On-farm feed dry matter techniques determine DM by difference between original feed weight and a dried weight. Oven DM techniques, though, have been documented to volatilize more than water, leading to underestimated feed DM measures. The objective of our work was to evaluate if feeds, dried through commercially adopted DM techniques, differ in total volatile compound content relative to undried. Corn (n = 14), grass (n = 5), legume (n = 15) and small grain silages (n = 14) were collected, divided into equal subsamples using a riffle-splitter, vacuum-sealed, and frozen until analyzed. Subsamples were thawed and then handled according to 5 different drying treatments; undried (CTL), on-farm type forced-air oven dry, 60 min (KOS), 50C for 48h forced-air oven dry (OV), freeze-dry (FD) or sequential microwave-NIR (LAB). Following treatment, and to assess DM technique non-water losses, undried and corresponding dried samples were analyzed for volatile fermentation products by HPLC (lactic, acetic, propionic, butyric, succinic, and formic acids, and ethanol). Each constituent was expressed as a % of DM, using sequential microwave-3 h 105°C oven dry as a standard DM. Fermentation products were then summed to determine total volatile compounds (TV). Technique difference from CTL represents significant non-water loss and a DM measure error. The resulting data were not normally distributed and were log-transformed before being evaluated using the Fit Model procedure in SAS JMPv11.0. Feed, drying treatment and their interaction were treated as fixed effects and assessed using backward-elimination. Significance was declared at P < 0.05. Feed type and drying treatment were significantly related to TV. Results presented here are converted back to % of DM. The TV means were compared using Tukey’s test, finding legume and corn silage (5.1 and 4.9) differed from small grain (3.6), which differed grass silage (1.80). The TV was the greatest for CTL (4.43) and was not significantly different from LAB, KOS, or FD (4.18, 3.67, and 3.13, respectively). The CTL differed (P < 0.05) from OV (2.7) while KOS and FD did not differ from OV. Results suggest OV underestimates feed DM.