3 ms·

Netflix can do collaborative filtering successfully because they have years of data. Amazon can also make recommendations because they know everyone's buying ha

Netflix can do collaborative filtering successfully because they have years of data. Amazon can also make recommendations because they know everyone's buying habits. Sounds like Redux is trying to abstract this to social media? It reeks of "technology in search of a problem".

aneesh 18y ago
Exactly. The collaborative filtering technology & algorithms is the easy part. The hard part is getting mountains of good data to apply it to. And by good, I mean monetizeable. That's what Amazon and Netflix have done. They should solve the data problem first, and being yet another "interactive social media mashup web 2.0 social networking thingy" may not be the best way to get lots of data. The companies that have done this best have monetizeable data - Amazon can actually SELL you the book it recommends. As for being a collaborative filtering platform: "1. If your idea starts with 'We’re building a platform to ...' and you don’t have a billion dollars in capital, find a new idea. Now." (http://diffle-history.blogspot.com/2008/06/postmortem.html http://diffle-history.blogspot.com/2008/06/postmortem.html)
- attack 18y ago
  No. Vastly more input data does generally allow for better output by these algorithms. BUT this in no way means that todays recommendation algorithms are good AT ALL. In fact, they completely suck. And in reply to GP, netflix's recommendations suck very much too, to be frank. Why don't companies care? Simply because it's hard to figure out how to make a lot of money even with very good recommendations.
  - wheels 18y ago
    For the record, more than 20% of Amazon's sales are a result of recommendations -- so just in the one case of Amazon, their recommendation system accounts for $3.4 billion in sales a year ... which incidentally is 22 times Facebook's revenue last year. Recommendations do suck, and VCs are throwing money at the problem because there's a market there that actually makes money.

aneesh 18y ago

Exactly. The collaborative filtering technology & algorithms is the easy part. The hard part is getting mountains of good data to apply it to. And by good, I mean monetizeable. That's what Amazon and Netflix have done. They should solve the data problem first, and being yet another "interactive social media mashup web 2.0 social networking thingy" may not be the best way to get lots of data. The companies that have done this best have monetizeable data - Amazon can actually SELL you the book it recommends. As for being a collaborative filtering platform: "1. If your idea starts with 'We’re building a platform to ...' and you don’t have a billion dollars in capital, find a new idea. Now." (http://diffle-history.blogspot.com/2008/06/postmortem.html http://diffle-history.blogspot.com/2008/06/postmortem.html)

attack 18y ago
No. Vastly more input data does generally allow for better output by these algorithms. BUT this in no way means that todays recommendation algorithms are good AT ALL. In fact, they completely suck. And in reply to GP, netflix's recommendations suck very much too, to be frank. Why don't companies care? Simply because it's hard to figure out how to make a lot of money even with very good recommendations.
- wheels 18y ago
  For the record, more than 20% of Amazon's sales are a result of recommendations -- so just in the one case of Amazon, their recommendation system accounts for $3.4 billion in sales a year ... which incidentally is 22 times Facebook's revenue last year. Recommendations do suck, and VCs are throwing money at the problem because there's a market there that actually makes money.