PrintConjoint Analysis
Introduction to Conjoint Analysis
Conjoint analysis tends to be among the most popular stated preference methodologies for developing new offerings, as it helps us understand exactly where the market finds utility in the offering. For example, we could apply conjoint analysis as yet another layer of understanding for our cognitive map, showing us strategic paths and combinations which are perceived to bring the most positive benefit to consumers. While having a consultant perform a full-blown conjoint analysis of the entire cognitive map would likely be prohibitively expensive and require significant sample sizes, conjoint analysis would be very well applied as we choose one or two strategic paths and begin to iterate offerings. In this topic, we will also see what powerful, yet simple software can do to automate both deployment and results of conjoint analysis.
Like many methodologies, conjoint analysis is by no means perfect, but where others attempt to offer a definitive "yes/no/how much" reading for our offering, conjoint analysis tends to be most helpful in showing us "hot spot" attributes and benefits in the offering, and how individual aspects of the offering can increase or decrease perceived utility and value. Please watch the following 4:09 video.
Video: What is conjoint analysis? (4:09)
One of the increasingly popular techniques being used in new product development is an analytical technique called conjoint analysis. The early academic work was done by Professor Paul Green at the Wharton School back in the nineteen-sixties and it's really come into very wide use now. It’s very useful for a number of reasons that I'll get to in a few minutes.
What is conjoint analysis? Well, I think the opening assumption is that if you ask customers, “Do you want this feature? Do you want this feature?” they want everything. But that's not the way it works in the real world. We have to make trade-offs between various features because we usually can't afford to have absolutely everything.
So what Green suggests is a technique where we give people combinations. We give people pairs or groups of products that are a combination of various features and ask them which one they prefer. And the example I always like to give is the following:
Let’s say you're gonna be flying to Paris from here in Boston and I'm gonna give you two options. Okay, option 1 is United Airlines. It's a Boeing 767. The seat width is this. The seat pitch, i.e. the distance in front of you, is this. The food quality is pretty good. The on-time performance is eighty percent and the price is 1450 dollars. So that’s option 1.
Option two is Air France. It’s an Airbus A340. The seat is a little bit wider but the seat pitch is a little bit less. The food quality - of course, it’s Air France so the food is terrific. The on-time performance is a little less good at seventy percent. The price is a little higher at sixteen hundred and fifty dollars
Which of those two do you prefer?
Now if I do that, what I'm doing is I am implicitly asking you to trade off a bunch of potential features or attributes in a product: airline; aircraft; seat width; seat pitch; food quality; on-time performance; and price. And if I create an experimental design and give you enough combinations of products like that I can derive out of that how much utility you derive or how much importance you place on each of those various attributes. And that’s referred to as conjoint analysis.
The term comes - it's a contraction of the words considered jointly. It's a somewhat complex thing to do although there are great tools that have made it quite a bit easier today.
The real benefits of conjoint analysis are two-fold. First of all, the variables can be categorical rather than continuous. So we could have Air France and United where there's no obvious higher lower interval - anything like that. They're simply categorical. The other thing is it's the only technique in all of market research that has been shown to be valid in evaluating price. That is, it answers the question of how much a customer would be willing to pay for a given feature or level of some attribute in a product. The old method of asking a customer, “how much would you be willing to pay for this feature?” is just shown to be completely invalid. The customer either doesn't realistically know or they’ll game the system. I mean no one in their right mind would, if a car dealer asked them, “how much would you pay for this car?” No one in their right mind would tell them the truth, and that's a problem in that old style of trying to get pricing.
Conjoint analysis presents price as a trade-off in a whole series of attributes. It works much, much better.
What is also useful about conjoint analysis is there are quite a few variations and tactics which can be applied to increase validity, ensure appropriate attention from participants, and improve the accuracy of overall findings. In some cases, these variations involve using actual product or credits as the incentive for accuracy, as the participant would actually be receiving the product they choose, or have the right to purchase the product at a certain price.
For one variant of conjoint analysis designed to increase experimental validity, please review "An Incentive-Aligned Mechanism for Conjoint Analysis" by Min Ding.
Options to Structure and Deploy Conjoint Analysis Easily
There are also quite a few reputable software packages that allow us to structure, deploy, and read the results of our conjoint analysis. For a very basic example, Sawtooth Software has made a simplified trial of their Choice-based Conjoint survey tool to allow potential users to try the package.
The video below gives you some feel for the power and usefulness of conjoint analysis software, in this case being the Number Analytics offering. Please watch the following 1:45 video.
Video: Number Conjoint Introduction (1:45)
With conjoint, one is trying to understand the preferences of consumers. And so the software allows you to recover the preferences of consumers at the individual respondent level by using fairly sophisticated hierarchical Bayesian methods when trying to understand the preferences of individual consumers.
Now we’ll look at the analytics of conjoint analysis. You can add the features and the questions and settings.
Now our questionnaire is generated from the conjoint design we just created.
Now we’ve got the results. The first shows importance. Then the utility for each attribute will be shown with different colors.
Check our summary of statistics for these attributes and importance.
Individual estimates.
Demographics.
So, the males dislike Dunkin’ and the Indians like Dunkin’.
Market simulator. So we can see that I changed price.
