But in effect, each one of these numbers is a utility level for
each of the levels of the attributes listed on the right hand side.
So here we have a golf ball example, and one of the attributes is brand.
And then I have these different brand names, High-Flyer Pro, Magnum Force,
Eclipse Plus and Long Shot, those are the levels of the attribute brand and
their associated effect or utility is right over here on the left hand side.
And you can really interpret that as a happiness level.
So larger positive numbers mean more happy and then negative numbers mean less happy.
Now, in most software output that does conjoint analysis,
the data has to be scaled and I'm just going to point out
one issue with respect to scaling because it's important as we interpret the data.
The software will usually scale conjoint analysis output such that the sum
of the utilities within any given attribute sum to zero.
Now perhaps you can see that just by looking at this.
Now, if you have trouble seeing the data on this screen, we've also put a paper
version or electronic version of this in the resources, so you can grab that and
look at that while you're looking at this particular segment of the course.
And that's actually a very easy way to do it.
Then you can look down at the data and look at the example that I'm presenting
and kind of square the two, make sure they make sense to you.
But if you look at these effects,
over here for brand, you see 0.56, .429, negative 0.38,
negative 0.608, if you add those up they will add up to zero and
that is true for each of the other attributes.
This is a distance attribute, how far does the ball go.
And finally down here, we have a price attribute.
That's what we're really interested in and those effects or utilities add to zero.
Now some of you taking this course may have done some advanced work in statistics
or done a lot in terms of regression analysis and
you would be familiar with the term t Ratios and standard errors.
If you're not familiar with them don't worry.
All the examples that I'm going to do use only the effect or only the utility and
that's all you need to focus in on.
Conjoint analysis generally does give us output
that tells us something about the t Ratio which is a measure of
whether these effects are something called statistically significant.
Now, why do I not think that is important here?
Every output from every conjoint analysis that I have ever seen that has been
done in any kind of a reasonable way has statistically significant output, okay?
In fact, if you had statistically insignificant output,
that would mean that the attributes that you put in the conjoint
analysis were in no way related to the people's choices that they made.
That's very unlikely, okay?
Especially by people who have done a lot of conjoint analysis.
So the practical reality here is what we have all of this output,
just by focusing in on the descriptors on the right hand side,
the descriptions of the attributes and the levels and
by focusing in on these numbers right here, which is the happiness levels
we can really do the kind of interpretation we want to do.
Now, so what can we do with that kind of output?
We can do a number of different things.
We can determine which product people prefer.
And we can do that both for the population and for
different segments of the population if you've asked those segmentation questions
as part of the conjoint analysis.
You can look trade-offs among different possible features.
You can determine the rank order of attributes in determining choice
which attributes are very important and can we quantify that and
which attributes are less so in determining choice.
We can also, very important for this course, compute willingness-to-pay for
different design changes that we might be thinking about for our product.
And finally, we can do something called Propensity Modeling.
For each one of these items that I have listed here, I'm going to do an example,
and I'm going to start out with determining which products people prefer.
So, here's how this is done in a conjoint analysis.