The development of science and technology had a strong impact on all branches of engineering, including geodesy and the possibility of acquiring and processing measurement data. The best example of this is the Terrestrial Laser Scanning, whih can perform measurements in the form of a multi-million-point cloud. The cloud, representing places and objects, becomes a spatial database. The current problem in engineering is no longer data acquisition and processing, but information excess and redundancy. The solution to this problem is optimisation, which is the process of reducing the amount of data. It should implement its assumptions in such a way as to remove or reduce unnecessary information without losing the information presented by an object. This issue is particularly important in the process of using point clouds in 3D modelling at various levels of detail. The appropriate levels of the LOD0–LOD4 model require a different type of data: on the one hand ensuring the obtainment of the appropriate accuracy class in the study, and on the other hand, the data source should not contain too detailed information that is unnecessary for the study, which makes the work harder and slows it down due to the need to operate on a huge amount of redundant information Therefore, the purpose of this paper is to determine the scope of work on the optimisation of the point cloud in order to adjust its number and quality to the needs of individual LODs. The results of the work allowed to determine the scope of data unification for the respective precision groups of 3D models generating.