The retention of trees bearing tree‐related microhabitats (TreMs) has become an important means of conserving biodiversity in production forests. However, we lack estimates of TreM formation rates and evidence on factors driving TreM formation. Based on the observation of 80,099 living trees from 19 species groups in Europe and Iran, we estimated the probability of TreM occurrence on trees and the associated rate of first TreM formation as a function of tree DBH, management, tree species group and random site effects. We built a separate model for each of 11 TreM groups. The hazard rate of first TreM formation (defined as the probability of formation of a first TreM forming on a tree that is known to have none, during an infinitesimal DBH increment) increased with DBH for some TreM groups like breeding‐woodpecker‐hole, rot‐hole or root‐concavity, indicating an acceleration in TreM formation during tree growth. However, it decreased with DBH for TreM groups like bark‐loss or dendrotelm, indicating slower formation on very large trees. Most TreM groups had reduced formation rates in managed forests (last logging less than 100 years ago) compared to unmanaged forests (no logging for at least 100 years), with the exception of dendrotelms. No general difference appeared between broadleaves and conifers but early‐successional species tended to have different TreMs than mid‐ and late‐successional species. Abies, Alnus, Betula, Fagus, Prunus, Quercus, Sorbus, Tilia and Ulmus displayed high formation rates for six TreM groups or more. Variability among sites was considerable. Synthesis and applications: The rate of formation of tree related microhabitats (TreMs) varies greatly among TreM groups, tree species, locations, tree diameters at breast height and forest management. The high rate of formation of some TeM groups on small trees implies that tree retention for biodiversity should concern trees of all sizes and start as soon as thinning operations have occurred. Biodiversity conservation should value not only forest stands and trees that already have many TreMs, but also those where the likelihood of future TreM formation is high due to species, maturity or local environmental conditions. The addition of quantitative models of TreM formation to forest stand dynamics simulators is necessary to better take into account biodiversity conservation in forest management.