We have always felt that one fascinating thing about market research is that it brings together mathematics and human behavior. Mathematics is beautiful, elegant, and abstract. It is much like art. While human behavior—in this case, that of consumers, buyers, and markets—is messy, contradictory, and desperately in need of a way to make sense of it.
Conjoint does that.
In this newsletter we review the basics of what conjoint analysis is (sometimes called “trade-off” analysis) and how it can help you.
Other items of interest include:
- When to Use Paper Surveys
- How to Measure the Un-Measurable
- Can a Focus Group Save Spider-Man?
- The Magic Numbers… Reappear!
- Allstate’s PR Misstep with a Silly Study
- Tips for Surveys on Smartphones
- How to Sell Your Boss on Research
- Make it Real with Adaptive Conjoint
- When “No Difference” Makes a Difference
- Survey Says: Call Me on My Cell Phone
- Getting Respondents to Love Your Survey
Not sure of the trade-offs involved in choosing one conjoint method over another, or of choosing an entirely alternative method? We would be happy to help sort through your options and decide upon the best approach.
In the meantime, understanding what conjoint is and how it works will put you on the right path. So please read on.
The Versta Team
ABCs of CBC: Understanding Conjoint for Market Research
Versta Research has been developing, fielding, and analyzing a number of conjoint studies over the last several months, including the full alphabet soup of ACA, CVA, CBC and other varieties of conjoint. In our experience, conjoint studies are among the most powerful techniques of survey research. In this newsletter we focus on some of the basic ideas, advantages, and uses of conjoint research. What is conjoint? How and why is it used? What insights can it give you? Furthermore, what are some of the pros and cons of fielding research using a conjoint method vs. other methods you might use?
What Conjoint Is and Why It Is Used
Conjoint is a method that considers the relative importance of multiple decision factors, and quantifies how buyers make trade-offs among those factors. Instead of just knowing that price is important, for example, a conjoint study determines the value of price relative to other important factors. It then allows you to model how preferences change as you modify price within the context of those other factors. What if you raised your price by 5% and your competitors held steady—what would happen to your market share and revenue? What if in addition to raising the price 5% you offered a satisfaction guarantee—would that protect your market share?
Instead of merely asking people what is important to them, a conjoint study has them make decisions and trade-offs where multiple important factors are presented together (conjointly—hence the name.) The value and importance of each factor is derived implicitly from the decisions made. Conjoint studies are like decision-making experiments, and with data about how each respondent makes trade-offs, we disassemble that decision-making into its component parts. We then re-assemble those components into a general model, from which we can predict thousands more decisions without additional survey questions.
The approach is exceptionally powerful because:
1. Conjoint overcomes the tendency of respondents to say everything is important. Buyers often say they want everything, so asking them directly about the importance of factors such as price, quality, speed, service, etc. may fail to differentiate among the factors. But buyers do make trade-offs and conjoint lets us observe those trade-offs to determine which factors drive the decisions and by how much.
2. Conjoint reveals the underlying structure of decision making, which allows us to build mathematical models of what buyers are likely to do and why. With conjoint, we know the outcome and the implicit calculus (the decision-making) that brought it about.
3. Conjoint is more efficient than concept testing. Concept tests are static and require large sub-samples to evaluate ideas one at a time, independently of all others. In contrast, a conjoint study can simulate thousands of concept evaluations, each within the context of any others.
4. Conjoint can be used to predict choices and marketplace activity with regard to products, ideas, and concepts, even those that were not explicitly addressed in the survey.
The “outputs” from conjoint studies are not the typical frequencies and banner tabs from survey research. Instead, we build a marketplace simulator (in Excel or with other tools) that allows us to test real or hypothetical scenarios of product versions and competitive sets. We can evaluate the relative appeal of product configurations, and predict the decisions that buyers will make. The tool also tells us the relative importance of each factor in driving decisions, and we can document the sensitivity of each factor within the context of others, including price elasticity.
Another important advantage of conjoint techniques is that the questions (and hence the data they elicit) are more realistic. In a recent survey we fielded among B2B decision-makers, respondents told us how much they liked participating in the study compared to other research studies they have done. They said it was “real” and interesting because it was confronting them with questions that reflect the kinds of decisions and trade-offs they make every day in their work. And even if they don’t necessarily enjoy the exercise, they tell us how effectively it can zero in on the decision drivers that matter to them:
“Interesting to take this survey it helped me realize what my priorities are as a shopper.”
“Whew — that was a complicated survey. Many of my choices were based upon budget/price. Features are very important, but ultimately I’m driven (today) by budget.”
“My head was hurting near the end, but I expect this approach resulted in more accurate info from me (checks and balances).”
On the down side, conjoint studies can be a time-intensive process for both researcher and respondent. There is rarely room in the survey to collect data beyond what is specifically required for the modeling. Forget about including a half dozen “nice to know” questions that others on the management team would love to have answered while you’re out there talking to their customers.
Another potential downside to conjoint is that if not done correctly, there is no way to “rescue” useful information from the data collected. The questions are built in tight, interlocking dependencies, each providing scant information on its own, and each dependent on all the others to be useful. As such, conjoint requires exceptionally careful thought and survey design; mistakes require scrapping the entire data collection effort and starting over (happy to say, we at Versta Research have never had to do this).
Questions Conjoint Can Answer
Conjoint can be used for a wide variety of studies, including product development and line extension research, concept testing, research to refine messaging and value propositions, competitive positioning, segmentation, pricing research, and brand equity research. Here are three quite different examples of studies we recently completed:
- A pharmaceutical client needed to know where it could reduce patient assistance without hurting its market share. Physicians, RNs, and PAs were presented with scenarios linked to competing products with various combinations of patient assistance, and asked to think about what proportion of their next ten prescriptions would go to each product.
- A financial services firm was hearing loud complaints from their network of advisors about low interest rates, but uncertain about whether and how to respond. Should they pay higher rates? And if so, where should they reduce benefits elsewhere to compensate?
- An IT software company wanted to identify areas for new development that would enhance the appeal of its SaaS product. Their goal was to find the two or three enhancements that would help them migrate customers from an entry-level subscription to a more profitable tier of functionality and service.
The financial services conjoint study provides a particularly striking example of the insights that this method can provide. Paying low interest rates helped profitability, but management was concerned it might drive customers away. To find out, we conducted online surveys (but recruited by phone in order to keep the survey sponsor unknown to the respondents) using a simple full profile conjoint method. We fielded parallel surveys to financial advisors and individual investors. This allowed us to build a conjoint model that quantified the importance of interest rates vis-à-vis other costs and services, and to compare the two groups.
The conjoint model showed that although interest rates were extremely important to both advisors and their clients (indeed, important enough to generate volumes of complaints), in trade-off scenarios both groups consistently chose low interest rates as a means to get other services and valued benefits. This was contrary to what the advisors predicted of their clients, but consistent with what our client believed was operating in the marketplace.
Should the client raise interest rates? The answer was no, not if it meant reducing other key benefits customers were coming to them for. But the research also affirmed the need for stronger messaging to help keep those other benefits top of mind.
Getting Started with Conjoint
So how do you get started with conjoint? As always, the critical first step is to clarify the questions you need answered rather than focusing on methods or the type of study you want to conduct. (See our whitepaper, The Art of Asking Questions.) Then, the nature of your questions will largely dictate the type of conjoint to use. There are several types, including traditional full profile, partial profile, adaptive, and choice-based methods. There are also choice-based techniques similar to conjoint worth considering, such as MaxDiff analysis.
Other factors that will affect your choice of method include sample size, the need for individual-level vs. aggregate analysis, how many factors you need to include in the model, and whether pricing is central to the research. For a thorough discussion of which types of conjoint to use, what analytical techniques to employ (regression vs. Hierarchical Bayes), and to begin deciding whether to buy or build your own conjoint tools, we recommend much of the information available on the Sawtooth Software website. The folks at Sawtooth have spent decades developing and refining both the theory and practice of conjoint, and they provide excellent primers and software for market researchers.
Or, better yet, we at Versta Research would be pleased to help you consider your options, including the option of working with us. We will help you understand, design, and implement conjoint studies in all of their technical complexities. Plus we will help you turn data and information into stories that provide incisive answers to your questions. Our approach may not be right for everybody, but it might be right for you. Which would you choose?
Stories from the Versta Blog
Here are several recent posts from the Versta Research Blog. Click on any headline to read more.
Despite new survey technologies, paper surveys can be far more effective in some circumstances. This articles describes when and how to choose this method.
This article summarizes new technologies and methods of research that can measure important and hitherto unmeasurable aspects of consumer and buyer behavior.
Just days after critics panned the musical, market researchers were brought in to help rescue the show. Here’s our take on how and whether research can help.
Arbitrary numbers like p=.05 for statistical tests are useful. But they are not “magic” and not a substitute for expertise and judgment in market research.
Research for PR often slips into silly surveys that gratuitously grab attention. Here are tips for designing research that truly supports your PR message.
Smartphones are likely to become the device of choice for many online tasks. Here are tips for online survey design optimized for smartphones.
Move discussion away from the value of research, towards the importance of the questions and the business value of having answers. This article shows you how.
Adaptive conjoint is a method that presents people with decision-making scenarios, learning and adapting the survey based on their answers.
Sometimes measurement scales for survey questions should contain neutral midpoints and sometimes they should not, depending on how meaningful that midpoint will be.
In the US, one quarter of adults no longer have landline telephones at home, so including cell phone numbers in your surveying is critical.
Recent research reveals qualities of surveys and the survey-taking experience that respondents love, which offers tips for best practices.
MORE VERSTA NEWSLETTERS