Feedlot steers (n = 180) representing 11 sire-breed groups were ultrasonically measured for fat thickness (FTU) and longissimus muscle area (LMU) at two 60-d intervals during the feeding period and four 21-d intervals corresponding to serial slaughter dates to predict carcass retail yield parameters. Two fat trim levels, 8 and 0 mm, were used to calculate percentage of trimmable fat (FAT8P and FAT0P) and retail product percentage (RPD8P and RPD0P) for each carcass. Regression techniques were used to evaluate best-fit equations that explained variation in retail product components. When FAT8P, FAT0P, RPD8P, and RPD0P were regressed on USDA yield grade (YG), R2 values ranged from 75 to 76% (P < .001). Comparatively, when live animal predictors of YG components (FTU, LMU, and final live weight) were used as the independent variables, equations predicting retail yield had R2 values of 61 to 65% (P < .01). Equations using final FTU as the independent variable explained 58 to 64% (P < .001) of the variation in FAT8P, FAT0P, RPD8P, and RPD0P. Equations with FTU, LMU, and either WT, AGE, marbling, or quality grade resulted in R2 values similar to those with only FTU, indicating the strong influence of fat on retail yields. These results indicate that ultrasonic predictors explained about 10% less variation in retail product percentage than did carcass measures.