80 million dollars is a lot to pay for a recommendation engine.
Netflix had their own and paid a million dollars as prize payout for an absolutely state of the art system.
If it really was just a tech acquisition they massively overpaid.
As someone who builds recommender systems for a living, let me assure you that the output of the netflix prize was perhaps 10% of what you need to actually build a good recommendations product.
Sure, I wasn't implying eBay overpaid by 80x but 79 million dollars is a lot of product development once you've paid 1 million for a proven state of the art proof of concept.
Obviously eBay must see other reasons for the acquisition than just the tech itself.
Also, if it's also in large part a talent acquisition then they're going to paying more than 80 million since they're going to need to pay for some new pairs of golden handcuffs to keep the talent around.
Don't forget that Hunch's system is real-time (this is very hard), and is more than just scalable collaborative filtering for items of the same kind. They can ingest a Twitter or Facebook profile, infer your "taste profile", and start recommending things instantly; "things" meaning books, films, music, events, places etc etc.
How does that vary much from what Etsy did with their gift recommendations application based on your/friend's facebook profile
I don't believe real time is very hard. Something that ebay can' already do. And if they can't already do it, how will hunch come in and solve it. It would probably require major changes to db's etc if that is why ebay can't already do it.
> I don't believe real time is very hard.
At scale it is. More here:
http://www.quora.com/Hunch/Is-Hunch-actually-doing-anything-... (this one got a vote from one of the founders)
See my above reply to suking--$80m is kind of a bargain.
Impressive back-end tech (the recommendation engine) that can now process millions and millions of items on eBay to help increase sales (something like a quarter of Amazon's sales come from recommendations).