The objective of this retrospective cohort study was to clarify the effect of re-modelling the breeding unit on farrowing rate. The original study population included 63 sow herds that participated in the Finnish herd surveillance system. In these herds, the breeding units were re-modelled between 1995 and 2002. Information about the production system and other herd data for the last year before and the second year after re-modelling were available for 47 herds. The herds had an average of 68.5 (SD 45) sows per herd. Data were collected during one farm visit per herd. Initially, all herds had individual cages in their breeding units. The piggeries had either re-modelled their cages or switched to a group housing system with deep-litter, solid or slatted (partly or totally) floors. Farrowing rates were gathered from the national database, and data were tested with a linear regression model. Re-modelling did not have an effect on farrowing rate if initial reproductive performance of the herd was neglected. However, when farms were stratified according to performance prior to the beginning of study, namely into 'farms with initially high farrowing rate' (HF) and 'farms with initially low farrowing rate' (LF), the re-modelling had a significant effect on farrowing rate. In the HF farms, re-modelling caused a decrease in farrowing rate. Correspondingly, farrowing rate increased after remodelling in the LF farms. Farrowing rate was also influenced by mean parity of sows and by proportion of sows culled because of leg problems. This study could not show a profound impact of the design of the breeding unit on the fertility of the sow in the modern commercial production environment if the farm's reproductive performance before re-modelling was neglected. However, when initial performance was taken into account, re-modelling revealed significant, although varying effects on farrowing rate.