Are you really trying to encourage a certain behavior or sell folks on a certain movie? If what you work on is a subscription service, the company has the money, you just need to keep the customer, maybe wring out an upgrade or a sale when possible.
The whole reason I made this post was that two separate flicks from Tom Selleck’s legendary ‘Jesse Stone’ made-for-tv movie series with nearly identical names & covers showed up in the recommendations box. I can’t expect a program to scan the image for similarities, but if you can beat that, you’ve got something on Netflix.
If they’d just sign a contract, we’d build it for them. We work for their competition. None of the information comes from the images, there are 3rd party data sources on all movies released in the US with information like that.
We try to encourage whatever behavior the person paying us wants. That can be very different for a purchase model vs a subscription model, or a rental vs purchase, and depending on what else is on the screen with the recos.
No, I did actually rent The A-Team. They’re one of several things I don’t mind watching when I’m in a 'splodey mood (they blow things up nicely). But why would that lead to a recommendation for Voltron?
I usually don’t bother with the recommendations Netflix gives me because they always seem to be way off. So what I do is usually just click a “Not Interested” on a movie it may recommend and hope it slowly fixes it after a couple years. : p
I think it is they are equally bad in terms of content but people think I remember watching that when I was a kid and decide to rent it.
I’ve found net flicks to be pretty good at suggesting movies. It is far better then blockbuster online or the other couple services I’ve tried.
My only issue is the endless lesbian movie recommendations. I watch allot of gay male movies. I find most gay movies are pretty bad to begin with so the shirtless guys make up for allot. Those shirtless gals just need to stay off my recommendations.
I just started a thread about Pandora.com similar to this… what kind of algorithms do they run. Their system seems to be much more in tune with recommendations, much more than the Amazon.com system I use to buy the music i hear on Pandora ha.
Netflix is convinced that I only like foreign films and hate comedies except for a couple stock 1980’s farces.
I swear, I told the stupid thing that I enjoyed Kieslowski’s Three Colors trilogy and Maria Full of Grace and now it shoves every foreign film ever made at me – “Hey! You enjoyed Red so here’s a Pakistani made documentry on bicycle factories in Qatar! Huh? Huh?”
These recommendation engines know nothing about individual movies. They work one of two ways:
Collaborative Filtering: looking at raw numbers, of people who liked (rented, rated highly) movie A what other movie do they have in common.
Analysis (various names): based on the attributes we know about movie A (genre, year of release, actors, length, country of origin) find other movies that are similar
There are combinations of the two, and quirks in the systems that weight certain selections higher then others. So, if the system allows you to rate movies, the more you rate the better the system will recommend you new movies, at least in theory. Usually, new ratings are valued higher then older ratings.
Click on the above link to go on my friend’s list.
To find your link, just go to Friends>Invite Friends>copy the link that will in the right hand box>post that link here so we can go on each other’s friends lists.
It allows the people on your list and those on yours to share recommendations, ratings, favorites, movies in common, percentages of how close you are to that person (based on ratings of movies) and queue info…
…however if you like the thought of a friends list but don’t like the thought of people maybe seeing some, ahem, personal movies in your queue, there’s also a new feature under Friends>Movie Privacy that allows you to set movies so that people on your friends list CAN’T see specific ones (just in case you want to rent that Skanks From Mars movie but don’t want everyone to know/see you’re renting it).
Anyway, feel free to click on that link and join my friends list. You can have up to fifty friends. Me, I have only three or four spots still open.
Maybe I got plebian tastes or something, but the Netflix recommendations work well for me – a little too well. When I’m going through them it’s, “Seen that, seen that, own that, seen that.” I find about one in eight is a movie not yet experienced.
There’s probably not much of an informative answer to be had. All they can say is that their algorithm has grouped you with people who like that particular item. At best, they could give you a big matrix that shows why you ought to like that movie. But it still wouldn’t have to do with the content of any of the media, only with the people who bought some of the same things you bought.
For a while Netflix was recommending nothing but gay movies. Gay gay gay movies. I have no idea why - but I suspect it is because one of my Netflix “friends” is gay. I mean, I like gay movies, there are plenty of good ones, but I am primarily a sci-fi and action fan and couldn’t understand why it was doing that. So I spent a good couple of hours rejecting page after page after page of gay movie recommendations. I must have rejected a hundred gay films. I went into their preferences system and did a lot of work there. Now at long last I am getting recommendations for movies I actually want to see.
Amazon is the same way. It can’t tell the difference between a random gift purchase and your personal purchases. I’ve learned to go in and remove those items in order to get better recommendations. But then I still get recommended stuff like every Cure album made after 1992, all the Depeche Mode albums, etc. If I wanted that stuff, I would have bought it 15 years ago. Show me something a bit newer, please.
Release year, genre, and age group are all probably pretty similar for these shows. Just looking at the attributes it’s very likely that they appear close. Also, there may be data that shows that people who liked The A-Team also liked Voltron. You might not, but when dealing with aggregated data you often find interesting and unexpected correlations.
Actually, Life of Mammals is damn good. I can’t think of anyone I wouldn’t recommend it to, no matter what their other movie interests are. Kept my middle school kids enthralled.
Amazon thinks that because I’ve bought a book by Peter Robinson, who writes crime fiction, I might be interested in Hard Labour: The Sociology of Parenthood. I have no idea how they made that particular connection.
None of the information comes from the images, there are 3rd party data sources on all movies released in the US with information like that.
You can have up to fifty friends. Me, I have only three or four spots still open.