Segmenting and Analyzing Local Mosaic Clusters

[daniel_mmi]

I'm working with a list of names and addresses I've run through the ConsumerView Matching, and thanks to this post, I have a good start on figuring out my demographic distribution:

http://community.alteryx.com/t5/Alteryx-Knowledge-Base/Identify-Top-Clusters-in-Your-Customer-File-M...

As I understand it, that method will give me the highest density clusters, relative to the general population within the area in question. For instance, if the model predicts that 10% of the records should be a part of mosaic group A01, and I have 12%, it reports an 'index' of 120. Using that, I can figure out which groups in my list are more strongly represented, not just more numerous.

Here's the question: since I have a nationwide list, is there a way to calculate those index values based on smaller geographic subsets, perhaps broken up by CBSA or within the boundaries of a state or county?

I already have the household mosaic codes assigned, and am able to get the results I want across the whole dataset, but as soon as I try to group or segment it within the Behavioral Analytics tools, it falls apart. I did figure out a way to do it running the analysis as a macro and feeding it the data already segmented, but it's not very pretty, and it makes me feel like I'm overlooking (or not understanding) something much simpler.

[dataMack]

Yes- you could absolutely do the indexing at smaller levels of geography, like a CBSA or even down to census block group.  The index is simply the portion of your customers that are in a given segment, for a given gepgraphy, compared to the portion in that segment entire target population in that same geography.  

[AmyH · Accepted solution]

You're correct, you can do this with the Behavioral Tools.  Take a look at the "Create Customer Profile with Penetration" Sample within the Behavior Analysis category.  It's very similar to the knowledge base article that you referenced, but maybe a little simpler.  This example allows you to profile your customer file and compare it to any other base.  In this sample the customer file is compared to the BGs that the customers fell within.  The clustering is based on dominant BG code, not using the Household cluster assignment from the ConsumerView file.  This would save you the step of having to summarize it yourself, and if you are already grouping it by BG, it should give you similar results.

[daniel_mmi]

Okay, took a little figuring, but I got there. I was looking to get the index and penetration based on block group, but also wanted a separate profile for each city. Making the following changes to that sample workflow #3 gave me the results I wanted:

1. Adjusted the Spatial Match tool to include the 'City' field in the results.
2. Adjusted the Summarize tool to group on 'City' as well as 'Key'.
3. Adjusted both Create Profile tools to group on 'City'.
4. Added a Detail Fields tool after the to profiles were joined, but before a new profile was created. (This was the spot where I'd gone wrong, previously.)
5. Filtered out all the unmatched clusters.

Looking at the resulting table, I can see the index for each cluster/city pair:

Thanks for pointing me in the right direction!