Imagine a man walking into a grocery store to buy a bag of coffee beans. In the coffee aisle, he is faced with myriad options. The different bags of coffee differ not only by brand, but also flavor, ground vs. whole bean, degree of roast, caffeine presence, and nation of origin. Somehow, he is able to psychologically sort through all of these options and make a decision about which bag of coffee to buy. Typically, researchers trying to understand this decision-making process would simply ask the consumer, "Why did you buy that bag?" However, his response might not be the most valid way of understanding his decision for several reasons.
First, he might suggest that the purchase was made because of a more socially desirable characteristic. For example, he might say, "I prefer a dark roast," but the true reason the purchase was made was because it was the cheapest bag. Second, he might overly simplify the decision making process. He might respond, "I prefer a dark roast," but that reason might only explain 40% of his decision—the other 60% might be split between his preference for unflavored, caffeinated, whole bean, and coffee grown in Ethiopia. He probably wouldn't mention these other attributes because the type of roast is the most important factor for him. Finally, and most importantly, he might not actually know why he made the decision to buy that particular bag of coffee. Even though he might give a response, the response might only be a post-decision rationalization for his decision making process.
One solution to these potential issues with asking consumers about their decision making process for purchases is conjoint analysis. Conjoint analysis (also called "policy-capturing" in other fields) is a technique that can be used to statistically deduce a person's decision making process. The technique has been used successfully to investigate a number of disparate research questions: Understanding consumer decisions in marketing, determining the optimal type of advice, and even understanding where hunters would go to hunt moose!
A conjoint analysis starts with the researcher determining a number of attributes that a decision maker might consider for a particular research question. In the coffee example above, the attributes would be type of roast, country of origin, price, flavor, etc. The researcher would then create different levels of each attribute. For attributes that have categories (for example, type of roast), the different levels of the attribute are easy to determine (for example, 3 levels: light, medium, and dark). For attributes that are continuous in nature (for example, price), the researcher needs to make a decision about the fixed point of each level. One major consideration is that unrealistic levels of an attribute can drastically bias the results. If, for example, price has two levels that were $0.01 per bag or $50.00 per bag, the results would probably show that price is incredibly important, explaining upwards of 90% of participants' decisions. This result, however, is merely a function of a poorly constructed attribute, not an individual's true decision making process.
Once the attributes and levels are set, the researcher creates scenarios by combining one level of each attribute. In a "fully-crossed" design, participants would then be asked to rate how likely they would be to buy or use all scenarios, with each scenario presented one at a time. In our coffee example, a single scenario would be one bag of coffee that has the following attributes: High price, light roast, hazelnut flavor, beans from South America, and ground beans.
Once the data are collected, researchers can statistically determine the relative importance of each attribute. For instance, we might find that for the average consumer, the most important factor in buying coffee is caffeine content followed closely by price. The other factors don't matter nearly as much. Further segmentation analysis might show, however, that there is a subset of "coffee snobs" who care very little about price, and instead are deeply interested in whole bean bags and country of origin of the bean. In this way, conjoint analysis allows researchers to more deeply understand how different factors impact an individual's decision making process.
For researchers planning on designing a conjoint analysis, there are several important considerations to keep in mind::
- Be considerate of the burden being placed on your participants. There is no way to get around the fact that a conjoint analysis is pretty repetitive for participants. To help with this, make the task as easy as possible. Use bulleted lists to describe scenarios instead of paragraphs. Make relevant changes between scenarios in bold. Consider the pros and cons of using a design where each participant only responds to a subset of scenarios.
- Make sure that each level of each attribute is realistic. As was previously mentioned, when the levels of one attribute are not at all reflective of reality, it could invalidate all the results of the study.
- Other methods to investigate decision making are important, too. Although conjoint analysis is a useful tool for understanding an individual's implicit decision making process, asking participants explicitly why they made a decision is important. When collecting both types of data, a complete picture can emerge. If the implicit and explicit results are consistent, researchers are able to make strong conclusions about the importance of different attributes; if the results are different, researchers can conclude that the reason participants say they made a decision is different than their implicit decision making process.
When implemented correctly, conjoint analysis can be a powerful tool for understanding the psychological mechanisms behind an individual's decision.