Herbert Simon (1947) observed that human learning and decision making does not strive for optimal solutions but merely ones that are good enough (known as Satisficing). However, the absence of a theory of what precisely counts as good enough has limited the utility of Satisficing in the social and behavioral sciences. We draw insights from recent work in cognitive science that provides a precise criterion for good enough learning (termed the “Tolerance Principle” or TP) which describes how humans form generalizations based on limited or inconsistent information. To date, TP has only been defined for individual learning. Here, we show how TP provides a precise model of how social norms emerge in a paradigmatic coordination game frequently used in research on cultural evolution and collective intelligence - i.e., the Wittgensteinian name game as designed by Centola & Baronchelli (2015). We compare different agent-based models in their ability to predict norm emergence in experimental runs of the name game involving human participants. TP considerably outperforms models which assume that agents leverage optimal rather than good enough predictions for social behavior. We also show how TP provides a more accurate model of tipping point dynamics in the name game. Altogether, we show how social coordination follows the same simple formula for “good enough” solutions as individual learning, providing a novel synthesis of psychological and sociological constructs.