Patient data is like an iceberg – only a small fraction of it is visible at any one time, with the rest either spread across disparate systems, or hidden in a glacier of unstructured text. Under the surface there is all too often information that would change a patient’s treatment if it were known about. In a group of patients, there are patterns in the hidden data that could offer us new insights into the clinical services we deliver or into the fundamental nature of disease and therapeutics. In clinical medicine it is hard to systematically identify what does and doesn’t work, find the treatments most likely to be effective for an individual patient, and learn from patient outcomes, without the massive expense of a clinical trial. As the the volume of data flowing in medicine scales up with increases in genetic testing and monitoring devices we believe we can no longer afford to base treatment on the limited information that can be gleaned from a review of the patient record.