In the highly competitive and dynamic fast fashion industry, accurately forecasting customer demand is crucial for efficient supply chain management and maximising profitability. This research presents a novel data-driven demand forecasting framework leveraging an integrated deep learning model—specifically, a combination of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The study utilizes data from H&M, focusing on the top-selling product type to ensure relevance and reliability. Considering limited data scenarios is a prevailing issue in the industry, the present study introduced lag features to capture temporal dependencies. The proposed CNN-LSTM framework is evaluated against traditional statistical methods (ARIMA) and single-model deep learning approaches (CNN and LSTM). Performance metrics such as Root Mean Squared Error (RMSE) and the coefficient of determination (R2) demonstrate the superior accuracy and stability of the CNN-LSTM framework, with significant improvements in the training and testing phases. This research highlights the robustness and effectiveness of the CNN-LSTM framework in handling limited and nonlinear data, providing a valuable tool for demand forecasting in the fast fashion industry.