Recently, the team at eCornell had the chance to sit down with Marco Vriens, Managing Director of Strategic Analytics and SVP Methodology at The Modellers and ask for his insights on marketing research using conjoint analysis. Marco has also appeared in eCornell’s Ask the Expert segments for our newest certificate Advanced Marketing Research.
How do big data relate to conjoint analysis?
So nowadays there’s a lot of talk about big data, and obviously big data has an incredible potential to inform marketing people. But for these types of problems big data is actually not very suited because big data only provides what’s already in the world, like how people respond to what’s already there. But conjoint actually allows you to get insight into things that don’t even exist yet. So I just want to point that out because that’s the discussion for today and I think that’s one of the reasons why I think everybody who pursues a degree in marketing should know about these conjoint methods.
Will big data analysis replace conjoint analysis?
Big data, I think, is an emerging field and there are a couple of marketing applications that I think for which big data is incredibly useful. Things like a marketing mix analysis and determining how efficient your marketing activities are. That is an area, I think, big data can help. Even in product development where you’re trying to understand what the trends are, what attributes and what features people like, and how that evolves over time, that scenario I think big data can help.
Keep in mind that conjoint is a snapshot. I’m interviewing people today and I’m getting insight into what people like today. But it doesn’t necessarily tell me how data will evolve over time and so that’s why big data would come into play. And so to some degree a conjoint approach and big data are supplementary if anything, but it’s not like big data is going to replace what you would do in conjoint, it’s a good supplement..
What critical information can a conjoint analysis yield?
Well once you gather the data, you estimate your model, keeping in mind that you have hundreds of respondents. And as many people know in marketing, not everybody has the same preference. Some people like this and some people like something else. And so it becomes a little bit of a computational problem, but conjoint analysis allows you to do that. Once you have your model estimated, you can create a simulation model. And that simulation model allows me to say, okay pick across this group of consumers, and pick the four or five profiles and [products] that would be maximally, optimally appealing to your target audience so that is a huge benefit.
A second benefit is also that you can get some insight into if there are maybe some potential discreet market segments that really have very different tastes from each other, so maybe you need to take that into account and it could even be that if those segments have different demographics that could actually mean that you may need a different marketing campaign for that.
Can you describe some common pitfalls to avoid in conjoint analysis?
There are a couple of common pitfalls and a couple of mistakes that people want to avoid. Make sure that your experimental design can actually estimate the model that you need. Nowadays there is a lot of standard software available from SPSS to sorted software. That software is in general very good, but you still have to be careful because there are situations where the software cannot handle your particular problem if you don’t still go ahead and come up with a standard design, that design might not be the correct design for your problem.
If you do start modeling, think about substitution and cannibalization and if you really think that those effects are going take place in your market, you need to make sure that you analyze the data that you use the right models, so that you can actually get insight into those kinds of dynamics.
And I guess follow the best practices as much as possible. I think this is really an area where you want to hire people who do this as a day job and who do this a lot so that you know that the people who are going to do this project for you will do it well.