Fecal nutrients and apparent total tract digestibility (ATTD) were predicted using near-infrared spectroscopy (NIRS) of feces collected from the pen floor or the rectum of feedlot cattle in 2 studies, and pen floor samples were assessed for their ability to predict NE, ADG, and G:F. In study 1, 160 crossbred beef steers in 16 pens (4 pens per treatment) were fed dry-rolled barley or wheat (89% of diet DM) processed at 2 levels. Study 2 utilized 160 crossbred beef steers in 20 pens (5 pens per treatment) that were fed dryrolled barley with 4 levels of barley silage (0%, 4%, 8%, and 12% of diet DM). Both studies fed steers to a target weight of 650 kg. Differences in composition of feces collected from the rectum and the pen floor of a subset of steers (3 to 7) were examined. Fecal pats from the pen floor of each pen were collected throughout the feeding period and composited by pen. Except for DM, which was higher ( 0.01) in pen floor than rectal fecal samples, there were minimal differences in fecal constituents between collection methods. In study 1, interactions between grain type and processing index ( ≤ 0.05) were observed, with a reduction in DM, OM, and starch and an increase in NDF and ADL concentrations being associated with more extensively processed wheat than barley. As grain was more extensively processed, ATTD of all nutrients increased ( 0.01). In study 2, fecal ADF and ADL linearly increased ( 0.01) with increasing silage in the diet, whereas fecal DM and N linearly decreased ( 0.01). Digestibility of all nutrients except starch linearly decreased ( 0.01) with increasing silage. Apparent total tract digestibility of GE predicted using NIRS was related to NEg of the diets as estimated by performance for the wheat-fed steers in study 1 ( = 0.58, = 0.03) and those fed increasing silage in study 2 ( = 0.43, < 0.01). Similarly, observed ADG could be predicted using NIRS for steers fed wheat in study 1 ( = 0.48, = 0.05) and silage in study 2 ( = 0.40, < 0.01), but G:F could not. Using NIRS of feces collected from multiple cattle off the feedlot pen floor demonstrated potential for predicting growth performance of finishing cattle. However, grain type and stage of maturity of the cattle impacted the predictability of equations. Increasing the sample size and sampling frequency may be necessary to improve predictions.