The next step is now to actually collect the data.
So we have to poll our customers about how they think or
how they feel about the different product profiles that we have just created.
Typically we can distinguish between two different types of asking customers to
respond to different profiles that we have created.
One alternative is called the choice-based conjoined analysis and
you now see one example on the screen.
It essentially replicates very much decision making processes that we would
go through as consumers if you walk into a store or if you walk into a supermarket.
Typically we are confronted with different alternatives and
we have to choose one of these alternatives that are offered to us.
Or we basically say,
we're not interested in any of the alternatives that are given to us.
Another alternative would be to go for a rating based conjoint analysis, and
we see an example of that now on the screen.
It is maybe less realistic, but
typically I've seen that it actually produces really good results as well.
In step five of the conjoined analysis, we're now going to leverage and
use the data we have just collected and try to make sense of it.
As you can imagine, there are a multitude of different statistical tools available
in order to analyze the data.
Essentially what we're now trying to do is we're trying to
put a numerical value to the utility or the value perceived by the consumers for
the different product attributes.
And within each product attribute to each product attribute level.
That is actually going to allow us to then make predictions about the preferences
of our consumers.
And again, as we said before, to construct fictitious products.
And make predictions about how customers would choose one product over the other.
Let's now come back to our BMW example.
Let's assume that the BMW managers use a statistical model called
regression analysis in order to make sense out of the data.
Essentially what we have to do is we first have to establish something
like a baseline product, let's just pick something.
Let's say the baseline product for us is in terms of the brand,
being Audi, the baseline brand in terms of the body style, the sedan.
And in terms of the engine type, let's say a gas run car is essentially our, or
gasoline run car is the baseline as well.
In terms of price, let's assume as well that $20,000 would be the baseline.
Here we see at the end, once we analyze the data,
that essentially the value that we would get for that particular product is zero.
We would also see that anything that shows up positive is
clearly superior to this baseline product.
And hence would be preferred by consumers, or would be chosen by consumers.
Every product profile that comes up that is negative, or
is lower than the baseline that we have actually identified.
Obviously will be perceived as well as inferior,
and would be less likely consumers.
And as a consequence, if they would be offered a choice between a negative
product and a baseline product, they would clearly go for the baseline product.
Let's now go to one specific customer, and
to the analysis that we have drawn up for this customer.
What we see here is that apparently this customer favors a sports car over a sedan,
because the customer has a positive util of 2.21.
And remember we established the sedan as the baseline, which is essentially zero.
At the same time, we actually see that he or
she likes the sedan much more than the SUV.
Because the util showing up for SUV is a negative 2.5.
So in summary, we could say just looking at the body style that apparently this
particular respondent.
This particular customer prefers the most a sports type car,
second, a sedan type car, and least prefers actually an SUV.
Looking at the regression report,
we see here as well something called the intercept.
The intercept actually represents the value or
the utility that customers perceive when choosing and when going for
the baseline model that we identified before.
In this particular case, it's a 4.12.
What we can now do is we can actually construct different product profiles.
And we can choose, out of the different product attributes,
different product levels.
And use the information that is given in this particular regression report.
So let's look at one particular example.
Looking at this particular product profile, we see that the resulting utility
or value of received value by the consumers, it says 0.91.
And we actually see that this is perceived as inferior, or less attractive,
as the baseline profile that we have just identified before.
This takes us now to step six, interpreting the part worth.
BMW is now ready to use the information from the regression output in order to
determine the importance of the different product attributes for a respondent.
The basic idea behind this is that looking at each product attributes.
The difference between the highest part worth and
the lowest part worth should represent the importance.
Let's actually look into our example.
And let's calculate the importance for their brand, for the price, for
the body style and obviously as well, for the engine type.
To get a better sense of how the importance of one attribute compares
actually to the importance of another attribute.
Marketing researchers typically look into the relative importance of the different
product attributes.
Actually, it is quite simple as well to calculate them.
We essentially just take the importance of each product attribute, and
divide it by the sum of all the product attributes.
For this particular respondent, we see that body style seems the most
important product attribute, followed by price.
On the other hand, brand and engine type seem a little bit less important.
This now takes us to step seven, predicting choices.
BMW has now gained a full understanding of how this respondent values
the different product attributes.
The importance, he or she attributes to the different product attributes.
And how much the different levels of each product attribute are valued.
As a consequence, what BMW can now do is create different product profiles,
different alternatives, or fictitious products.
And can determine what is the overall value that is perceived by this particular
respondent for the different alternatives that are presented to him or her.
Let's actually look at one example.
Here we have three different alternatives.
Leveraging the information that we have seen before from the regression output,
we can easily determine as well.
What is the resulting utility for alternative A, B, and C?
Now we can try to make predictions as well about how this respondent would choose
between the three different alternatives that have been presented to him or her.
The simplest rule that we can actually use here is called the maximum utility rule.
As the name already implies, it just simply assumes that the customer's going
to choose that alternative that has the highest utility.
In this example, it will be very straightforward,
alternative A seems to be the one that gives the highest level of utility.
And hence, should be the perfect choice for this particular respondent.
We're now coming to the final step in our conjoined analysis process,
and as you can imagine, this is the step we have all been waiting for.
Overall, this is a course on pricing.
So we now have to see how a conjoined analysis can leverage in order to learn
more about prices.
And more specifically, about the willingness to pay of our consumers.