A team at Wharton did some Long Tail analysis on the Netflix ratings data the company released for its Netflix Prize. Although I don’t agree with many of the conclusions in their paper (like some other academics, they got confused over definitions of “head” and “tail” and fell into the common trap of doing percentage analysis in an absolute numbers world), the data was interesting. They kindly shared it with me before publication and incorporated some of my analysis in their paper. But for some reason they didn’t use the best part, which was this chart:
The vertical axis is percentage of total demand (with ratings used as a rough estimate of rentals), and the horizontal axis is the popularity rank of the DVD titles. Between 2000 and 2005, the Netflix selection grew from 4,500 DVDs to 18,000, and the effect on the demand of this increase in variety is shown above.
You say that "percentage analysis" is a trap, and yet the y-axis of this chart measures percentages. What gives?
Posted by: James Grimmelmann | September 24, 2009 at 07:48 PM
So how far can this trend be expected to go?
Posted by: twitter.com/shay_guy | September 24, 2009 at 08:06 PM
I'd love to see the graph before and after they made it horribly hard to find the new releases and put movies in your queue as "Long Wait" and my favorite "Very Long Wait". Are people turning to the alternate movies from the slight demand for the long tail, or being driven to the long tail because they can't get the latest movie?
Posted by: www.facebook.com/profile.php?id=507316383 | September 24, 2009 at 08:20 PM
James,
It's using percentages in the inventory (x axis deciles, etc) that confuses everyone. They're not used to inventory growing by orders of magnitude. Everyone, including me, uses percentages in the Y axis (percent of total sales).
See this for more:
http://www.longtail.com/the_long_tail/2006/07/factchecking_my.html
Posted by: Chris Anderson | September 24, 2009 at 09:25 PM
Really a educative and informative post, the post is good in all regards,I am glad to read this post.
Acai Force Max
Posted by: JanCry | September 25, 2009 at 05:08 AM
I was thinking along the same lines as Bryan. Netflix has a fixed number of DVDs and only so many of their new releases can be rented. They send you any old movie from your list so I think that the supply limitation would shift the rentals like we see here. I don't think you would see this trend if you were dealing with a product of unlimited supply (like mp3 sales)
Posted by: Chris Caldwell | September 25, 2009 at 08:39 AM
could it be that the movies being made now compared to the older inventory just suck
Posted by: Thinking | September 25, 2009 at 05:26 PM
People are only beginning to learn the power of anarchic purchasing.
They are learning to buy what they like as individuals instead of what they are told to buy.
This trend will continue.
Posted by: Grant Morgan Czerepak, Canada | September 25, 2009 at 07:20 PM
I'm a very long tail Netflix viewer. I was given a membership for my birthday three months ago. Most of viewing has been films that my local Blockbuster would never stock. I have three conclusions from my brief experience.
1. Choice: Netflix provides choice in a manner that a physical video store or a traditional television, whether over the air or cable, cannot provide. My tastes are highly eclectic, so being able to choose between 18,000 titles gives me endless possibilities.
2. Timely: It isn't instant gratification, unless they offer it online, but it is close. Makes a huge difference.
3. Concentration: There is something satisfying about being able to watch an entire season of a show that you missed, all within a matter of days. When the series is written well, the characters live as they do in those classic long novels of the 19th century. It reminds of Thackery's Vanity Fair which was written and published serially so that each chapter was self-contained and also progressed the story along. The best series do that. All the more reason to watch them in a concentrated timeframe.
The key is choice and the technology to fulfill it.
Posted by: Ed Brenegar | September 28, 2009 at 03:32 AM
Could be just a measure of the changes in the user interface at Netflix. It's possible it got easier to find, keep track of, and be recommended titles in the "middle".
Posted by: Throttled-Netflix Guy | September 28, 2009 at 12:02 PM
Also agree with Bryan Smith. By 2005, Netflix was full on "throttling" its customers so that if you rented more than 9 DVDs a month, the chances of getting the Top 500 were decreased, so you were more likely to get something something from the "middle" in your queue. A phenomena not experienced in the earlier days of the company.
Posted by: Throttled-Netflix Guy | September 28, 2009 at 12:12 PM
Chris,
How wonderful there is a paper to show it's all about you in the end.
Thank you for that. I am so happy for you. So fucking happy.
Posted by: Fake Chris Anderson | September 29, 2009 at 12:07 AM
While Netflix is popular, I hope they know they do have competition and will have more competition in the future. Downloading movies online will be a viable and legal option for a lot more people in 5 to 10 years, I assume.
Posted by: Jessica Deal | September 29, 2009 at 12:13 AM
Part of what drives this behavior for me is the convenience of being able to update my Netflix list whenever inspiration strikes. A number of times I have been watching an interview on Charlie Rose (online), heard them discuss an interesting film, and, without even pausing the interview, opened Netflix in a new tab and added the movie to my list.
Posted by: Cris Cohen | September 29, 2009 at 05:52 AM
Chris,
How does Netflix's ability to stream (at least part of) the catalog -> infinite inventory affect the analysis?
Posted by: JP | September 29, 2009 at 11:32 AM
Was I misunderstood to assume that long tail was about lines on a log-log scale plot?
Posted by: Bertil | October 05, 2009 at 07:09 PM
Yeah I can see why with redboxes and other means of watching movies now available. Although, I did jut sign up for Netflix again because I can watch movies on my 360 instantly, which is pretty cool.
Posted by: Vera Bradley | October 13, 2009 at 07:57 AM
manner that a physical video store or a traditional television, whether over the
chances of getting the Top 500 were decreased, so you were more likely to get something
مقاطع
-منتديات
Posted by: مقاطع | October 26, 2009 at 02:04 PM
Interesting study. Downloading movies online will be a viable and legal option for a lot more people in 5 to 10 years which should shake things up
Posted by: Dana DeArmond | October 29, 2009 at 11:27 AM
I loved the type ad commenting system very clean and very efficient.
Love to read more from you on the topic above
Posted by: watch friends | October 29, 2009 at 11:55 PM
Thanx for the valuable information. This was just the thing I was looking for, keep posting. Will be visiting back soon.
Posted by: r4 ds karte | October 30, 2009 at 04:46 AM
I just read the link and I am blown away with how badly they missed on their analysis. Taking a 10% sample of ratings and then arguing that many of the niche offerings had no ratings at all (implying they had never been viewed) points to a real misunderstanding about the nature of niche markets. The point isn't supposed to be that a few dozen people watch the movie, love it so much that they tell their friends and so on until it becomes a top 100 offering. The point is that someone somewhere got to see a rare find and this is something they would not have been able to do 20 years ago (and the important point from the business perspective is that this customer will happily PAY for this entertainment opportunity). Chief to the academics' findings seems to be the notion that niche offerings are not very popular (...?) What part of niche did they not understand?
Posted by: Slimming Pills | November 09, 2009 at 03:31 AM
While Netflix is popular, I hope they know they do have competition and will have more competition in the future. Downloading movies online will be a viable and legal option for a lot more people in 5 to 10 years, I assume.
Posted by: sasha grey | November 10, 2009 at 10:16 PM
Coca Cola has a long tail?
Not related to Netflix, but definitely related to the 'long tail', is that Coca Cola is introducing vending machines which serve up drinks with flavors customized according to the customer's in-that-moment taste preference. The machine uses flavor concentrate dispensers much like ink-jet cartridges in a color printer, though apparently the actual technology is derived from medical dosing intrumentation. The machines are connected back to the Internet and provide feed-back to Coke in real-time.
Where does the long tail fit in? Well, if you look at a chart of vertical bars representing the number of drinks or flavors chosen by customers and sold via the vending machine, you get a distribution with only a few large, tall vertical bars at the left (eg: Coke Classic), and then a rapid drop-off to a flat, long extended base, where customers are buying a specific, customized, flavor combination only occasionally. As with other long-tail phenomenon, there are however, a fairly large number of these 'micro-sales'.
It seems Coke is experiencing double-digit growth with these sales in a business segment (soft drinks) which has been declining of late, as customers seem to love the element of choice.
Brilliant, no?
Posted by: Rimas | November 17, 2009 at 04:27 PM
I am not able to see the graph. Can you show it what you are explaining here?
Posted by: Colon Cleanse | November 23, 2009 at 08:49 PM
I only signed up for Netflix to gorge on old movies. I very seldom rent new movies. I tried Blockbuster, but they had a very poor selection of old movies. I want what I saw in an ad once "Every movie ever made, online, instantly." Of course, it wasn't true, but I want a movie library. A massive, online, beginning to end, movie library.
Posted by: jaen | December 11, 2009 at 09:11 PM