Over the last decade, oil and gas well productivities were estimated using decline-curve analysis for thousands of wells as part of U.S. Geological Survey (USGS) studies of continuous (unconventional) oil and gas resources in the United States. The estimated ultimate recoveries (EURs) of these wells show great variability that was analyzed at three scales: within an assessment unit (AU), among AUs of similar reservoir type, and among groups of AUs with different reservoir types. Within a particular oil or gas AU (such as the Barnett Shale), EURs vary by about two orders of magnitude between the most productive wells and the least productive ones (excluding those that are dry and abandoned). The distributions of EURs are highly skewed, with most of the wells in the lower part of the range. Continuous AUs were divided into four categories based on reservoir type and major commodity (oil or gas): coalbed gas, shale gas, other low-permeability gas AUs (such as tight sands), and low-permeability oil AUs. Within each of these categories, there is great variability from AU to AU, as shown by plots of multiple EUR distributions. Comparing the means of each distribution within a category shows that the means themselves have a skewed distribution, with a range of approximately one to two orders of magnitude. A comparison of the three gas categories (coalbed gas, shale gas, and other low-permeability gas AUs) shows large overlap in the ranges of EUR distributions. Generally, coalbed gas AUs have lower EUR distributions, shale gas AUs have intermediate sizes, and the other low-permeability gas AUs have higher EUR distributions. The plot of EUR distributions for each category shows the range of variation among developed AUs in an appropriate context for viewing the historical development within a particular AU. The Barnett Shale is used as an example to demonstrate that dividing wells into groups by time allows one to see the changes in EUR distribution. Subdivision into groups can also be done by vertical versus horizontal wells, by length of horizontal completion, by distance to closest previously drilled well, by thickness of reservoir interval, or by any other variable for which one has or can calculate values for each well. The resulting plots show how one can subdivide the total range of productivity in shale-gas wells into smaller subsets that are more appropriate for use as analogs.