I am absolutely convinced Change Research is a fake pollster. The sheer number of polls they're putting out is immediately suspicious, since these each would be quite expensive to conduct were they real and they're putting out fifteen or more a week without a client sponsoring them. They've also missed races by huge margins abnormally often. It's statistically unlikely for them to have had huge misses this often were they a real polling outfit.
Their methodology, while imperfect, is incredibly inexpensive and has a fast response rate.
I understand that much of their statistical programming process is automated as well, which explains their fast turnaround.
They are attempting to do something that, by every measure, is new to the field.