JMP for marketing and consumer research — an update

[#1]

In a previous post, I compared what’s available for marketing and consumer research in JMP to what’s taught in an MBA classroom as represented by this textbook. Now that JMP 13 and JMP Pro 13 have been released, I’d like to provide an update on the newly added platforms: 

MDS (Multidimentional Scaling), previously available only as a JMP add-in, is now fully developed into a new platform in JMP Consumer Research. It is a very useful analytical tool for visualizing similarity or dissimilarity of objects such as product or brand similarity in a 2D or 3D plot. 

New MaxDiff (Maximum Difference Scaling) Design and Analysis use a single-factor design to get consumers' picks of "best-worst" pairs from multiple choice sets and then fits a random utility model to estimate the probability of consumer’s preferring one option to another. The MaxDiff method complements the existing Choice Design and Choice Models so researchers can choose an appropriate approach to investigate consumer choices in different situations.

Association Analysis, also known as Market Basket Analysis or Affinity Analysis, can sort through a large transnational data set to discover interesting patterns like "80% of shoppers who buy a bag of tortilla chips also buy salsa. This association occurs more often than chance alone by a factor of two."
The Association Analysis platform is available in JMP Pro since it also uses an advanced statistical technique called singular value decomposition (SVD) to group items into topics. 

JMP 13 brings together different clustering methods such as K Means Clustering and Variable Clustering into a new menu called Clustering. A new technique called Latent Class Analysis (LCA) is added. Compared with K Means or Hierarchical Clustering. LCA is probability model based and works with categorical data. Just like other clustering methods, it can be applied for customer segmentation by estimating two sets of probabilities: the probably of belonging to a cluster and the conditional probabilities for each response level based on the assigned cluster.

 (Clustering Probabilities from LCA)

Last, but not least, JMP 13 adds Text Explorer that opens the door for you to understand your customers’ opinions and experiences through unstructured texts as never before. (Note: Some of Text Explorer’s advanced capabilities such as Latent Semantic Analysis and Topical Analysis are only in JMP Pro). 

To learn more about these new platforms, stay tuned for my 2017 Advanced Mastering JMP live webinars.

[stanmaklan]

Looking for a marketing data set for MDS... have you got something?

[jiancao]

Two data sets from the book, MBA Programs and Beer Brands, are good for MDS. 

https://community.jmp.com/t5/JMP-Add-Ins/Addin-for-accessing-data-sets-used-in-Modern-Marketing-Rese...

[stanmaklan]

Thanks got them up. At the risk of looking silly... 1. Not sure how to get names against the dots on the matrix (e.g. MBA works but cant get the tags Harvard et al.. to appear) and 2. How to transform the Beer data so that I create that symetrical matrix that MDS requires

[jiancao]

1. To enable the label to appear on MDS plots, highlight rows on the JMP table, and then select Rows > Label/Unlabel from the menu.  

2. JMP supports data format being attributes, so you donot have to use distance matrix.

[stanmaklan]

OK hopefully one final quesiton. When I highlight the rows and label what I get instead of the name of the school are numbers one to nine... How do I substitute the name of the columns for numbers of the rows... Same issue on the other data set. Thanks

[jiancao]

Please also select School on the JMP table and then right click and choose Label/Unlabel as shown in the screenshot.

[stanmaklan]

Sorry this is just not working. Nothing will generate labels for me. Using JMP for MAC but should not make a difference

[stanmaklan]

Further thought... using JMP 14 does that explain the problem??

[jiancao]

Stan,

In order to display column values (i.e., school names in this example) and enable the label to always appear, you need to label both rows and column.   Check out the documentation for details:

http://www.jmp.com/support/help/13/Label_Rows_and_Columns.shtml#481050.

[TCM]

Jian,

In your JMP 14 webcast, you talk about Multifactor Analysis using a wine dataset. My question is this:

I have structured my data similarly such that each row in the table is a product and the columns are the n product features by panelists (i.e., the columns are grouped by panelist), and I have performed PCA on all the columns.  What is most important to product developers is the score plot- a 2-D plot where products fall on the Dim 1 vs Dim 2 map.

Can new products be introduced into the table to find out where they fall relative to the original products that were used to create the score plot?  I know this can be done in a "regular" PCA map obtained from data where each row is a product and each column is the mean rating/value for each feature.  One simply enters a new row for the new product and inputs its features in the columns.  The PCA map reveals the placement of the new product.  How may this be done in MFA mode?