This study aims to predict the success of crowdfunding campaigns using the Kickstarter dataset and various machine learning models, including XGBoost, CatBoost, and LightGBM. These models, though powerful, are inherently black-box in nature, necessitating the use of SHapley Additive exPlanations (SHAP) to elucidate their predictions. To further understand what modifications can lead to successful campaigns, we generated contrastive scenarios and counterfactual examples. We then evaluated these counterfactuals using user-centric metrics such as Actionability, Relevance, and Impact, creating a comprehensive DataFrame. A combined score from these metrics facilitated an overall assessment, enabling the ranking of counterfactuals based on their usability and interpretability for stakeholders, including campaign creators and investors. This holistic approach not only demystifies machine learning predictions but also provides actionable insights to optimize crowdfunding campaigns.