Clinical trials provide our core backbone of evidence but are restricted in eligibility criteria, as well as including less than 10% of our patients. Clinical practice data provides us with an opportunity to provide additional clinical evidence for treatment decisions and we collect it for all our patients. The more clinical practice data we can harness the more we can learn, especially for rare cancers and smaller population groups. Federated learning enables aggregate analysis and prognostic model development from datasets that are stored separately. The Australian Cancer Data Network enables automated de-identification of data to achieve a local research database at each institution and then uses federated learning to learn from this data. It has been initiated with radiotherapy datasets which contain granular imaging and radiotherapy dosimetry data. Using this approach we can learn from radiomics as well as clinical features from a large and varied cohort of patients both nationally and internationally to provide additional evidence for our patients and their clinicians.