Microgravity time series at active volcanoes can provide an indication of mass change related to subsurface magmatic processes, but uncertainty is often introduced by hydrologic variations and other noise sources that cannot easily be isolated. We empirically assessed seasonality and noise by conducting four surveys over the course of May-October 2017 at Yellowstone caldera, Wyoming. Yellowstone experiences frequent changes in the rates and styles of seismicity and surface deformation, but the mechanisms of these changes are poorly understood. Past gravity data from the caldera have yielded ambiguous results, possibly due to hydrologic noise. Given the strong visually observable changes in surface water and snow conditions over the course of our surveys, we expected to see significant variations in gravity. The net change in gravity, however, was less than 20 μGal at most sites, and there was no strong correlation with river and lake levels or snow conditions. Seasonal changes in gravity are therefore small compared to those that would be expected from magmatic activity, although they may be on the same order as those associated with Yellowstone's hydrothermal system. We did find that noise levels in gravity data were highly dependent on site characteristics, with bedrock sites away from trees yielding the lowest levels of noise, and thin concrete pads in forested areas the highest. These results can be used to plan future surveys at Yellowstone and to reinterpret past data, and they provide guidance in terms of best practices for repeat gravity work on volcanoes worldwide.