The consolidation of streetwear as one of the most relevant segments in contemporary fashion, especially among Generation Z consumers, has intensified classic demand forecasting challenges in a context marked by short product life cycles, limited-stock drops and strong influence from digital media. In this scenario, brands that operate with small collections and planned scarcity face high uncertainty regarding the performance of individual items, which directly affects production planning and the financial results of each collection. This study aims to develop and evaluate a predictive model for the success of new product launches in the context of the Brazilian streetwear brand CienaLab, focusing on the T-shirt category. The research adopts an applied, quantitative and predictive approach, structured as a single case study. A historical database of past collections is used, in which each T-shirt is described by visual and production-related attributes (color, materials, types and sizes of prints) and by sales performance indicators in two time windows: the item s share of collection sales in the first 15 days and in the total sales of the collection. Based on these indicators, a categorical performance variable is constructed, classifying items into groups such as best performer, hyped and normal. The problem is formulated as a multiclass supervised classification task. Different algorithms are tested and compared, and a Random Forest model is selected and tuned using stratified cross-validation and grid search for hyperparameters. Model interpretation is conducted with SHAP values, allowing the assessment of the relative contribution of visual attributes to the predictions. In addition, a mapping procedure between CIELAB (L a b ) color coordinates and the brand’s discrete color categories is implemented, enabling use of the model during the product design process. Finally, the classifier is integrated into a Python-based graphical interface and validated on a recent CienaLab collection, discussing its potential as a decisionsupport tool for product planning in highly volatile streetwear environments.