I was reading the section about summed pricing in "Becoming an expert...", and there was a remark about that the price attribute becomes correlated with other attributes that contribute to the total price, and a considerable random shock is needed to reduce this correlation. I guess this is because in this case Price is added on top of generated design.
But then I thought that maybe other studies that don't involve summed pricing could benefit from applying random shock on predefined price levels?
If, for example, we have a pricing study with Price varying from -20% to +20% of the market price, and first define our levels (as if for part-worth coding) as -20%, -10%, ..., +20%, then generate a good design, and after that apply random shock within +/-4% to each level, we'll be able to keep Price more or less independent of the other attributes while testing many more price values.
This way it should be possible to see if there are price thresholds which may differ brand to brand (given enough data) and also disguise the highest and lowest price that may be cause psychological bias.
Do you think this approach may work or am I overlooking something?