Part 1 of this tutorial showed you how to prepare for choice based conjoint analysis by setting goals and designing a questionnaire. Now it’s time to take a look at how to analyze the results from this questionnaire and how you can further use the results for marketing purposes.
SAS & Conjoint Analysis
Having translated the choice design obtained in Conjoint Analysis Tutorial 1 and send it to the respondents, it’s time to analyze the data. To compute the utility of the hypothetical products for (each of) the respondents, the objective is to obtain (the individual) part-worths of the product attributes. This will also be useful for building a choice simulator. Your data will look somewhat like this:
| Respondent # | Choiceset # | Choice | Attribute 1 | Attribute 2 | Attribute 3 | Attribute 4 |
| 1 | 1 | 1 | 1 | 0 | 1 | 0 |
| 1 | 1 | 2 | 1 | 1 | 0 | 0 |
| 1 | 2 | 2 | 0 | 1 | 1 | 1 |
| 1 | 2 | 1 | 1 | 0 | 0 | 1 |
(For the choice-column, 1 refers to the alternative that is chosen, 2 symbolizes the alternative that was not chosen. For this data it means that respondent 1 chose alternative 1 given the alternatives that were presented as choice set 1. For choice set 2 he or she opted for alternative 2. )
To get from the raw data to part-worths, SAS has several macros that will come in handy. The ”Proc Print” macro will print the dataset. ”Proc Phreg” will fit a multinomial logit model on the choice sets. Make sure you specify these macros to match your research and objectives. Among the output from these macros you will find a table with columns labeled as parameter estimate, standard-error, chi-square and several others. This is an important table since it contains information on the importance of the different attributes. The parameter estimates are the part-worths. They can be used to create a choice simulator, which will be showed below. In this table you can also find ”ChiSq” which represents the significance of the product attribute. The rule of thumb is that this level should be below 0,05. Or in other words, there should be less than 5% chance that the results occurred by chance.
Choice Simulator
One way to use the results from your survey would be to create a choice simulator. This can be done in Excel. The idea behind the choice simulator is to estimate market share based on utility derived from different products (here read products as ”combination of product attributes”). There are three steps to obtain these market share forecasts:
- Define your product and that of the competitors.
- Calculate utility of each of these products.
- Use logit rules to compute choice shares from the utilities.
- Experiment with different levels of attributes for your product to see what works best for your market share.
Defining the product and that of the competitors consists firstly of gathering product information. Which product attributes and levels do they have? Code these levels into dummies in the same way that you have done when creating the conjoint survey in SAS. Now that you have the products standing at the ready, it’s time to compute the utilities. Multiply the dummy value of each attribute of the individual products with the part-worths of the corresponding attribute. The part-worths can be found in the SAS output under the name ”parameter estimates”. For each product, sum these part-worths times levels. This value equals the utility for the product.
There are several rules for determining which product is the most desirable. You could simply say, everyone chooses the product with the highest utility. But it’s safer to use a logit choice rule. Use Excel to compute e (Eulers number) to the power of the utility of the product. (Excelcode: =exp(U), with U representing the number that correspond to the utility of that product). To get to the market share for a product, you need two things: the e to the power utility of the product, and the sum of these numbers for all products. The market share is simply the re-calculated utility divided by the total re-calculated utility.
Try to experiment with different attribute levels. Can your current product be re-designed to gain a larger market share? What attributes should your product in development have to establish a healthy starting position on the market? Use your choice simulator to experiment and come up with an optimal product offer.
That in a nutshell is how you perform a choice based-conjoint analysis.
The post Conjoint Analysis Tutorial (2) appeared first on ODMGuide.com.
