Food for thought 1: a cool 3.0 product idea

I was talking to a friend of mine the other day, and he came up with this idea:

"I live on a block at one end of which there is a Duane Reade, and at the other a CVS. There are some items that are on sale/cheaper at one, and others cheaper at the other. Why can't someone build a smart consumer-driven shopping agent that given a shopping list will tell me at any point, which items to buy from which store?"

Kind of like a smart consumer shopping optimization engine, right? I thought the idea was brilliant in that it exposes directly the fact that the success of this is NOT a technology issue, but a business and product marketing issue. Here's why: while I think this is very doable technologically (biz dev with Duane Reade and CVS to access their inventory and/or promotion offer API's - if they exist, if not, create it/scrape the site, etc., etc.), the business value prop for them is painful and hits right at the core of the substantive dichotomy between consumer needs and data owner needs that I mentioned in an earlier post; while I want to have this capability as a consumer, the incentive for the data owner (CVS, Duane Reade) is yet unclear for why they should open up their data. This has to do with the mentality of "opaque" branding/selling/marketing, as opposed to "transparent". There are two issues here, and I would like to hear your opinions/answers on these (please consider i am the Chief Data Officer at CVS):

1. Why would I want to provide this service to the consumer, and open up my offer data? Since I plan my promotional efforts in part based on what Duane Reade offers, to get more in store foot traffic, why would I risk them knowing my every move?

2. How do I make this process more efficient, given that my promotional offers change on a daily basis as function of manufacturer/CPG deals, market demand, competitor (Duane Reade) promotion counter-offers, holiday retail, location, etc., etc.? In other words, how do you see this service being intelligent but also dynamic enough to pull of these data points considering that this process is mostly manual today?

Any takers? By the way, this is the first in a series of "wild and crazy" posts that will actively seek reader comments in solving a concrete idea.