A nice article over at Read/Write Web about recommendation engines. They break down current systems into 4 categories:
- Personalized recommendation – recommend things based on the individual’s past behavior
- Social recommendation – recommend things based on the past behavior of similar users
- Item recommendation – recommend things based on the thing itself
- A combination of the three approaches above
I take issue with the label “Social recommendation” because that’s not what social is. Social means being part of a society, being connected.
When Amazon tells you that 63% of people who bought product X also bought product Y, there is nothing social involved. This is math.
What I want to see is a recommendation system that takes advantage of the fact that true social information is now available in the form of social networks. Don’t tell me what 63% of other shoppers bought, tell me what my friends and experts bought.
I was thinking about this kind of thing earlier today after reading a blog post about microformats. Amazon would do better to use XFN to improve their recommendations. I wish I’d thought about this before I wrote my latest blog entry.
It’s much worse than Amazon not providing ‘true’ social recommendation. It’s this sentence:
“The Amazon system is phenomenal. It is a genius of collaborative shopping and automation that might not be possible to replicate.”
What’s genius about browsing for kids toys for a friend, and then having the damn bot keep recommending Wiggles DVDs to me? It’s cool to see what others bought (stats) but when they try to guess what I want, they always do a terrible job.
I hope the Amazon system will never be replicated. A far better solution is just off the horizon.