DM: RE: Telecom Credit Score

[Tcasey]

Jerry,

It may be difficult to "intuitively" pick a single variable that 
determines
the customer's credit risk.  Predictive modeling techniques (Neural
Networks, CHAID, etc.) can help, however, the difficult issue with 
using
predictive models is that you can't put a "face" on the customer.  It 
is
therefore difficult to develop a solution specific to that target 
group.  

One variable that I find to be a good start is geography.  View the 
credit
history of groups of customers by the various geographic designations 
you
have at your disposal (urban vs. rural, metro vs. suburb, by state, by
wirecenter, by zip code, etc.).  It helps if you already have a 
geographic
classification in place.  Regardless, if you can identify a trend (GIS
systems help with visualization) then you can start applying 
variables to
the geographic segments.  Through simple analysis you should find at 
least
some other variables that are worth investigating in greater detail.  
I
typically find this approach to work well when trying to make
generalizations such as the one you describe.

Let me know if that helps you at all.

Tom Casey

> -----Original Message-----
> From: DLaney 
> Sent: Sunday, 6 December 1998 03:01
> To:   Musial Jerry; datamine-l@nautilus-sys.com
> Cc:   SBarnes; Tcasey
> Subject:      RE: Telecom Credit Score
> 
> Jerry,
> 
> From my limited experience with such data, your initial results 
>might be
> skewed by customers who have churned and have left outstanding 
>balances. I
> have copied a couple of our telco DW experts to see what their 
>experiences
> in this area might be. If you have difficulty identifying churned
> customers from your sample-set (we know how difficult this can be in
> telco), we can help you with this as well.
> 
> Cheers,
> Doug Laney
> Solutions Business Development Manager, Prism Solutions
> 8750 W. Bryn Mawr, Chicago, IL 60631
> office: (773) 399 9175 --- fax: (773) 399 9435
> dlaney@prismsolutions.com --- www.prismsolutions.com
> 
> 
> -----Original Message-----
> From: Musial Jerry [mailto:Jerry_Musial@bscc.bls.com]
> Sent: Thursday, November 05, 1998 10:51 am
> To: datamine-l@nautilus-sys.com
> Subject: DM: Credit Score
> 
> 
> Hi,
>       I am looking for help/resources related to credit scoring of
> 'existing' customers.  In particular, I am interested in how 
>researchers
> deal with tenure as it relates to whether or not one of our 
>customers will
> continue to pay their bills in a timely manner.  We intuitively 
>feel that
> the longer a customer has been with us (and paying us) the better 
>credit
> risk.  However, my initial results show that the longer a person 
>has been
> with us, the odds of his not paying increase.  Anyone have any 
>ideas?
> 
> Thanks
> Jerry Musial
> BellSouth Cellular