Weighing up attribute value
Conjoint analysis has some similarities to comparison shopping. It asks respondents to consider a range of descriptions, known as profiles, each with strengths and weaknesses in different attribute areas.
For example, imagine you’re an amateur photographer looking at these 3 camera profiles:
|Product 1 – CrazyCam||Product 2 – WackyCam||Product 3 – Camerama Deluxe|
|15 mega pixels||17 mega pixels||21 mega pixels|
|Weighs 150g||Weighs 200g||Weighs 400g|
|Focus from 1cm away||Focus from 5cm away||Focus from 5cm away|
Which model you choose depends on your priorities, but none of these is obviously the ‘perfect’ camera. If you want to take very high-resolution shots, but also to focus in on tiny objects, both Product 1 and Product 3 are potential choices. You need to make a trade-off according to what’s most important to you.
Based on the respondent’s choices, the researcher can assign the utility for each of the attributes – i.e. how much they value this attribute relative to the others.
Setting up your conjoint analysis
When designing your survey, you need to select which attributes to present to the respondent, and also what ranges of these attributes you’ll work within.
Conjoint analysis variants
There are a wide range of conjoint methods available, allowing you to collect and assess your data in different ways and to different levels of detail.
Depending on the variant, respondents may be asked to choose between profiles, to rank them in order, or rate them on a scale. They may be shown a complete range of profiles to select between, or sets of two at a time.
Which one you choose will depend on how many attributes you are working with, your data collection method, and how robust your results need to be.
Self-explicated conjoint analysis
This type of conjoint analysis is pre-built on the Qualtrics Product Experience platform. It produces utility scores for attributes by presenting the different levels to respondents separately, rather than as part of profiles.
Respondents are asked to eliminate any attribute levels they wouldn’t ever accept in a product. Then, the remaining levels are evaluated for desirability. Finally, the most desirable levels of the attributes are ranked according to importance.