This paper discusses the current state of development of on-farm cattle welfare assessment systems with special regard to the approach of Welfare Quality® that focuses on animal-related measures. The central criteria validity, reliability and feasibility are considered
with regard to selected welfare measures. All welfare measures incorporated into the Welfare Quality® protocol possess face validity, but for most of them construct or criterion validity as, eg shown for lameness, have not been demonstrated. Exemplarily the cases of qualitative
behaviour assessment and measurement of avoidance distance towards humans or social licking are discussed. Reliability issues have often been neglected in the past and require more thorough investigation and discussion in the future, especially with respect to appropriate test statistics and
limits of acceptability. Means of improving reliability are the refinement of definitions or recording methods and training. Consistency of results over time requires further attention, especially if farms are to be certified, based on infrequent recordings. Considering feasibility, time constraints
are the main concern for assessment systems that focus seriously on animal-based measures; currently they require several hours of on-farm recordings, eg about 6 h for a herd of 60 dairy cows. The Welfare Quality® project has promoted knowledge and discussion about validity,
reliability and feasibility issues. Many welfare measures applied in the Welfare Quality® on-farm assessment approach can be regarded sufficiently valid, reliable and feasible. However, there are still a considerable number of challenges. They should be tackled while using the
present assessment system in order to constantly improve it.