Conjoint analysis
Also called choice-based conjoint, CBC, discrete choice, trade-off analysis
Asking people to choose between whole offers that differ on several attributes at once, then working out how much each attribute and level drove the choice.
Ask people to rate price, storage and support one at a time and everything is important. Make them choose between whole offers, each a price, an amount of storage and a kind of support, and they have to trade one against another, as they do when they buy. Over many such choices, each respondent seeing their own random set of offers, a statistical model recovers the part each level played.
The usual model is a multinomial logit. Its part-worth utilities say how much each level pulls a choice towards an offer, on one scale, so a price rise can be weighed against a feature. Importance is each attribute's range of utility as a share of all of them: how much that attribute could swing a decision within the levels you tested.
Results only cover the levels you showed. Importance depends on the range: test prices from $10 to $50 and price will matter more than if you test $10 to $12.
Worked example
Utilities for price of +0.9, 0 and −0.9 and for support of +0.3 and −0.3 give price a range of 1.8 and support 0.6: price accounts for 75% of the importance, support 25%.
Where it goes wrong
Too much in one choice. Past about two dozen details per screen, people simplify, picking on one attribute and ignoring the rest, and the model learns their shortcut instead of their trade-offs.