Running a Conjoint Analysis

It can be daunting to setup a conjoint analysis for the first time. We recommend first running a simple study with only two attributes and two levels to familiarize yourself with the software, and then proceed to the full survey design.

1. Define product attributes

Conjoint analysis is based on the assumption that a product is a bundle of attributes.  Choose attributes that matter to the consumer and can be technologically modifiable. For example, a cookstove has the following attributes:

  • Price level
  • Color
  • Size of cookstove
  • Longevity / lifetime
  • Service level (as provided by retailer or manufacturer)

Find out what attributes you should use through focus groups, interviews, brainstorms, etc.

2. Select Number of Levels for Each Attribute

Each attribute can be further specified into different levels.  Make sure that the range is broad enough, but be careful as the higher the number of options used for each attribute, the more burden that is placed on the respondents. For example, we could define the following levels for our cookstove example.

  • Price level – $5, $7, $10
  • Color – red, yellow
  • Size of cookstove – small, medium, large
  • Longevity / lifetime – 1 year, 2 years, 3 years
  • Service level (as provided by retailer or manufacturer) – lifetime warranty, no service

3. Define hypothetical products

Choosing all combinations of attribute levels will generate too many products. Instead, aim for a subset of products that still represent enough levels.  If you have more than 6 different combination of features (e.g. 3 attributes, 2 levels), it’s easiest to use a tool to generate the product profiles.

4. Design and conduct survey

There are several ways to collect conjoint analysis data. The first is rank ordering all the products by writing out every possible combination on individual cards and then asking respondents to order their purchasing preferences from highest to lowest. The second is rating all products on a scale (e.g. 0-100).  The last is choosing the preferred option from two options, each option with pre-defined attribute levels.

5. Run statistical model

Input the data into your statistical software (note: this can also be completed through Microsoft Excel, but requires some data manipulation), and run the model. The model will allow you to estimate the importance of each attribute to understand the “weight” of price level, color, size, etc. to the users. Depending on the program, you can play around with the different attributes and levels to understand how to maximize utility for the customer.

JMP Software

Discrete Choice Designs

JMP(8) is SAS software designed for dynamic data visualization on the desktop. It is a powerful modeling tool that will allow you to construct graphs interactively, develop surveys and studies using the Choice platform, and more. Unfortunately, the software is geared towards corporate users and is expensive ($1595 for a corporate license) but you can download a 30-day free trial. There are also academic licensing agreements for 6 months or 12 months at a highly discounted rate.

In JMP, you will be creating a “Discrete Choice” design. JMP recommends that you create an example choice experiment to hone the survey design before administering the full survey. There is an excellent set of step-by-step tutorials available on JMP Help.  Under the index, select “Discrete Choice Designs” and then follow the instructions as written.