'Most practitioners use summary metrics (for example, mean, median, standard deviation, and so on) to communicate about distributions. However, you should usual

'Most practitioners use summary metrics (for example, mean, median, standard deviation, and so on) to communicate about distributions. However, you should usually examine much richer distribution representations by generating histograms, cumulative distribution functions (CDFs), Quantile-Quantile (Q-Q) plots, and so on. These richer representations allow you to detect important features of the data, such as multimodal behavior or a significant class of outliers.' I don't disagree with the advice above, but the median is a robust estimator of location, and if the median differs from the mean by an appreciable number of standard deviations, that is an indication of non-normality.