In Diagnostics projects (click here), I occasionally run into clients who thoroughly understand the dynamics of their customer file.
This happens infrequently.
But when it happens, I get to have a ton of fun!
Look at this image. What the heck are we looking at?
Here's what I do. You can do the same thing - go talk to your analytics expert - and when the analytics expert balks at your request, come talk to me (email@example.com).
- I take "x" years of data ... typically 3-5 years.
- I create three slices of the file ... 2014 - 2016 ... 2013 - 2015 ... 2012 - 2014.
- I create a ton of variables at a customer level ... three year purchase frequency, average items per order ... average price per item purchased ... % of demand by merchandise category ... % of demand by marketing channel ... % of demand by full price vs. discounts/promos ... % of demand by mobile vs. desktop/laptop vs. call center ... and other key variables.
- I select only customers with 1+ purchase in the past year. No lapsed buyers in this analysis.
- I run a Principal Components Analysis ... where I reduce the 20-30 variables down to just 2 variables.
- I create 4%-tiles of each of the 2 variables created by the Principal Components Analysis.
- I create a new dataset - which generates a 25x25 grid of 625 customer segments.
- Within each of the 625 segments, I calculate the average of every attribute that went into the Principal Components Analysis.
- I use Graph Expert Professional (click here) to generate the image above. I plot the X/Y average of the 625 segments against the average (in color) of any/all variables used to create the analysis.
I know, I know, that's too much math. But that's what I am doing. If you want the SPSS code that I use to produce the analysis, send me an email message (firstname.lastname@example.org) and I will send the code to you.
Tomorrow, I'll explain to you what that "yellow sun" in the western portion of the image means.