Energy flexibility of buildings is one solution that could help a better integration and an uptake of renewable energy sources. Energy flexibility consists in shifting some energy usages of buildings (e.g. heating, domestic hot water, wet appliances) according to the needs of the energy production systems. In order to better integrate the needs and constraints of occupants, indirect control strategies are foreseen to activate this potential. In this approach, the flexible loads are not directly controlled by the Transmission System Operator (TSO), but the control rely on signals sent and interpreted at the building level. Occupants will therefore interact with this signal, decreasing the level of confidence in the flexibility potential. The district scale increases the reliability by taking advantage of the coincidence of usages and the diversity of occupants. The objective of this article is to evaluate the sensitivity of the response of flexible buildings to occupants’ interactions. This indicator used to characterise flexibility is the decreased load during a 3-hour flexibility event at the end of January. Bottom-up simulations of buildings and occupants are performed for a new district composed of 337 dwellings located in France. The sensitivity analysis is conducted according to the Morris methodology, accounting for the diversity of occupants in terms of presence, behaviour, comfort and flexibility preferences. Finally, the influence of aggregation on the reliability of the potential is evaluated.