Selection of replacements for the sow herd is one of the most important facets in swine production. Although our current methods of selection are effective, there is still a large amount of variation in sow reproductive performance traits such as pigs per sow per year (PSY). Therefore, the objective of this study was to determine if on-farm phenotypic traits at 21 d postnatal (PN) or perinatal environmental factors could predict sow reproductive performance. Data were prospectively collected from 2,146 gilts born on a commercial sow production facility and included birth and weaning weights, vulva length and width at 21 d PN, birth and nursing litter size, days nursed, average daily gain from birth to weaning, and age at first estrus. Of the initial animals, 400 (17%) were selected for the sow herd, 353 remained after removal of animals culled for non-reproductive reasons. Animals were assigned to 1 of 5 reproductive performance categories based on observation of estrus or pigs per sow per year (PSY) across two farrowings: High Fertility (HF; 23%; n = 82; ≥26 PSY), Middle Fertility (MF2; 12%; n = 43; 20–25 PSY), Low Fertility (MF3; 15%; n = 54; <20 PSY), Infertile-Estrus (IFe; 10%; n = 36; estrus, no pregnancy), and Infertile-No Estrus (IFno; 39%; n = 138; no estrus, no pregnancy). Generalized linear model analysis indicated vulva width (P = 0.03) was related to PSY, however, it only explained 1.5% of the total variation in PSY. To determine if preweaning variables were predictive of gilt fertility outcome, animals were grouped as those that became pregnant (n = 179) or not (n = 174). Vulva width tended to be greater in fertile animals versus infertile (P = 0.07). Binomial regression analysis revealed a positive relationship between vulva width and gilt fertility, however, this relationship is not strong enough to make sow herd selection decisions.