Ok, i`ll respond using your numbers to keep it organized:
1.) I can`t argue with the fact that it may take too long, if a site doesn`t get too much traffic. But i personally see no harm in exposing 5% of your site visitors (assuming you get enough traffic) to a variation and an original using something like AB in order to get a quick idea of which may perform better. Technically you can use multivariant testing in a serial manner and test 3 or more variations at once.
2.) If multivariant testing is done correctly you are finding the best combination of those different factors while testing your variation. Keep in mind that a testing methodology does not have to be used strictly to its original purpose. As an example, AB testing is actually quite dangerous and can cause inconsistencies if used the way it was originally intended (hence why others were developed). But if you apply AB testing in a serial manner (testing one element at a time), you side step a lot of the issues. Similarly you can apply multivariant testing in a manner that makes the testing more accurate.
3.) If your target customers are one`s wanting a quick opinion based on a variety of potential customers comments, then i think you`re good. My original point is that i am not sure how many online marketers behave this way. Your idea is pretty close to that of focus groups, which are quite helpful and are used in the initial "pre" stages of marketing. However, focus groups are mainly only useful for offline marketing attempts, as things are more tangible. But with online where the marketing material is so dynamic and the inclusion of dynamic material, a popular choice of an advert through your program may in fact behave poorly once placed in a dynamic environment, where the opinions were developed in a vacuum/independently. Online marketers have often found that it is the combination of page elements that work, and there is a danger in testing elements alone/by themselves.
4.) I feel tools, like website optimizer, can easily be used as a simple test bed for starter ideas as much as anything else, which i`ve already explained above.
5.) I completely agree on this point. All of my complaints thus far have been its usefulness for online. TV and anything offline... i think the idea is useful. Its just mainly the dynamic nature of anything on the internet that makes opinion polls not as useful
as other methodologies. As an example, what if everyone loved logo A over logo B, but when actually placed into its intended environment, it can be seen that logo A completely clashes with the color scheme, and so logo B may be better suited than A initially seemed. While the opinions matter, they are not getting the complete picture, and so, it can be argued that the result from their opinions is of less value compared to other methodologies.
Though the internet marketing market is huge, so i wonder if there is a way you can alter your system, or perhaps an addition, so that it cater to the online world more. I`ll think about it and post back with any ideas.
Then again, perhaps i am just being too serious about "initial" ideas. But to me, your sites main purpose boils down to one thing, making a choice. And if that choice is not made in the most optimal manner, even if its just the initial idea elimination stage, it would not seem as if some of the choices based off of your site may not be the most optimal ones all the time.
And again, i too would be interested in seeing if you agree with my rebuttal of your reply
Edit: One idea may be to have the ad, logo...etc for some online place, displayed within its environment. So you can display the element by itself, to show the people what to look for, then show them the element in its destined environment.