If you use data from the U.S. Census Bureau, you have probably already noticed the radical change to the American Fact Finder tool. If you are frustrated by these changes, know that you are most definitely, not alone. The new AFF (or AFF2 as the Census Bureau advertised) no longer has the Turbo Tax-esque guided search feature that allowed users to build a query step-by-step. Instead, the new AFF thrusts users into a convoluted and overly-designed interface that makes it very difficult to figure out where to start.
Being fair to the Census Bureau, the old AFF could have used a facelift, but why change the mechanics completely? The old AFF prompted people to start at the most crucial starting point - CHOOSE GEOGRAPHY FIRST. Census data availability is always affected by the size of the area you want to study. At smaller geographies (tract, block group), less data is available due to confidentiality issues (and sometimes things just aren't collected at that lower level). If you are planning on using decennial census data or American Community Survey data on the new AFF, here's what you should do - pick your geography, then pick any topics you want to limit your search by (this includes specific population characteristics such as race because the Race and Ethnicity filter can be too narrow), then pick your dataset (also listed under Topics). If you're going to be looking at one table or just a couple tables at large geography, the new AFF isn't so bad. The problems start when you want to download multiple tables matching on geography to use that data with ArcGIS or any other statistical package.
Here is my advice to you - if you're going to use decennial census data or the American Community Survey, use Social Explorer instead. You will notice that it has a very familiar and friendly interface. It also has decennial census data back to the beginning of the decennial census, as well as some other goodies. It took me 3 minutes to read a downloaded file (all Baltimore City Tracts, multiple tables) into Stata. Seriously, 3 minutes. If I were to do the same within the new AFF (and I tried) I would have to remove all the excess text at the top of my file, rename all the variables to unique names, merge discrete tables together, and then write a program to read the file into a statistical package. If you're up for the challenge, I admire you - I chose the simpler path (are you listening Census Bureau??).