Netflix: Big Data on film and television creation

Many subscribers to Netflix may feel the same way that they opened the Netflix device in Friday a few weeks ago and happened to see a TV show called "Women's prison" on the Netflix recommendation, and clicked on to start watching the first episode, Then wondered how he never heard the name of the play. As we all know, Netflix's other original programmes, including the TV series "Solitaire" and "stunted development", have all had great market benefits, but they have been less successful than women's prisons.

Why has the women's prison been so successful? The answer, like many other questions, is the data. Netflix does not need to spend millions of of dollars on new programs and guide users to the table because it knows that users will refer to their recommended columns and try to watch the recommended new play, and that the user will love it. According to the company's surplus conference call in Monday, the "Women's prison" was more highly rated and the overall playback time longer than the TV series aired before Netflix.

Of course, the success of a play, such as a women's prison, is not entirely dependent on data, it must have a good script and first-class acting, in addition, the play is based on memoirs, rather than the interpretation of the concept of ordinary television. But it is undeniable that the data do help Netflix analyze the audience for the cold and the way the plot is interpreted. It can be said that with the development of digital television, the television industry itself must be transformed, it should be more like a Web page, not only can analyze the behavior of visitors, but also through the data to understand the layout of the page or the user experience of the subtle changes. Of course the plot is kingly, but to adapt to the audience's taste and achieve success, any small factor is extremely important.

Neflix executives ' answers were vague when asked what kind of audience they liked to watch. However, we can understand one or two only by looking at the data analysis behind the company's proud program recommendation system. According to one of the company's former data analysts at the Hadoop summit last year, we can see Netflix's tracking data as follows:

More than 25 million subscribers

About 30 million times per day (this data includes the number of times a user has to rewind, fast-forward, and pause the video)

In the last three months of 2011 alone, streaming media video viewing time has exceeded 2 billion hours

About 4 million ratings per day

About 3 million times a day

Geographic data

Device information

Watch the date and week (current analysis shows that TV dramas are more popular on weekdays, while movies get more viewers on weekends).

Meta data from third parties, such as Nelson

Social media data from Facebook and Twitter

Addition：

"One of the most interesting aspects of Netflix's data application may be its attempt to analyze the film itself." Netflix is now able to capture JPEG images and record the exact time that the video caption began scrolling, and Netflix is planning to expand its analysis of other aspects of the film, such as volume, color, and scenes, which can reflect viewers ' preferences. ”

We do not understand the impact of the data in the plot content creation, but it should be a data acquisition type, but also with the user fast forward and backward driven by the relevant factors. The data can reflect much more than the traditional metadata can refract, and traditional metadata is generally considered when deciding to launch a program, and the New York Times media reporter David Carr published an article in April this year and elaborated on it (citing the above data points as a case of Netflix's focus on data). Of course, this data acquisition and Epagogix model is not the same, epagogix use of artificial analysis and algorithms to predict the box-office revenue of the screenplay.

It can be said that it seems to be very easy to figure out what kind of program the audience likes to watch. For example, from Netflix and the women's prisons, we can almost see the results of predictive models that are deeply analyzed for high ratings programs such as prisons, FireWire, sopranos, Nurse Betty, bloodthirsty judges, and single parents. In this respect, we can also look for data analysis from series, prison dramas, crime dramas, black humor dramas, female main dramas and other episodes.

But is over-reliance on data leading to monotonous and formulaic plot content? There is a mixed argument about this. There is no problem in predicting when an aircraft engine will fail, or whether an email is spam, but the duplication of content is overshadowed if it comes to originality and innovation. For example, the introduction of black humor in prison plays should not be defeated, but it is not possible to red the sky.

Of course, we should not underestimate these data, they may be the final success of the series is an important factor. Think about how we interact with content on the Internet or mobile devices. For example, if you are interested in a topic or post, we will click on this topic or the title of the post to read, but if the first 400 words of the article are bragging about unimportant things and away from the focus, or the design of the page is bad or a lot of advertising bombing, then we will definitely choose to exit directly. If it is influenced by the page ads, it is likely that the reader will leave and no longer return, because in addition, the audience can find the information they want to read in many other places.

Whether it's Netflix, DirecTV, or cable TV, it's the same as providing television broadcasts. At some point, we can assume that the director must also start thinking about how to display titles. When the program is playing, what factors will make the audience decide to suspend viewing or give up completely? Should we avoid certain factors that prevent viewers from making fast forward moves in watching programs? Should we add more content to make viewers willing to go back and watch again? (digression: Women's prisons are really a good play, but on the other hand the play is full of sex and nudity.) ）

Of course, what we're talking about here is big data, big data that can help your program succeed, for example, you can tell a yuppie football father's life as an eco-terrorist, full of mystery and adventure, and believe that such programs will be more interesting and more enjoyable than other programs.