Author: Melinda Thielbar
Industry: Consumer Package Goods
Functional Area: Market Research, Survey Analysis,
Product Life Cycle Stage: Market Research
This is a made-up brand awareness study that I created for visiting some consumer research customers in the CPG space. It shows how to use Maximum Difference Scaling (MaxDiff) to get customers to rank a large set of brands of chocolate bars without the cognitive burden of ranking all of them at once.
This is a common problem in consumer studies. You want to know how average consumers (i.e. an untrained consumer panel) will react to your product. However, in any situation where you are asking an untrained group to identify what they like and why, you can get a noisy ranking that could change if you asked them on a different day. MaxDiff allows you to present them with small sents and then infer the overall ranking. Each set is a cleaner result, and so the overall result is better.
The Latent Class Analysis used for the Respondent Info allows the researcher to pare a large group of attitudenal columns into two latent class columns that can be used in further modeling.
The data tables are:
1) A list of the products to rank, which can be used to show the MaxDiff designer,
2) An Excel Spreadsheet with the survey results and a spreadsheet with the resondent info (This is a good chance to show how easy it is to bring Excel data into JMP. Also, the Excel data has a mistake where products are ranked -1,0,1 in one sheet and 1,2,3 in another sheet. This is a good opportunity to show recode.)
3) A JMP table with the survey results, including table scripts MaxDiff for Text Explorer (non-Pro). The text exporer script lets you explor what people say when they choose each kind of chocolate first.
2) Respondent table with table scripts for analyzing the respondents, including a table script for Latent Class Aanalysis. The LCA shows that many different demographic and attitude variables can be brought down to 2 sets of latent classes for this data (4 total) based on 1) How much chocolate the respondents eat and their attitudes about chocolate (LCA: Candy Tried and Attitude), and 2) Consumer attitude questions about when it is appropriate to eat chocolate. The LCA allows us, then, to create two respondent columns representing 4 out of the potentially hundreds of different categories we'd have to use without it.
showSlides.jsl needs to be edited so it points to the directory with the slide images. Alternatively, you can use the PowerPoint presentation, which has the same informaiton.
Scene Setting: You often have to see how your brand compares with other brands. A MaxDiff study like this one can be fielded quickly and easily in an survey tool. This study asked individuals to rank well-know brands of chocolate bars in sets of 3 and then comment on why they ranked each group the way they did. We'll show how to analyse the MaxDiff experiment and how to use the Data Filter in with text explorer to understand the consumer comments. We'll also see how Latent Class Analysis can let you quickly group your survey respondents based on their habits and attitudes.
