Point clouds collected by laser scanners have been used in various architecture, engineering, and construction (AEC) projects. Effective uses of laser scanners on job sites require reducing the interferences of data collection with field activities while ensuring the data quality. Existing laser scan planning methods focused on visibility analysis with limited considerations about the level of detail (LOD) of collected point clouds even though LOD is a function of factors other than visibility. This limitation leads to insufficient LOD of certain features in point clouds for decision support. This paper presents a laser scan planning approach that integrates geometric feature clustering and data quality checking methods for addressing this limitation. The inputs of this approach include an as-designed model integrating LOD requirements of geometric features. The algorithm first clusters geometric features to reduce the computational complexity of scan planning. It then uses a sensor model of laser scanning to generate a "feasible space" for scanning each cluster of features. After that, it assesses the number of features covered at candidate scanning locations and pinpoints scanning positions covering more features with less scans. The results in a case study show that this approach outperforms manual scan planning in data's LOD.