A monitoring study was carried out in Gumara watershed, upper Blue Nile basin, with the objective of evaluating livestock water productivity (LWP) using a life cycle assessment method. Sixty two smallholder farmers were selected for the study implemented between November 2006 and February 2008. Data on crop and livestock production were collected to allow assessment of livestock water productivity. Study sites were situated in three different rainfed mixed crop/livestock farming systems; barley/potato based system (BPS), tef/finger-millet based system (TMS), and rice/noug based system (RNS). LWP was found to be significantly lower (p < 0.01) in RNS (0.057 USD m−3 water) than in TMS (0.066 USD m−3 water) or in BPS (0.066 USD m−3 water). Notably, water requirement per kg live weight of cattle increased towards the lower altitude area (in RNS) mainly because of increased evapo-transpiration. As a result, 20% more water was required per kg live weight of cattle in the low ground RNS compared to BPS situated in the upstream parts of the study area. Cattle herd management that involved early offtake increased LWP by 28% over the practice of late offtake. Crop water productivity expressed in monetary units (0.39 USD m−3 water) was higher than LWP (0.063 USD m−3 water) across the mixed farming systems of Gumara watershed. Strategies for improving LWP, from its present low level, could include keeping only the more productive animals, increasing pasture productivity and linking production to marketing. These strategies would also ease the imbalance between the existing high livestock population and the declining carrying capacity of natural pasture.