Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.tools -> Re: how does amazon.com implement its smart book suggestion depending on previous purchases by customers ?
We do this all the time with PC games.
We compute the cross-sell for each game purchased, that is, what is the most frequently purchased game for households who've purchased any given game. [For games with no purchase history, e.g., new games, we match on cross-genre]
For us, this job runs once a day.
When the consumer visits (round 2?), we know what games(s) they've already
purchased
and can "offer" them the top 3 or 4 most frequently purchased "companion"
games.
[BTW: we also know the average length of time (interval) between purchases.
so
if the consumer visits for round two within interval 1, then we offer the
cross-sell
from the 1-list, and so on for up to N intervals.]
[BTW(2): we also know life-to-date (LTD) purchase patterns; if this
consumer visits
for round 2 and should order from a specific cross-sell interval list, and
has LTD=3, then we
could determine if a special discount can be applied (or free shipping,
etc).]
This sort of thing has been going on in direct-marketing (aka database
marketing)
circles for quite some time. Amazon just does a good job of presenting
parts of
it in "real" time.
"kilidire" <yc_at_hwcn.org> wrote in message news:3b563067.486005158_at_news...
> Hi,
> I am sure you've all been to amazon and done some purchases there. I
> would like to know if anyone knows how they do this. Once you have
> purchased a book, they have a smart feature which tells about other
> books which might interest you based on purchasing patterns of other
> customers. Amazon.com is not the only site doing that, there are more
> and more sites implementing it. Does anyone know how they implemet
> that kind of feature ? Can it only be applied to books or to any other
> product ? Thanks all.
>
>
> I'm thinking that they must query the purchases table to retrieve
> this info (which is to sugesst 3-4 similar books based on other
> customers purchase patterns) , but it will be a complex and intensive
> query and I don't think this is done in real time. Maybe they use some
> kind of procedure and run in batch mode and save the results ( 3-to 4
> books) in a different table that is linked to the product table. Any
> ideas ? Thanks.
>
Received on Mon Jul 16 2001 - 19:00:03 CDT
![]() |
![]() |