amazon design doubt

Question

I was looking at the amazon website and was wondering how one of the feature would have been implemented. The feature : what customers buy after viewing a particular item.

If i were to develop such a feature i would probably generate a session id for each user session and store the session id-page id combination in a log file. and if a book is bought set a separate flag for the session id-page id. A separate program can then be run on the log file periodically, to identify the groups that were bought together/viewed together and that information can be stored in a persistent file.

This is ofcourse a simple solution without taking into consideration the distributed nature of the servers - but would this suffice or can you help me identify a better design.

Answer 1

What you can do is track what each viewer has looked at, by having (user_id, product_id, date) in a database. You can then look at what people bought the day they looked at that item.

But, you will want to improve on this, as, if I looked at a cooking pan, and bought a book, you don't want to show the two, unless it was a cooking-related book, perhaps.

But, storing the session id is probably not the best choice, as, what happens if I had looked at something, then later, after talking with my SO I decided I could buy that iPad, your way it wouldn't show up as something I bought after looking at a notebook.

So, start with writing out some user stories, basically just some scenarios that show what you want to do, then the design will probably be easier.