Mitigation hierarchy is applied in many countries to avoid, reduce and compensate for impacts on the environment due to development projects. The design and sizing of measures thus depend directly on the assessment of impacts. However, this assessment is rarely framed by precise methods and is based on ad hoc evaluation that are carried out by environmental consultants by relying on own and more or less sophisticated methods. This article thus proposes to start from the ground up to analyse how impacts on biodiversity are assessed in practice, according to what method, and for what results. A statistical analysis was performed on 29 environmental assessments. We hypothesize that project size and number of species impacted are positively correlated with large impacts; and that measures and their costs are positively correlated with high impact reduction and low compensation. No correlation could be found for any of the factors studied. This result shows that there is considerable flexibility in the assessment of impacts, leading to inconsistent treatment of impacts. In the absence of a standardized method, this should encourage the adoption of sizing rules that compensate for uncertainties.