Stand spatial structure plays a key role in forest management, and in particular the random structural unit, comprising random tree and its neighbors, largely determines forest stability and its productivity. However, how the spatial structure of the random structural unit affects productivity remains unclear. The study focused on four larch forest types from Hebei and Shanxi provinces, China: 35‐year‐old larch (Larix principis‐rupprechtii) plantations (35LP), 39‐year‐old mixed larch–birch (Betula platyphylla) forests (39LB), 58‐year‐old natural larch forests (58LN), and 73‐year‐old mixed larch–birch–spruce (Picea asperata) forests (73LBS). The forest spatial structure index (FSSI) was employed to comprehensively evaluate the stand spatial structure. Additionally, the uniform angle index was used to discern whether the stand structure units were uniform, random, or clumped. A regression model was used to elucidate the effects of species mingling, diameter dominance, and crowding on the productivity of random trees. Results showed that the FSSI varied among the stand types, ranking as 35LP < 58LN < 39LB < 73LBS. The values of distribution frequency and the percentages of basal area increment (BAI) for random trees were above 0.5% and 40% in most stand types, respectively. The diameter dominance of random structural units in four stand types showed a significant negative correlation with the BAI of random trees, whereas the mingling and crowding of random structural units in the 73LBS displayed a significant positive correlation with BAI of random trees. Thus, increasing the size of random trees, as well as mingling and crowding in random structural units, can facilitate the formation of a rational stand spatial structure, thereby enhancing the productivity of larch forests.