Between 70% and 85% of the healthcare budget in OECD countries is spent treating chronic patients. Yet, behavior and lifestyle are at the root cause of nearly 80% of these chronic diseases and they are at least in principle preventable. Generic measures to achieve behavior change and prevent disease have proven to be not successful. Giving advice that is not timely, not actionable, and not personalized leads to low compliance of the user. Exactly because we are all different, and evolve over time, a key to success will be achieving personalization well beyond what is offered by today’s wearables and APPs. To achieve such personalization, we are creating digital phenotyping methods, which combine vast personal physiological information from diverse custom created wearable sensors, smartphone information and contextual information to learn individual behavior as well as habits and triggers. This will be the basis for giving the right actionable recommendations to the right person at the right time. Such highly perceptive and just-in-time feedback contrasts with today’s mainly time-based recommendations that are at best location aware and are not based on longitudinal nor personal physiological data. As the domain of preventive health is vast and diverse, we currently focus on three pilot applications: personalized stress management (for healthy people as well as patients), smoking cessation and eating behavior. For each application, in our living labs we conduct large-scale (i.e. 1000 persons), long-term (weeks or months) trials to account for the variability among people and over time. This research brings together multidisciplinary expertise and leverages clinical collaborations with psychiatrists, behavioral scientists and psychologists.