We propose a notion of r-rationality, based on the idea that the choices of individuals are guided by a single preference order, but rather than always choosing the very best available alternative, agents are content with selecting one of the r-best. This proposal provides a purely ordinal and relative version of the classical idea of satis…cing behavior. No level of satisfaction is exogenously fixed, agents are not full maximizers, but they follow a clear pattern of behavior whose consequences generate testable implications, which we fully characterize. The notion of r-rationalizability is further extended to individuals whose ordinal satisficing level may vary depending on the set of available alternatives: a similar characterization obtains. Since any choice function F is n-rationalizable, we can ask for the minimal r(F) for which F is r(F)-rationalizable, and take that value as a measure of the agent´s degree of rationality. We provide an algorithm to compute it for any given F. Special cases of ordinal relative satisficing behavior are shown to result from a variety of choice models proposed in the literature. Our notion allows for further flexibility, yet still provides precise restrictions on observable data.