Good project management is an important support for decision-making in the management of an organization. Today, there is a trend in having transparent and objective decision-making procedures, often based on some calculations like in risk assessment for project prioritization. These methods sound very attractive and have indeed a real potential, but they can also reveal a kind of fake objectivity. Based on real data science projects, Rodolphe Dewarrat will present the potential and limitations of data-driven decision-making.