Date: Fri, 26 May 2006 10:36:33 -0400
Reply-To: Amw 5G <anwright@UNCC.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Amw 5G <anwright@UNCC.EDU>
Subject: Re: Students like treats
Fair 'nuff. It's a marketing optimization problem, whereby every customer
is scored for a different product, and we want to drop marketing campaigns
on certain days of the month. The score is the expected NPV for the
product, adjusted by propensity to respond (typically in the range of a few
cents to $50+). Anyone who already owns the product is given a null score.
The constraints are to make sure we don't overburden customers with
marketing messages (the 21-day window), and that we give customers an
adequate time to respond to a product-sepcific message (the 2-month window).
The goal is to find the customer-specific marketing plan so that the NPV is
maximized over the 6-month calendar, subject to these temporal & frequency
constraints. It's complex because those are the rules in which our
marketing process operates today. I'd like to see if there are any trends
that come out of the calendaring process. For example, if I see a trend
towards offering more of product A in months 1, 3, 5, then I know that that
campaign for A should really be effected every other month, instead of
wasting resources doing it every month.
I also intend to find out the overall expected marketing value for the
entire customer base, if we were to go with the "optimal" marketing
calendar versus what the individual product lines want. They are naturally
product-centric, whereas my approach would be customer-centric. M. D.
Cohen of SAS has a really good write-up in Information Systems on customer-
centric marketing optimization via SAS, but his constraints are more around
capacity, whereas mine are time-based.
I figured that if I phrased it in the context of something more "fun" than
boring ole marketing, it would be a more interesting read and more likely
to get suggestions. My apologies if that made the project description
needlessly frivolous or if it injected needles complexity. I thought it
still described the problem adequately and accurately, but I guess not...
Thanks for your consideration.
On Thu, 25 May 2006 22:04:07 -0700, David L Cassell <davidlcassell@MSN.COM>
wrote:
>This seems really bizarre. What is your *real* problem? Or is this a big
>homework problem? Why would you be trying to maximize a problem
>with constraints like these?
>
>You say "My objective is to maximize the value of each Student through
>optimal Treat allocation." That doesn't make sense. You need to specify
>your exact optimization criterion. Mathematically.
>
>You have a value scale of 1-100 *plus* a null that represents "doesn't
>like".
>Which is silly. With a 1-100 scale, anything below perhaps 20 is already
in
>the "doesn't like" area. Why can't 1 or 0 be "doesn't like at all" instead
>of a null?
>
>Why are you forcing your data into 181 days? You only have 5 available
>days in each of 6 months. That's only 30 usable days. Tops. So of course
>your Cartesian joins are a nightmare.
>
>So perhaps you need to write back to SAS-L and explain what you are
>*really* trying to do, and why, and why you have made the problem so
>extremely complex.
>
>HTH,
>David
>--
>David L. Cassell
>mathematical statistician
>Design Pathways
>3115 NW Norwood Pl.
>Corvallis OR 97330
>
>_________________________________________________________________
>Express yourself instantly with MSN Messenger! Download today - it's FREE!
>http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
| 
