational costs, highlighting the risk of overfitting. Key factors influencing consumer response include income, education level, and recent company interactions. These variables help determine purchasing behavior and engagement levels, making them crucial in refining marketing strategies. Class imbalance introduces bias, affecting model performance by favoring the majority class while underrepresenting minority groups. The modified ID3 model outperforms ID3 Shannon, offering better precision for the majority class but lower recall for the minority class. Limiting campaign offers to one or two improves consumer responsiveness and prevents information overload. A data-driven marketing strategy ensures promotions align with consumer preferences and market trends. The developed model enables businesses to better target campaigns, increase conversion rates, and optimize resource allocation, ensuring an effective balance between tree depth and model accuracy