In our study, we hoped to evaluate the level of halo effect in our data and decrease it by randomizing brands within attributes in comparison to the control group where attributes would be randomized within brands. We found the halo effect is higher when brands are asked within attribute.
