# Compare normalized utilities across individuals?

Hi there!
I have seen multiple similar questions, but I wanted to just make sure before I continue with my analysis that what I am doing is permitted within the rules of conjoint analysis.

My question: Are we able to compare utilities across individuals if we normalize them?

For example, Attribute A for participant 1 is 80 importance and the same attribute is 20 for another participant. Within those attributes are 3 levels. We care about one of the levels. Can we multiple 80 by that individual level and 20 by that person's level to normalize the data for that level?

EX: 80 * (20.3 utility of level) = normalized level score
and 20 * (30.4 utility of level)= normalized level score

If not, please let us know how we can compare part-worth utilities across individuals! We really appreciate your insight.

Strictly speaking, the only quantity that has any real meaning is the difference between two RAW utility scores within the same attribute.

So if you have raw utility scores and you have one level with utility 2 and one with utility -2, the difference of 4 is meaningful.

If a different respondent had a corresponding utility difference of only 3  between the same levels, then the second respondent you could judge to have less modeled difference between those two utility scores.

The exact number, e.g. 2, is meaningless, but differences matter, and they are only strictly comparable on the raw scale or uniformly applied functions of the raw scores (e.g. a logistic transformation to turn utilities into probabilities like we do in the simulator).
answered Jul 12, 2019 by Bronze (3,920 points)
I just wanted to follow-up and ask if I use zero-centered diffs, are those normalized?

I saw in a research paper that the whole population's utility levels were averaged, but I thought you could not compare utilities across people for a level.

Please let me know what you think, this is the research paper:
The effect of extrinsic attributes on liking of cottage cheese.