NBS8604: Marketing Analytics
Week 9 Tutorial: Conjoint Analysis
Lab Notes
Introduction
New management has taken over a local ice cream shop. They are collecting opinions from
their customers before relaunching the shop. As a marketing analyst, they have enlisted
your help in order to understand the following:
• How do customers make trade-offs between various attributes of ice cream?
• Which attributes (e.g., flavour, price) are most important to customers?
• How should we set the price if ice-cream when the product changes?
• How much market share would we receive if we launched a new product?
Step 1: Defining the Attribute List and Attribute Levels
For simplicity, we make a number of assumptions about the available ice cream. Five
attributes and the possible levels that the company can offer are identified as follows:
• Flavour: vanilla, chocolate, strawberry, or black raspberry.
• Packaging: cone or cup.
• Production/Fresh: fresh or factory- produced.
• Low fat: low fat or not low fat.
• Price: $1.50, $2.00, $2.50 or $3.00.
Step 2: Constructing Hypothetical Products
In total, there are 128 possible combinations/products! Fortunately, we do not need all of
them to run conjoint analysis. We will determine a subset of product profiles sufficient for
estimating attribute effects. SPSS provides a function to generate an appropriate subset of
products:
• Select DATA ORTHOGONAL DESIGN GENERATE. As Factor Name and Factor Label,
enter “flavour”, then click the Add button. This new design element should now appear
in the window. Click on “flavour” in the window and select the Define Values… button.
• In line 1 of the Define Values window and Value “1” and Label “vanilla”. In line 2, enter
Value “2” and Label “chocolate”. In line 3, enter Value “3” and Label “strawberry”. In
line 4, enter Value “4” and Label “black raspberry”. Now click Continue.
We then continue to specify the attributes and attribute levels for the other attributes in the
same way. They are as follows:
• Factor Name and Factor Label “packaging”. Values: 1 = cone, 2 = cup.
• Factor Name and Factor Label “fresh”. Values: 1 = fresh, 2 = factory.
• Factor Name and Factor Label “lowfat”. Values: 1 = Yes low fat, 2 = No low fat.
• Factor Name and Factor Label “price.” Values: 1 = $1.50, 2 = $2.00, 3 = $2.50, 4 =
$3.00.
1
• When all attributes and attribute levels have been specified, select the Options button.
This allows us to specify the minimum number of products in the subset. Enter “16” into
the Minimum number of cases to generate box. Click Continue.
• To save our new design, select Create a new dataset and enter “design1” as the Dataset
name. Now click OK.
The data set will appear in the new window. Click the button indicated in the figure below to
show attribute labels.
2
Alternatively, if you do not do this step you can use a design prepared earlier. In this case,
save and open the dataset icecream_design.sav.
Step 3: Collecting Customer Opinions
Once we have the design for our study, in the form of 16 product cards in this case, we need
to collect data. We do this by surveying potential customers. An example of the survey
design for this study is given here: [Link]
Now assume that we have surveyed 40 customers.
• Download and open the [Link] file.
Looking at the response data file, we see that customer 1 gives a rating of 5 (out of 5) for
ice-cream 1 (‘Pref1’), indicating that s/he would definitely buy this product.
Step 4: Estimating Attribute Effects
Download the following files:
• Profiles of 16 test products (“icecream_design.sav”)
• Consumer ratings for each product (“[Link]”)
• Programme code to run conjoint analysis in SPSS (“syntax1_mac.sps” or
“syntax1_windows.sps”)
• Product profiles for simulation (“icecream_design_simu.sav”) – for simulation
analysis
• Programme code to run simulation (“syntax2_simu_mac.sps” or
“syntax2_simu_windows.sps”) – for simulation analysis
Save all files on your Desktop!
3
We will now run some SPSS program code (or “Syntax”) to conduct the analysis:
• Select FILE OPEN SYNTAX. Open the file “Syntax1_windows.sps” or
“Syntax1_mac.sps” (according to your computer).
• Replace the “XX” in the file with the username of your computer. Mine is “stuar”.
• To run the analysis, select all the code, and click the green button.
• IMPORTANT: If your files are not saved on the desktop, modify the directory, accordingly
before running the analysis.
Customer One. How happy is customer 1 with each attribute level?
• To answer this, find the results for Subject 1: Customer 1 in the Output window. This
shows the level of Utility (happiness) attached to each product attribute. For example,
customer 1 appears to like chocolate more than vanilla.
How important is each attribute for customer 1 to make a buying decision?
• This is shown in the Importance Values table. For customer 1, price, low fat, and
freshness are most important, followed by flavour. Packaging is much less important.
4
How happy are all 40 customers (on average) with each attribute level?
• Find the Overall Statistics section of the Output and the Utilities table. Here, for
example, we see that customers appear to favour fresh ice-cream over factory, cheaper
prices, low fat ice-cream, cones, and chocolate.
How important is each attribute (on average) for all 40 customers in making a buying
decision?
• If we examine the Importance Values table, we see that, on average, price and flavour
are most important to all 40 customers.
• To generate a graph, double-click on the table. Then right-click and select Create Graph
and then Bar.
5
Implications for Pricing
Imagine we are selling an ice cream that is “Chocolate” in a “Cone”. It is “No low fat” and is
“Factory-produced”. We sell is for $2.50. We are considering changing this ice-cream.
If this product is changed, would the average consumer rather have a low-fat ice cream or a
price reduction to $2?
To do this, we must calculate the utility gain or loss between different product options.
These are summarised in the following table:
Attribute level Utility
No low fat -.091
Low fat .091
$2.5 -.181
$2.0 .238
• If we move from no low fat low fat, the utility increases from -0.091 to 0.091, a
difference (increase) of 0.182.
• If we reduce our price from $2.50 to $2.00, the utility increases from -0.181 to 0.238, a
difference of 0.419.
• Therefore, reducing price provides the higher utility and is preferred by customers.
How much MORE would the average consumer be willing to pay for a change from factory-
produced ice-cream to a fresh ice-cream?
To compute consumers’ willingness-to-pay we need to complete the following:
1. Calculate utility gain from “factory” to “fresh”.
2. Find the new price such that the utility loss between the original price and the new price
exactly equals the utility gain from “factory” to “fresh”.
3. The difference between the new and original price is the increase in willingness-to-pay.
Attribute level Utility
Factory -.116
Fresh .116
$2.5 -.181
$3.0 -.806
• Utility gain for moving from a factory to fresh product is from -0.116 to 116, a difference
(increase) of 0.232.
• The fall in utility from increase the price from $2.50 to $3.00 is from -0.181 to -0.806, a
difference (fall) of -0.625.
6
To see how much more we can charge, we examine absolute utility values.
• The utility value of moving to a fresh product from factory is calculated as a proportion
of the utility change from the price of $2.50 to $3.00.
• Therefore, we find that the proportion of utility is: 0.232 / 0.625 = 0.3712.
• Multiplying by the $0.50 price change we find: 0.3712 * $0.50 = $0.19.
• Adding this to the original price of $2.50 we find that: $2.50 + $0.19 = $2.69.
• Therefore, customers are willing to pay $2.69 for the fresh version of the product.
Predicting Market Share
Conjoint analysis can be used to predict the outcome when the market changes (e.g., when
a new competitor enters the market or when a current product is revised).
Suppose currently there are three ice-creams (products 1 to 3) in the local market. If we
plan to launch a new product (product 4). What is its predicted market share?
Our new product is “Raspberry” in a “Cone”. It is “Fresh”, “Low-Fat” and sold for “$3.00”.
• To begin, download and open the data set icecream_design_simu.sav. This data set
includes the market simulation products mentioned above.
7
• Select FILE OPEN SYNTAX and open the file “Syntax2_simu_mac.txt” or
“Syntax2_simu_windows.txt”
• Replace the “XX” in the file with the username of your computer. Mine is “stuar”.
• To run the analysis, select all the code, and click the green button.
• IMPORTANT: If your files are not saved on the desktop, modify the directory, accordingly
before running the analysis.
In the output window, find the Preference Scores of Simulations table and Preference
Probabilities of Simulations table.
• The results show that Product 4 will have the lowest overall utility, a score of 2.553, as
compared to competing products. Competitor 1 has the highest score at 3.050.
• Product 4 will have a market share of 21% if launched.