Creating a control group to match a pilot group

[Adkook]

Hey all,

I need help with an analysis I am attempting to conduct. I need to identify a control group from within a large population to loosely match the average characteristics of a smaller pilot group.

i.e., my pilot group has:

• 709 individuals
• Average orders = ~43.8
• Average Total Revenue = ~4,200 
• Average Online Revenue = ~500.

From the broader list of 25k individuals, I need to identify 709 that will closely mirror those three dimensions on aggregate.

From my research it seems like the AB Control tool would be helpful in this, but the ways I have attempted to configure it I get an error message (in screenshot below).

Any suggestions on how to either 1) correctly use this tool, or 2) another method to complete my analysis?

I have consulted other colleagues who have suggested doing propensity score matching in Python; however, if possible, I would prefer to keep this analysis in Alteryx. 

Thanks in advance!

[oly]

Hi @Adkook ,

Both control and treatment(new records pool) group should have the variables "#orders", Total Revenue and Online Revenue. I slightly modified your workflow to add average on the # orders and sum on the both revenues.

Also, since you have several lines per UniqueID, I added another Summarize Tool.

Additionally I added summarize across all lines (no grouping) to compare the results of your selections as average of orders and revenue totals. You can append those individual records if necessary. You can see the result workflow.

AB testing will find closest matching elements to the individual selections in your control group. Hope that's exactly what you need here.

Let me know if you need further help,

Best,

Oly