Welcome everyone to the session of the
marketing research and analysis. Today, we are going to discuss on something,
in the last session we discussed about multi-dimensional scaling and its importance,
today we are discussing something similar, a tool, a technique which is highly utilized by
marketers. In fact you do not see too much of theoretical research in this but its practical
utility is extremely high, extremely high.
So, this technique that we are going to talk
about is known as conjoint analysis. So, conjoint analysis is a technique where a marketer
identifies the possible combinations of several attributes and then which attribute is going to
be, or which combination is going to be the most, having the highest value and most sought
after combination in terms of the from the customer s point of view. So, to do this conjoint
analysis is highly used. Now there is a choice, for example let us see there are three
Colas available in three different packs.
One is you can see, there are three different
price levels, maybe the quantity is different so when a person decides to buy or a customer
decides to buy how does he make a choice? So today you must be seeing companies coming up
with something called decoy effect and all, where they do nothing else but they make
several combinations of attributes and then try to place in front of the customers.
So, that the customer then identifies that which one he likes most and most of the, and the
one which is mostly in demand that becomes more a product more in demand for the marketer and so
that he can maybe streamline his production and his supply chain according to the demand of the
market. So, what is this conjoint analysis?
It is a technique developed in the early 70 s, it
measures how buyers value components of a product or service as a service bundle, product bundle
or service bundle. That means you are not talking about, when you are talking about a product you
are not talking about individual compartments, like the product, the value of the cost of
the product or the amount of its size or its volume or something or nothing like that.
Rather when a person or a customer selects a product he finally selects it on the basis of
a combination or a bundle of benefits and he makes his own comparison like a tradeoff,
may be, he likes A but in a combination of A and let s say X he does not like it and
suddenly he prefers something which is less preferable than A which was B but with
the combination of let s say x or y.
So, the person s a person s ability to finally
select a product changes with his total holistic thinking that means what individually you would
have liked, maybe it does not happen when you take in totality the entire, all the attributes. So,
what is the definition? It says conjoint is joint/ together or combined so it s a combination
so and no product has individual features, all products have individual features but
they are available in the combination.
So, in the combination finally how does a
person select a combined attribute or benefit of attribute or combined bundle of attributes,
so features considered jointly, marketers say it is as features considered jointly.
So, how does conjoint analysis work? We vary the product or service features
which has the independent variables to build many product concepts, so what does
the marketer do, the marketer combines in different combinations all the attributes
which are the independent variables.
Basically for example in case of a let us
say a cool drink, a soft drink the price, the packaging, the color, the volume or the amount
of soft drinks, this could be the four features, four attributes. Each attribute has got
several levels, you can understand like factors, each factor has several variables.
Similarly here each attributes has got several levels, so it could be 500ml, it could be 250ml,
1000ml, similarly, and in color it could be green, it could be red, different colors. So we ask
the respondents to rate or rank rate or rank or choose among the subset of the product
concepts which are the dependent variable, so the dependent variable is choice, your final
choice, so which one what is your ranking or what is your choice among those various combinations
which one which combination do you like most?
Let us say, he says a particular combination
of materials, he says is no.1, and similarly some other he would say is 2. Let me explain it
with an example. Let us say the attributes could be on the basis of price, volume, let us say
color, I am taking only three, so price is let us say Rs 100, Rs 50, let us say Rs 75, volume
is let us say 1000ml, this is let us say 400ml, this is let us say 750ml.
This is let us say color is green, red, and blue. Now there could be several combinations
and now at least there are 3, 3, 3, so 3, 3, 3 is let us say 27, so 27 combination possible
but then the question is how many, which one is most preferable for the consumers? So, so the
choice, he makes a choice, let us say the price, let us say he makes a particular combination
let say this combinations his choice and he says as rank 1, this is rank 2, this is rank 3.
So, whatever the ranks, so these are the ranks which he is giving finally, so based on the
respondents figure, evaluation of the product concepts we figure out how much unique value.
Or utility or which is termed the part worth functions which we say as unique value or utility
or part worth function each of the features added; that means a particular feature had it be
not there, what would have happened?
When it is present what it is happening, so its
presence or absence will impact the utility value of the entire function. So, regress dependent
variables on the independent variables and estimated betas equal to the part worth utility,
so the part worth which I said this utility is the ultimate utility and the ultimate utility
is the summation of the part worth utility.
So, let us see this case, what is so good
about conjoint, which more realistically which product will you prefer, you want a 210
horsepower and 17 mpg or 140 horsepower with 28 mpg? Now suppose somebody says he will choose the
left one 210 that means he is preferring power, but if he chooses the right one then
he is preferring the fuel economy.
So the point is on the basis of this marketers
can identify what customers want and accordingly they can place their products. Rather than ask
directly whether you would prefer power over fuel economy will present realistic trade off
scenarios and preferences, infer preferences about the product choices of the respondents.
When respondents are forced to make difficult trade off, when you are not given an option and
you are asked to make a choice, there you have to make a real difficulties in life, do you
want to watch a movie or you want to play a game of cricket so you cannot do both because
you will only be given to choose one, because rest of the time the student has to study.
For example the parent says you need to study so either you choose cricket or you choose
a movie, which one would you like to see? So there it is a difficult tradeoff for the
student. So we learn here what is true value, suppose the students says I would like to
play a game of cricket then it means that he values cricket over movies. These values are
associated with the specific and actionable attribute levels relevant to the problem.
Now let us see this case, this is a building a model, this is where conjoint analysis comes
into play, so the inputs are attributes, that various attributes, size, price, all these levels,
each attributes have several levels, respondents, their prior knowledge, external data, experimental
design, and finally conjoint method. Outputs are the utility scores for each level so what does the
score, the particular level, different levels the scores are available to the marketer.
Importance course for each attribute, so how important is a particular attribute and
finally the ability to perform simulations, that means if I bring in a new respondent can I
find out on basis of its, the part functions can I say this whether the new customer or the new
respondent would prefer the product or not.
So, defining attributes, attributes are
nothing but the independent aspects of a product or service, so brand, price, size, color
etcetera. How many attributes should you take, this is the important question, so the rule
of thumb is the number of attributes should be maximum up to 7, so you do not go more than
7, more than 6 or 7 as it says because if you take too many attributes with this number
of levels it becomes more complicated.
So, 2 to 3 is I think ideal, attributes
should be independent and mutually exclusive, there is no doubt, this should be clear
from each other separate from each other. Each attribute has varying degrees or levels, so
let us say in terms of price if you can see 1$, 2$, 3$, color, green, black, blue for example
are the three different levels, Each level is assumed to be mutually exclusive again, so that
a program has only one level for that attribute.
So, attributes are assumed to be mutually
exclusive, attribute add on features, level one, sun roof for example, level 2 GPS system, level
3 a DVD player, so if you define levels in this way you cannot determine the value of providing
2 or 3 features at the same time or none of them, so the question is how do you formulate
the attributes or the levels? Let us see.
For the solution for a 8 level attribute for
example if you can see, now this is the 8 level attribute so the features are, there are
no features, sunroof, GPS system, DVD player, sunroof and GPS, sunroof and DVD, GPS and
DVD, sunroof, GPS and DVD all together. So, all the 8 are different levels, similarly the
binary features they are you can see binary attributes, sunroof none or there, GPS system
none there, DVD player none there, so this is the presence or absence of a binary attribute.
So, this is how you formulate the attribute levels, don t include too many levels for any one
attribute, if you include it becomes complicated. So the usual number of attributes is 3 to
5 per level level per attributes, so you at 6 to 7 attribute maximum and 3 to 5 levels
maximum, so in a maximum if I can understand 7*5=35 there should be a combination for only
one case. Make sure levels from the attributes can combine freely with one another resulting
in utterly impossible combinations, without resulting utterly impossible combinations.
Now let us see this case, on the basis of cost there are three levels, brand there are
three levels, color there are three levels.
Suggestions for determining which attributes
and levels to include, how do you identify which attributes should I and which levels should
I include. Simple, talk to the stakeholders, the people who are involved, the customers, the
managers in the company, the different people through a focus group interview, we have done
focus group in qualitative research search of competitors websites, sales material that mean
your competitor and what they are doing now, so this will help you to determine which attributes
to include and which one not to include.
Now conjoint utilities that are called
part worth which I said, so numerical values that deflect how desirable different
features are. For example look at this case, now vanilla, chocolate, three price
levels are 25 cents, 35 cents, 50 cents, the utilities are 2.5, 1.8, 5.3, and 3.2,
1.4 so the higher the utility the better, why, because it simply works as the regression
function so this is like the beta weights.
So, if you have the higher beta weights
automatically the total utility, because the total utility is something like u =
summation of alpha, alpha is the part function, is the part part worth function, part
worth so if you have the utility if your alpha values are high the coefficients are high
automatically the total utility will be high.
So predict as market share for 35 cents
vanilla account versus 25 cents chocolate cone, this is how the conjoint analysis work
actually you can see, so the vanilla, let us go back, you have 2.5, chocolate is
1.8 right so 2.5 plus 35 cents was 3.2 so 3.2, that is equal to 5.7. For the chocolate it
is 1.8 from the chocolate part and from the 25 cents that is the price function it is 5.3.
Now look at it, although people do not, might be preferring chocolate but with the combination
of the price that 25 cents suddenly the whole utility value has drastically changed and vanilla
which was the preferred taste has gone down and the chocolate which was the less prefer taste has
gone up and the total utility has become 7.1.
So, the respondent chooses 25 cent chocolate cone
over the vanilla. So this can be repeated for the rest of the respondents and understanding
can be developed, how does the marketer, how does the respondent makes his
choice, makes his combinations.
So, this is a popular example we have brought
from, this is from Green and Wind is a marketing research book which is very, very popular
book, you can reserve, so this book is, I have used also myself when I was a student from
that day, a company interested in make marketing a new carpet cleaner wants to examine the influence
of five factors, what are the five factors, package, design, brand, name, price, a good
housekeeping seal and a money back guarantee. Now this 5 features he has taken right.
The three factor levels for package design, in each one deferring the location of
the applicator brush, three brand names, three price levels, you can see here, so the
package, the levels are A B C brand K2R, Glory, Bissell. Price, three again levels, seal, yes
or no if it is there, yes, no, zero, so it is a kind of binary, money yes or no, either money
back guarantee is there or it is not there.
Now there could be other factors and factor
level that characterized the corporate cleaners but these are the only once of the interest
to the management see the question is there could be several attributes but the question
is we can take into play all the attributes it becomes extremely complex so what are the most
important once we have selected them right.
So, these are important point conjoint analysis
you need to indentify the most important ones right you want to choose only those factors that
you think mostly influence the subject preference using conjoint analyzes you will develop a
model for customer preference based on these five factors so these five factors that we have
chosen now on the basis of this five factors we will see how are the consumers making a choice.
So, the first step in the conjoint analyze is to create a combination of a the factor levels
that are presented as a product profile to the subjects right since even a small number
factors and a few levels for each factors will lead to unmanageable number of potential profiles
you need to generate the representative subset known as an orthogonal array now what is that
mean so if you have so many combinations.
Then it could be a huge built up right and you
cannot do it so it is better to have something which is orthogonal means completely opposite not
opposite they do not meet that means parallel to each other right so the 90 degree to each other
so those representative only we will take those subsets which are exclusive extremely
exclusive mutually exclusive right.
So, to generate the orthogonal design procedure
which is orthogonal design what is done is we can create one right if you already have a active data
s you can either replace it or save the orthogonal design or create a orthogonal data file right
so I will show through a SPSS file how do you do it right theoretically you have understood key
that you have make several combinations right and then you can say which combination is the most
preferred combination right so one two three, one two three so several combination
27 combination can come right.
Let see this conjoint analyze so to do that go to
data orthogonal design and then generate right so I have in steps so that if ever you want to do
conjoint analyses you do not want you do not have to search for the steps right it becomes
easier for you to do it so this is on SPSS.
So, once you have suppose for the package for
example package so the label of the package is package design you name it and start adding
right so how does it look a package design this creates the item level package design select
this item ok and define the values here so one let s say one is A two is B three is C.
Similarly you can do it for the other factors also remaining factors like
brand price seal and money.
So, this is how it will look like finally you can
see this space right this is how it will look like so package design three brand name 1 K2R to 2
glory 3 Bissell price similarly 1 2 3 we have again three levels ok seal 1 is no 2 is yes money
back 1 is no 2 is yes so you can do this right.
And once you have created this data set right
so what you do is we do a random number you know for iteration we use it but you may
not you can kept it a default which is 2 lakhs right what is done is new data file
is created out here now this new data file of interest to us right yes but one thing is
important when you are creating new data.
And once you created a new data file please
understand that to for the validity of the test that what you have tested is valid and
highly is valid thing you have to split the data file into two parts the one part is where
you are doing the normal analyses and other part is called the hold out part as we are done in
other MDS also we have done MDS and other things so sorry so this is called the hold out case.
So, the data file will have now two data files one is holder and other one is the one which
are using right so once you have done it right then you can run the analyses.
So, let me go to the file so this is how the final file look like right the final
file will look something like this.
So, 1 2 3 4 5 ok so data we have already done
it am I going back so this is the how you have done it so listing of experiments so you have
taken the five factors lets go back once so now so you have your display or from the menu
choose the data orthogonal design display go to display and then this factors you take
it here right select package brand for factors right format is listing off for experiment.
Because it is an experimental design correct you are experimenting this is purely the experimental
design like ANOVA analysis of variance you are trying to experiment which combination is going
to be the most give you the highest worth utility or highest utility or part worth function
ok so orthogonal and click ok so the output resembles the look of the orthogonal design
as shown in the data editor one row for each profile with the factors as the columns so the
factors are these are the factors 1 2 3 4 5.
However the column headers are the variable
levels rather than the variable names that you see in the data ok fine so this is the
hold out case now when there is a hold out case that will be a foot not and it is written 6
and if you can check it right so this is hold out case so the SPSS gives you two files ok now to
display each profiles in a separate table what you can do is you can also see for each profile
individually for the each respondent you can see key whether how it is what is there response
right you can go for profile for subjects here can you see you can profile for subjects.
So, deselect listing for experiment and select profile for subjects and click okay.
So, this is what you will get so if you can look at this file the information for each product
profile is displayed in a separate table, a profile number 1 say 1 this is 1 package design is
1A, glory brand name prices 1.39 housekeeping sale he wants it money back guarantee is not important
similarly two package is a profile brand name is Bissell price is this one housekeeping seal not
required money back guarantee not required.
So, that process also you can do it for each
case right so by doing that and then you can finally may be through a subjective analysis
interpretation you can see and tell which is the one which is the which is mostly in
demand and you can make a interpretation out of it or right so otherwise what you
can do is you go the complete file right the complete file and find the utilities.
So, this table there is a table which gives you the whole statistic table and utilities this
table shows the utility or the part worth scores and their standard error for each factor level
higher utility once you get the utility function which are the beta weight, for example I have said
to you once you get the utility functions then you can from here you can calculate you know that
the final the dependant value or the dependant score right so high utility values indicate
greater preference so suppose for somebody says now for example here let s say package A.
I think there must be example let s see okay let s say package A then glory price 1.39 and seal yes
and A yes so let say automatically by putting in those values of utility function you can finally
find the total value or total score and this score you can compare and say whether it is high or low
higher utility value indicate greater preference because obviously the more the utility that means
the people are finding more value in it right so obviously the preference would be higher for
example look a this price now it s says there is inverse relationship between price and utility
higher prices corresponding to lower utility.
So, once the going up the utility value is going
down that is not the desirable thing so similarly if you can look at seal right so if there is
a seal then the utility function goes high the preference goes high for the respondent
and similarly money back guarantee if it is there still high but you can see right
there are some let us say for example the package B has got a very high positive
value right similarly you can use this values ok this is already I have explained.
Now let us look at this since the utilizes are all expressed in a common unit they can
be added together that the beauty so I said utility is equal to summation of alpha
right so now the alpha is the part worth ok for example the total utility of a cleaner now
look at this part the total utility of cleaner with package design B brand K2R price this
much and no seal of a approval or money back guarantee is how much this much let s see 1.867
now this is 1.867 B ok plus 0.367, 0.367 K2R.
This one, so it adds on it comes to 11.759 suppose
it says if the cleaner had the package design seal brand Bissell and come changes have been made
now how much is the total utility function is coming is 10.909 so now this was 11.759 and that
was that is 10.909 so obviously you can see from here that the utility function is growing with
the change in it so in conjoint analysis the most important is to you calculate the utility
functions right this utility sometimes you can use dummy variable coding also to use it.
But let me not get into it right now so but you need to calculate this coefficient and then
do it right so now what it is saying it has the range of the utility values for each factor
provide a measure of how important the factor was to overall preference factor with greater
utility play a more significant role then those with smaller ranges right so you can see this
suppose factor this factor has minus 2.23.
1.867, 0.367 brand is this is one ok this
much price yes you can look at the impact if you forget the negative and the positive sign
take the absolute value price plays a very, very, very significant role in impacting the
preference right similarly seal also has an impact but if you look at the brand hardly
it has got a very important role to play with ok so here the marketer can play with the
consumer and its wants and then accordingly he can decide so this is the final beta coefficient.
So, price point -5.542 seal is 2 money is 1.25 so these are the things that are very important
in conjoint analysis so once you can do that conjoint analysis helps you to identify which is
the best combination and which combination has got the highest utility function and then accordingly
that can be placed to the market or before the market so that the consumers can you know show
a greater demand and ultimately that also has an impact in doing a better forecasting as a marketer
and streamlining your supply chain accordingly, right, okay that is all for the day we have.
Thank you so much. Well thank you very much.
