One example of why I like tests rather than just looking at it is when I see someone say, "this shot had more accurate colors, but the other looked better" and then they give the win to "the other". I understand WHY they're doing that, but it is technically incorrect. Now, I don't think that DXO and others have figured out the end all of camera analysis, but their methodology is definitely better than having some random person take random shots with their unknown skill level and make observations with their unknown knowledge level.