The First Step to a Successful Conjoint Study: Creating the Attribute List

So, you’ve decided conjoint is for you. Now what?

Creating your attribute list is the most important step towards a successful conjoint study. This is what your respondents will be evaluating in your survey, and these are the items you are interested in investigating.

In a conjoint study, respondents evaluate product profiles that are constructed from the features in your attribute list. A typical conjoint question might look like this:

Setting up your attribute list correctly is imperative to getting the right data you are seeking.

What is an attribute?

An attribute is a product characteristic that can have multiple levels. For example: brand, processor speed, RAM size, and monitor size are different attributes of a laptop computer. Each of these attributes has multiple levels. Dell, Sony, and HP are three such levels of the attribute for brand. Here is an example of an attribute list for investigating different computer options:

Defining attributes and levels for your conjoint study 

Because the attribute list is used to construct the conjoint questions (or task exercise), it must conform to specific guidelines.

• Attributes should have independent meanings. For computers, “brand” is a completely different characteristic from “processor.” Having two attributes like “Dual core processor” and “Quad core processor” is inappropriate. Those attributes have overlapping meanings. They are both processors.

In a conjoint study, product profiles are constructed by using one level from each attribute. So looking at this example, the problem of attributes with overlapping meaning becomes clear:

• Attributes must have more than one level. A key tenet of conjoint analysis design is that respondents value certain attribute levels over others. We cannot understand their preference for the “Dell” brand if it is not evaluated in relation to the other brands

• Attribute levels are mutually exclusive of each other. Remember, only one level within each attribute is utilized when constructing product profiles. If you have an attribute called “add-ons,” the product profiles will have either “carrying case” or “webcam,” but not both. See the figure below. If you would like to test for both “carrying case” and “webcam” within a product profile, there are ways to get around this mutually exclusive rule.

  

• Be specific about your attribute descriptions. Don’t say “A lot of RAM, ” say “4GB of RAM”

• Make sure your attribute list contains the full range of attributes and levels that you are interested in investigating. Typically, these are the key features that impact a person’s product decision.

• Keep your attribute levels confined to your target market. Utilities derived from conjoint analysis reflect relative preference of the items tested. A lineup of laptops that include HP, Puget Systems, Olivetti and Lanix will overwhelmingly favor HP and overstate the importance of brand. Take HP out of the list and you can see how brand would become a less important attribute. It is no longer as strong a factor when respondents select their choice of laptops in the task exercise.

• Ideally, conjoint studies have no more than 12 attributes and 7 levels within each attribute. More can be incorporated if properly set up.

Creating a good attribute list is the first step towards a successful conjoint study. If you need help evaluating your attribute list, don’t hesitate to contact us. At Decipher, we offer the full range of capabilities to employ ratings-based and choice-based conjoint techniques in your next web survey.