Comments
To Buy or Not to Buy: Where's the Analytics?
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
Re: Welcome!
  • 10/30/2013 11:56:40 AM
NO RATINGS

Martijn, - That caught my attention too, that netherlands just got netflix, then it hit me netflix wasn't just an site available from anywhere. I had never checked for it until today :)...and its certainly not available in Kenya. On recommender systems, i think they are usually good because they get you started at least from when you have no idea. As you keep using, they get reliably adjusted to one's taste from what i've seen.

Re: Welcome!
  • 10/30/2013 10:09:36 AM
NO RATINGS

Seth, could not agree more. The more data it has to work with the greater the accuracy. It's not fair to rate the system if you are new or sparse user as it is working with limited personal data. As for influence, definitely does affect viewing choices. The choices are so numerous that we default to it.

Re: Welcome!
  • 10/30/2013 2:43:46 AM
NO RATINGS

@Seth, it's indeed unbelievable that NL only just got netflix. It was introduced with lots of attention - up to the primetime news reporting about it. But since we have it, we also consider it as electricity as it gets ingrained in your system pretty fast.. Netflix learning me is still a bit in the bootstrapping phase, but gets better along the way with how many films I watch (which is growing exponentionally fast lately)..

Timeliness
  • 10/30/2013 12:08:41 AM
NO RATINGS

Recommenders are good for about a week I find and then they age and are no longer useful. The recommender needs to predict not just offer options based on past behavior. Then it will not age as quickly.

Re: Welcome!
  • 10/29/2013 10:13:25 AM
NO RATINGS

Yes, I would certainly think that Amazon is on the forefront!

Re: Welcome!
  • 10/28/2013 8:36:18 PM
NO RATINGS

The Netherlands just got Netflix last month?  It has become so ingrained here in the U.S. it's like hearing you just got electricity. 

I find the Netflix recommender is pretty accurate as it learns what you like and don't like. I was an avid user for a few years and it got to know me pretty darn well. 

Amazon's recommender works almost as well, however it is playing in a different arena. It might learn better for book recommendations.  However, just because I just purchased a cook book, doesn't necessarily mean I'm looking for a spatuatta. I think it is much more difficult with physical products.  

Re: Welcome!
  • 10/28/2013 6:49:02 PM
NO RATINGS

Beth, I agree, though I suspect algorthms within a retail site have not caught onto the discoveries found in the search engines within Google, Facebook, Bing.  Social media and the web have distinguished search as conversational, a form of questions and queries. Much of this is through mobile. I am not sure marketers and consumers have fully caught on that some of those attributes can apply to a search within a retail site as vast as Amazon. But I also suspect Amazon (and Netflix since they've been a big analytics player) are woring on their own variation of contextual search that can take advantage of the new features such as voice search and maybe webRTC in the near future.  

Re: Welcome!
  • 10/28/2013 6:19:57 PM
NO RATINGS

So the lingo is a bootstrap.  That is interesting.  

 

At the least every recommendation thrown up by a recommender should have a 'this is not relevant to me.' button.  Like those ads on hulu.

Re: Welcome!
  • 10/28/2013 3:12:19 PM
NO RATINGS

Jeff, thanks for your comment. It's a great idea to bootstrap a recommender a bit. In machine learning, bootstrapping is a much used technique to limit (in a good way) the possibilities that the algorithm considers.

Re: Welcome!
  • 10/28/2013 1:56:48 PM
NO RATINGS

Is it bad karma to ask a few questions for new users?  Like, do you like action films? Then run the recommender based on that.  I'd like to set some options on Amazon.  Like say I'm into a new hobby, I'd like to go on Amazon and say 'show me all new products related to basket weaving.'  I'm sure there is something like this but I don't know where it is...my point is:  Even after initial set up what is wrong with having the user set a few options?  Is this against the idea of a recommender in general?

Page 1 / 2   >   >>


INFORMATION RESOURCES
ANALYTICS IN ACTION
CARTERTOONS
VIEW ALL +
QUICK POLL
VIEW ALL +