The best analytical approach to an analysis must take into account the availability of and quality of the data as well as the business objective. It doesn't make sense to analyze data from the 'distant past' unless you believe the relationships will still hold for current data. In general, you would likely get a lot of different answers even with that detail available. In many cases, the business needs will help to identify how large each window is.
My assumption is that your recovery model would be tasked with identifying how to obtain the most revenue from people who still owed money after disonnecting their service. These problems might be well-suited to a two-stage model -- one model which identifies whether you receive any money from someone after they disconnect, and the second model which predicts how much money you will be able to recover after they disconnect. The product of these two models predictions generates an 'expected revenue per disconnect' which you could use to prioritize those accounts that will provide you the greatest revenue. There is a TwoStage node that accepts both a binary (Yes/No) target and a value(e.g. revenue) target and allows you to build this model automatically. You can also model the two targets separately.
I hope this helps!
Doug
Don't miss out on SAS Innovate - Register now for the FREE Livestream!
Can't make it to Vegas? No problem! Watch our general sessions LIVE or on-demand starting April 17th. Hear from SAS execs, best-selling author Adam Grant, Hot Ones host Sean Evans, top tech journalist Kara Swisher, AI expert Cassie Kozyrkov, and the mind-blowing dance crew iLuminate! Plus, get access to over 20 breakout sessions.
Use this tutorial as a handy guide to weigh the pros and cons of these commonly used machine learning algorithms.
Find more tutorials on the SAS Users YouTube channel.