Abstract Autogenic fluvial dynamics, including river avulsion, influence the distribution of channel sand bodies in alluvial deposits. Over long timescales, autogenically organized avulsions can generate stratigraphic patterns such as clusters of sand bodies when avulsions preferentially return to previous channel locations, or evenly spaced sand bodies when avulsions preferentially fill topographic lows. Consequently, quantifying stratigraphic patterns may provide an avenue for reconstructing paleoavulsion dynamics from ancient deposits. Several quantitative approaches have been used to quantify the degree to which channel-belt deposits are distributed randomly, evenly, or with clustered patterns; however, to date, there are only a few examples where these metrics have been applied in outcrop studies. Here we present a quantitative analysis of stratigraphic architecture in the lower Williams Fork Formation (Cretaceous, Colorado) to quantify the paleoavulsion pattern in this interval. A spatial-point-process statistic (the K function) and the compensation statistic are applied to stratigraphic data mapped from a terrestrial lidar digital outcrop model. Both analyses show random channel-body distributions and random basin filling at short (less than 200 m) spatiotemporal scales, which suggests that lower Williams Fork channels avulsed randomly. To evaluate the sensitivity of the K function to different degrees of stratigraphic organization, we use a two-dimensional (2D) geometric model to build synthetic stratigraphy with different degrees of sand-body clustering. Model results show that the lower Williams Fork data set should be of sufficient size and resolution to detect strong clustering signals, if they were present. This type of sensitivity analysis is helpful for comparing results of spatial-point-process analyses among outcrop examples with confidence. The random paleoavulsion pattern inferred from lower Williams Fork stratigraphy in this locality contrasts with previously published analyses that show qualitative clustering at larger scales; however, these results are not incompatible if avulsions remained clustered regionally over long timescales.