Forum Moderators: buckworks
[webmasterworld.com...]
Curious if anyone has actually tried it and what your results have been. I am working on it now with some tweaks based on what you last looked at and what items are in your basket
I have actually been suggested a product that I hadn't looked at on amazon before and bought it, so I guess it works.
In fact, I find this kind of personalization to be one of the major factors that makes amazon.com the best place to buy stuff on the internet today.
Behind the scenes, it used to be:
"People who bought this have also bought this"
But now it is:
"People who have **recently** bought this have also bought this"
It weights the output more heavily towards recent behavior in the customer record, since recent behavior is more predictive overall. This slight adjustment is quite powerful, because it more closely follows current trends, rather than working with potentially "stale" data.
This idea of recent activity being a more powerful predictor than overall history is one that has been used offline for many years in the catalog and TV Shopping businesses. The metric is referred to, appropriately, as "Recency" - Purchase Recency, Category Recency, Complaint Recency, and so forth. Online you can add Visit Recency, Download Recency, Game Play Recency, etc.