Abstract Biases in climatological and extreme precipitation estimates are assessed for 11 global observational datasets constructed with merged satellite measurements and/or rain gauge networks. Specifically, the biases in extreme precipitation are contrasted with mean-state biases. Extreme precipitation is defined by a 99th percentile threshold (R99p) on a daily, 1° × 1° grid for 50 °S–50 °N. The spatial pattern of extreme precipitation lacks distinct features such as the ITCZ that is evident in the global climatological map, and the climatology and extremes share little in common in terms of the spatial characteristics of inter-product biases. The time series also exhibit a larger spread in the extremes than in the climatology. Further, when analysed from 2001 to 2013, they show relatively consistent decadal stability in the climatology over ocean while the dispersion is larger for the extremes over ocean. This contrast is not observed over land. Overall, the results suggest that the inter-product biases apparent in the climatology are a poor predictor of the extreme-precipitation biases even in a qualitative sense.