Data annotation is the process of labeling data to make it usable for machine learning models, enabling them to understand and learn from the input data. It is crucial for supervised learning as it provides the ground truth that models use to make predictions and improve accuracy.