Testing the Trade-offs

Chrstine Birkner
Marketing News
Current average rating    
Key Takeaways
  • Choice modeling can help you explain the choices that people make with models that tell you the probability that a person would choose one good versus another.

  • Choice modeling should be coupled with competitive research since your competitors’ options will impact your target customers’ purchase decision-making.

  • Choice modeling can help marketers narrow down their options to create the best possible product.​

​​​​Choice modeling mimics real-life consumer decision-making to help marketers position their product offerings for success. Experts explain the concept and how to put it into practice.​​

When marketing researchers want to survey potential customers about potential new products, features or package designs, it’s helpful to simulate real-world purchasing behavior, wherein consumers must make trade-offs on specific product attributes when deciding whether to make a purchase. One way to do that is through choice modeling, which measures consumers’ preferences for product characteristics by showing them several product profiles, often side by side, and asking them either to give a rating to each one, or to choose the product that they most likely would purchase.

Choice modeling was invented in the 1920s, but it started being used in marketing research in the 1970s. Today, most marketing practitioners rely on automated software to set up the research. First, you must decide which product features to test, and software will help you rank the importance of each feature. “On the back end, we’re able to put the pieces together to show a preference for any possible combinations. That might be thousands or millions. The more attributes we’re considering in the model, the more we include, the more complex the design becomes, which means we’ll show more choice scenarios,” says Jay Heyboer, vice president of Morpace Inc., a Farmington Hills, Mich.-based market research and consulting firm.

Choice modeling can help you explain the choices that people make with models that tell you the probability that a person would choose one good versus another, says Greg Allenby, professor of marketing and statistics at The Ohio State University in Columbus, Ohio. “What come out of these models are predictions expressed as probabilities. Your goal is to find the coefficients in your model that maximize the probability of the goods that are chosen. … Software packages can help you design a study and will estimate coefficients for you. Then, you’re ready to make predictions.”

The research method is particularly effective if you’d like to change multiple items in your marketing mix, says Caroline Roe, director of quantitative research at Insights in Marketing, a Wilmette, Ill.-based market research company. “Maybe you’re looking to shift your price up, but you’re not sure how high up, and you have three different prices you want to test and four different package designs you want to test, and you have a matrix of all of these combinations. Choice modeling can be a great way to see which combination is going to be the best.”

For example, if you’re developing a new frozen pizza product and you want to figure out the attributes that will appeal to consumers in terms of flavor, ingredients, size and nutritional content, you should use choice modeling, says Kathryn Korostoff, president and founder of Southborough, Mass.-based market research consultancy Research Rockstar. Rather than designing a simple questionnaire and asking people to rank each attribute on a scale of one to five, you could present them with mock-ups of frozen pizza boxes that mix and match the product’s features. “If you’re asking them to rate each feature one by one, you run the risk that they could check everything off as ‘very important,’ and then you don’t have any priorities. … [Choice modeling] lets you simulate, as closely as possible, the more complex decisions they would have to make in the form of trade-offs.”

Here, experts weigh in on best practices for leveraging choice modeling.

1. Only test what’s necessary. “Sometimes folks will say, ‘As long as we’re at it, let’s add this, this and this to it.’ That’s really shooting yourself in the foot,” Heyboer says. “You’ll have lower precision in the stuff that really mattered to begin with.”

Jordan Louviere, research professor at the Adelaide, Australia-based University of South Australia Business School’s Institute for Choice, suggests conducting a stage-one prioritization study to get the number of attributes or features down to a usable level. Then you can quantify how much each feature matters in consumers’ decision-making processes. “Once you’ve done that, you need to go through the process of deciding which values to assign to each of the features,” he says. “That will vary. Some features are either there, or not. Other features like price are much more complicated: You have to decide the range of price and which values of price you’re going to vary.”

2. Keep surveys brief. Most marketing researchers use software to generate the appropriate number of choice combinations to use in a session, Korostoff says. “The software figures out the minimum set of tasks to present to people so it can determine the relative importance of those items, so that if you were taking [an online] survey, you wouldn’t have to work through 1,000 screens, but rather 20 different screens.”

Adds Roe: “You want to have the right number of observations but still make it so that individual respondents aren’t overly burdened. The last thing you want is for it to be tedious and have the respondent just start to click at anything, which will ruin your data.”

Choose participants who have a more vested interest in your product, and then put yourself in their shoes, Korostoff says. “Ask yourself: ‘Would you complete this task? Would you go through 15 screens that look like that?’ If you won’t, your customer probably isn’t going to.”      

3. Be flexible. Your results could shift based on changes in the marketplace, so you need to be able to land on your feet, Korostoff says. “If you’re surveying one set of features, your competitor could unveil a new feature tomorrow that changes the way people make their trade-offs. Just because you’ve just spent $50,000 on a research project doesn’t mean it’s going to have a long shelf life. Some markets change rapidly and some do not. Be honest with yourself about the shelf life of your data.”

4. Incorporate choice modeling with other research tactics. “Take your data with a grain of salt,” Roe says. “Choice modeling is very powerful, but it isn’t able to account for every single factor going on in the marketplace.”

Choice modeling should be coupled with competitive research since your competitors’ options will impact your target customers’ purchase decision-making, as well, Korostoff says.

Overall, choice modeling can help marketers narrow down their options to create the best possible product, Roe says. “When you have 42 concepts that are variations on a theme or when you have multiple marketing components moving around and you’re trying to figure out what’s going to work best, choice modeling can be very useful.”


This was originally published in the March 2015 issue of Marketing News​.


Author Bio:

Chrstine Birkner
Christine Birkner is the senior staff writer for Marketing News and Marketing News Weekly. E-mail her at cbirkner@ama.org.
Add A Comment :

Become a Member
Access our innovative members-only resources and tools to further your marketing practice.