The elusive Caspian red deer Cervus elaphus maral lives at low densities in rugged forest habitats of the Caucasus and the south Caspian region, and its declining population requires urgent attention. We here address the precision and reliability of dung counts (fecal standing crop approach FSC) and camera trapping (random encounter model REM) for estimating its population size. We surveyed 36 km of strip transects arranged in systematic random design and applied 1585 camera trap nights of effort in the mountainous forest habitats of Golestan National Park, Iran. We also conducted a dung decay analysis of 80 samples. Dung decay rates were not habitat-specific and the mean time to decay was 141.8 ± 15.1 days, i.e. only ca 52% of the most reliable estimate available for red deer dung. Estimated deer population size and density from dung counts was lower (194 ±46 individuals, 0.46 ±0.11 individuals km^-2, 2012–2013) than from REM (257 ±84 individuals, 0.61 ± 0.20 individuals km^-2, 2011), but this difference was insignificant. Both these estimates confirm a sharp decline of the population from an estimated 2096 animals in the 1970s. Density estimates reached a stable level and were most precise at a sampling effort of 15 transects (FSC) and 1345 camera trap-days (REM). Our results confirm that FSC and REM can both be reliable for assessing populations of Cervidae.