Conjoint analysis uses an experimental design to create balanced, uncorrelated profiles to show people. For example, you can break down a laptop into component parts, like Brand, Screen Size, Hard Drive Size, etc., with different values for each attribute. You then combine these to create many profiles and show them to the respondent typically either one at a time, in paired comparison, or show several and ask which is best. It's not really well suited to show everyone the same thing and then ask for feedback about that thing. You would need to, for example, create different advertisements that systematically vary (music, background, gender of lead, things like that) and then show people different variants of the advertisement and gather their feedback.