After getting a great amount of real work done today (nothing to do with this Netflix stuff) and seeing a play (see, I have a life!) I made an update to my report with significant editions in the way of "editorializing" (is that a word... I'm too tired to think of another way to say what I mean).
Added summary of the state of my queue ("A") before I began my tests. Added section guessing why Netflix might be doing this and suggestions for improvement. Added additional warnings about what I was not testing.
Thanks for clarifying this point for everyone angle_slam. I have hundreds of movies in my rental queue ("A"). I picked 45 of non-Now titles and added them to 4 other accounts. The point was then to track over time the availability differences for *each* movie in the different queues. I did delete other movies from my queue but found that this did not affect availability scores.
Without looking at the data, the other queues probably averaged 10 other movies each and I added my test movies to the end of their queues so as to not impact my friends and family.
If you send me your Netflix queue, 90 day history, and billing history as HTML I will run it through my scripts and see what comes up. My e-mail address is dvd_rent_test@hotmail.com. Just XXX out the last 4 digits of your credit card if you are concerned about it.
No ax to grind. I started this project to determine *what* was going on, not to flame Netflix. And I found many customers asking similar questions on USENET with no definitive answers. So I thought others would be interested as well.
And I was able to hack around with CPAN's XML classes... they were new to me. And I think I am a fairly competent programmer, so this was really not all that difficult of a task. Past projects have involved hard-core automated data collection on Websites so it was not difficult to leverage my tools and classes.
I would like Netflix to document their behavior or tweak it. The part about the algorithm seeming not to take into effect the rental plan the customer is on (for example, 5 vs 3 disc) is a problem that needs to be addressed regardless, however.
I am still a customer, at least for now. When I have time, I plan to check out the Walmart service.
This is only true if you did not keep any other movies in your queue and no movies were shipped. But I think most heavy renters will have have a bunch in waiting and will continue to have problems getting movies.
Base the selection alogirthm on the length of time the movie has been in your queue and possibly the average position within the queue. Thus someone who has had Trees Lounge in their queue for a month has priority over someone who just added it.
My new "B" account got Trees Lounge. It took a little but to ship but the Netflix system definately flagged me as "top priority, give this guy Trees Lounge vs. the poor sap who has been waiting for it for 2 months."
This is the way to do it. I add movies to my queue when they are in the theater. That way I do not forget about a movie I might miss when it comes out on video. And Netflix often ships the day before it comes out so I get them the same day!
I don't think so. If you have a decent sized queue and at least a few movies with "Now" availability you will still get a movie and it will still cost them postage. I do not get fewer movies because of this, just fewer hard-to-get movies.
I think it has more to do with keeping the light users who pay as much as the heavy users happy so they don't cancel and go to Blockbuster. And of course to keep the new users very happy.
I tried not to editorilize too much in my report I feel more comfortable guessing here.
I had no problems getting *a* movie. So when I was a heavy renter I still received a movie very quickly, usually the same day it was received by Netflix. And the service center is so close to where I live that I get excellent turn around time.
The problem I have is that the allocation scheme is not documented and not uniform. More the undocumented part. When I was researching this on USENET people indicated that Netflix customer reps told them the allocation method was completey random. I have not contacted Netflix myself. It is far from that.
Slashdot Mirror
http://dvd-rent-test.dreamhost.com/
Added summary of the state of my queue ("A") before I began my tests. Added section guessing why Netflix might be doing this and suggestions for improvement. Added additional warnings about what I was not testing.
Without looking at the data, the other queues probably averaged 10 other movies each and I added my test movies to the end of their queues so as to not impact my friends and family.
Based on some excellent feedback I updated the document. Changes are marked in yellow.
I got a chance to make some updates based on your input. Thanks. See http://dvd-rent-test.dreamhost.com/. Changes/additions are marked in yellow.
There were only three movies in the new account ("B") for a few days. I then added at the 45 test movies plus other movies that I wanted to see.
Thanks. I will check this out and get back to everyone and update the report as necessary. Nice to have "peer review" by folks smarter than I.
If you send me your Netflix queue, 90 day history, and billing history as HTML I will run it through my scripts and see what comes up. My e-mail address is dvd_rent_test@hotmail.com. Just XXX out the last 4 digits of your credit card if you are concerned about it.
No ax to grind. I started this project to determine *what* was going on, not to flame Netflix. And I found many customers asking similar questions on USENET with no definitive answers. So I thought others would be interested as well. And I was able to hack around with CPAN's XML classes... they were new to me. And I think I am a fairly competent programmer, so this was really not all that difficult of a task. Past projects have involved hard-core automated data collection on Websites so it was not difficult to leverage my tools and classes. I would like Netflix to document their behavior or tweak it. The part about the algorithm seeming not to take into effect the rental plan the customer is on (for example, 5 vs 3 disc) is a problem that needs to be addressed regardless, however. I am still a customer, at least for now. When I have time, I plan to check out the Walmart service.
How did you guess :-).
This is only true if you did not keep any other movies in your queue and no movies were shipped. But I think most heavy renters will have have a bunch in waiting and will continue to have problems getting movies.
Base the selection alogirthm on the length of time the movie has been in your queue and possibly the average position within the queue. Thus someone who has had Trees Lounge in their queue for a month has priority over someone who just added it.
My new "B" account got Trees Lounge. It took a little but to ship but the Netflix system definately flagged me as "top priority, give this guy Trees Lounge vs. the poor sap who has been waiting for it for 2 months."
I did not find that queue size affected availability. I had the same scores whether I had the 45 test movies in my "A" queue or 300 movies.
This is the way to do it. I add movies to my queue when they are in the theater. That way I do not forget about a movie I might miss when it comes out on video. And Netflix often ships the day before it comes out so I get them the same day!
I think it has more to do with keeping the light users who pay as much as the heavy users happy so they don't cancel and go to Blockbuster. And of course to keep the new users very happy.
I tried not to editorilize too much in my report I feel more comfortable guessing here.
The problem I have is that the allocation scheme is not documented and not uniform. More the undocumented part. When I was researching this on USENET people indicated that Netflix customer reps told them the allocation method was completey random . I have not contacted Netflix myself. It is far from that.
I rented U.B. for Ms. Richards! But she could not hold up that movie. Sorry.