Which Flavor is Right For You?
Conjoint Analysis has become one of the most powerful quantitative methods in Marketing Research. Conjoint Analysis answers the question of which attributes are important to consumers and how important they really are.
Over the years, many different “flavors” of Conjoint have evolved. Each “flavor” has its own distinct advantages and disadvantages. So how do you know which “flavor” is right for you? Let’s first describe the main types of Conjoint and then we’ll help you understand which one is right for you.
Terms
Before we get started, it’s important to understand a few terms.
- Attribute: An attribute is a general feature of a product or service. (i.e. size, color, price, speed, etc. )
- Level: Each attribute is then made up of specific levels. So for the attribute color, levels might be red, green, blue, and so on.
- Part-worth utility value: Researchers can derive a number (from the responses) which describes how valuable each level is, relative to the other levels around it. This value is known as the part-worth utility value. High part-worth utilities are assigned to the most preferred levels, and low utilities are assigned to the least preferred levels.
Full-Profile
In a Full-Profile Conjoint study, the respondent is shown a single profile of all attributes at the same time and is asked to rate the profile by their preference or likelihood of purchase.
Since Full-Profile conjoint studies ask respondents to consider all attributes at the same time, it is only appropriate for measuring a limited set of attributes (generally about six or fewer).
Partial-Profile
Partial Profile conjoint differs from Full Profile in that it shows only a few attributes at a time. Partial Profile is used when the number of attributes and the levels within the attributes are large. Due to the large number of possible combinations, every respondent does not see every different combination of attributes and levels.
Using an experimental design, profiles are designed in such a way that the occurrences of attributes are evenly distributed. In other words, each attribute is shown the same number of times as every other attribute. By controlling the attribute pairings, the researcher can estimate the respondent’s utility for each level of each attribute tested using a reduced set of profiles. Using the attribute “color” as an example, you can measure the individual weight of red, blue, green, or any of the colors included in the design.
Choice-Based (CBC)
CBC has become a very widely used conjoint technique. CBC tries to imitate the purchase process for products in competitive markets. Instead of rating or ranking product concepts, respondents are shown a set of product profiles and asked to indicate (i.e. “Choose”) which one they would purchase.
In a CBC exercise, an experimental design is often implemented to reduce the number of profile sets shown and help avoid respondent fatigue.
Adaptive Choice (ACA)
Adaptive Conjoint Analysis (ACA) is designed for situations in which the number of attributes/levels exceeds what can reasonably be done with more traditional methods (such as CBC or Full Profile Conjoint). In an ACA study, a computer interview at the front-end customizes the experience for each respondent. The study starts by asking the respondent to rank the levels under each attribute in order of preference. The idea is to eliminate levels of attributes that the respondent would never consider.
ACA concentrates on those attributes that are most relevant to the individual respondent and avoids information overload by focusing on just a few attributes at a time.
ACA is often used for product design and segmentation research, where the number of attributes is greater than six. ACA is not a good approach for pricing research, as it tends to understate the true weight of price. When people are asked to rank the levels under the attribute “price” in a silo (considering no other attributes), they tend to choose the lowest price. Nobody wants to pay the higher price, right?
So which “flavor” is right for you?
A few key characteristics can help determine which Conjoint method is most appropriate for your situation.
Number of attributes
If the number of attributes being studied is high, then ACA may be an approach worth considering. In the case of a lower number of attributes, CBC could be preferred.
Desired real-life purchase simulation
You want to align your Conjoint exercise with how consumers will interact with your product in a real-life purchase environment. For example, if the product message you’re testing will be on grocery shelves next to competitors, you want to simulate a competitive purchase environment using CBC.
Desired Analytics
There are several approaches to the analysis of conjoint data. For example, Hierarchical Bayes (HB) algorithms can be used to model individual respondent effects as opposed to aggregate approaches. This can be important when the part-worth utility scores do not follow a normal distribution. Interaction analysis is used to identify attributes which are not independent – when 2 or more attributes combine to jointly impact the respondent’s purchase decision. TURF analysis (used to identify the smallest number of items which maximize total reach) and cluster analysis (looking for groups of people who respond similarly) are examples of back-end analysis approaches which can be useful to “push” conjoint data further for specific issues or desired outcomes. Knowing if specific downstream analysis approaches will be used can have an impact on the type of conjoint approach used – the structure of the data must support the desired analysis.
Start with the problem you’re trying to solve
The best place to start is by identifying the problem you’re trying to solve. From there, different conjoint “modules” can be matched together to create the best solution for your problem.
These are only a few of the Conjoint “flavors” available. There are a multitude of flavor variations in use today, so it’s sometimes hard to keep track of. In the end, it comes down to identifying the “flavor” that best fits your problem.
Leave a Reply