We've been doing some low-level work in this area which may help people with some comfort with coding -- see https://github.com/censusreporter/acs-aggregate There is example code with lots of documentation that might help you get what you need, but you have to be comfortable at least copying and editing python code for now.

For Census Reporter itself, to support importing files and such adds not only its own complexity but also user authentication, secure user resource-storage, and such... so that's a lot of friction. But we're thinking about this.

We like this idea, although note that the American Community Survey doesn't go down to the block level -- only block group.

We've experimented with aggregation--there are open questions about how to aggregate medians, the GINI index, and a few other non-count estimates--but mostly, this is a user interface challenge. Our requests for more funding are focused on adding historic ACS data and representing change over time on profile pages and in data downloads, but arbitrary geographies are a close second.

For future conversation, there are also precision challenges with using GIS files to select Census geographies: for example, as implemented now, our query for "Census Tracts in Chicago" includes some from Indiana as well, because it's done with GIS instead of against the Census explicit delineation.

Another question: how to handle split geographies? Block groups are kind of small, so maybe it's based on a percentage of land inside the selection region? But the land in the region may not be the land where people live...