Netflix: Bringing Data Analytics to a Creative Industry

Can data-driven decisions create a sustainable competitive advantage in creative industries?

 

The film and TV business has historically been considered hit-driven and relatively unpredictable. More than half of new broadcast series are cancelled after just one season, and nearly every year we see at least one highly anticipated film flop at the box office.[1] Netflix, however, is revolutionizing the media industry, bringing data-driven decisions to a business historically driven by creativity. As a direct-to-consumer platform, Netflix has access to data traditional media players lack. While traditional media players receive “ratings” for their content indicating viewership among key demographics (broken down by gender and age), Netflix knows who its customers are, what they are watching, where and on what device they are watching, how much of a movie or TV show they watch, the sequence in which they watch, and a plethora of other attributes. Although this post could not possibly not cover all of the ways in which Netflix uses data (in fact, that is likely not public information), it will attempt to highlight a few of the key ways the company uses data to create and capture value.

Creating Value

One of the most readily apparent ways Netflix creates value for customers is through its recommendation system. In 2012, Netflix estimated that 75% of the content its subscribers watch stems from some sort of recommendation.[2] Essentially, the goal of the recommendation engine is to “recommend titles that each member is likely to play and enjoy.”[3] Netflix’s recommendation engine uses a combination of popularity and predicted rating to rank its recommendations, employing machine learning to determine the weights of each independent variable. Over time, the company has added additional variables to further improve recommendations, including member viewing queues, social media, search terms, and demographics.

Netflix is also a proponent of A/B testing as a means to improve the user experience. The company outlines its basic approach to hypothesis testing in its tech blog. Netflix first uses offline testing to as an indication of whether it should go on to pursue online A/B testing. If a hypothesis is validated offline, Netflix then tests the concept online over thousands of users, with A/B tests running in parallel. The process is illustrated below.

Amatriain, Xavier and Justin Basilico.  “Netflix Recommendations: Beyond the 5 Stars (Part 2).”  June 20, 2012.  http://techblog.netflix.com/2012/06/netflix-recommendations-beyond-5-stars.html
Amatriain, Xavier and Justin Basilico. “Netflix Recommendations: Beyond the 5 Stars (Part 2).” June 20, 2012. http://techblog.netflix.com/2012/06/netflix-recommendations-beyond-5-stars.html

Value Capture

The recommendation engine and constant focus on improvement of the user experience not only create enormous value for customers, but also help Netflix to capture value by increasing customer retention. Customers’ preferences are saved even when they unsubscribe from the service, and this personalization often draws customers back.

Netflix also uses data as a source of value capture in content acquisition. Detailed viewership data not only tells Netflix what type of content its subscribers enjoy, but also gives the company an indication of how much it should be willing to pay for that content. Regarding Netflix’s content acquisition strategy, the company’s Director of Global Media Relations says, “We look for those titles that deliver the biggest viewership relative to the licensing cost.”[4] Netflix has also used this detailed data on viewer preferences in deciding which content to create itself. Before investing nearly $100m to order two full seasons of House of Cards, Netflix reportedly considered the share of its subscribers who had streamed director David Fincher’s previous work, the performance of Kevin Spacey films on the streaming service, and the popularity of the British version of House of Cards. According to Chief Communications Officer Jonathan Friedland, this data “gave [Netflix] some confidence that [they] could find an audience for a show like House of Cards.”[5] This data seems to allow Netflix to place bigger bets with a higher degree of confidence, although only time will tell whether this is a true, sustainable advantage.

Finally, Netflix uses data analytics to better market its products, capturing even more value from its content. As an illustration, Netflix created ten different trailers for House of Cards, each created for a different audience segment based on viewing history.[6]

Upcoming Challenges

Netflix’s use of data analytics has benefitted the company enormously, propelling the streaming service to 69m subscribers globally today; however, success is not guaranteed. Netflix faces a proliferation of competitors (Hulu and Amazon Prime, among others). The company also faces a potential threat from its reliance on traditional media players’ studios to supply its content. These competitive dynamics in part explain Netflix’s more recent push into creating its own content, a more expensive and risky endeavor given unpredictable consumer tastes. Although the company has had success in much of its original content thus far, only time will tell if data-driven content creation will create a sustainable advantage for the company.

 

[1] Ocasio, Anthony. “TV Success Rate.” May 17, 2012. http://screenrant.com/tv-success-rate-canceled-shows-aco-172162/

[2] Amatriain, Xavier and Justin Basilico. “Netflix Recommendations: Beyond the 5 Stars (Part 1).” April 6, 2012. http://techblog.netflix.com/2012/04/netflix-recommendations-beyond-5-stars.html

[3] Ibid.

[4] Stenovec, Timothy. “This Is Why You Won’t See Oscar Blockbusters Streaming on Netflix.” The Huffington Post. March 3, 2014. http://www.huffingtonpost.com/2014/03/03/netflix-oscar-movies_n_4892961.html

[5] Carr, David. “Giving Viewers What They Want.” The New York Times. February 24, 2013. http://www.nytimes.com/2013/02/25/business/media/for-house-of-cards-using-big-data-to-guarantee-its-popularity.html?_r=0

[6] http://gigaom.com/2013/02/12/netflix-ratings-big-data-original-content/

Previous:

EFL: The psychology behind risk management

Next:

HypeMachine: Using Data and Aggregation to Surface The Next #1 Song

4 thoughts on “Netflix: Bringing Data Analytics to a Creative Industry

  1. Netflix also uses data to understand exactly what episode a user will get “hooked” on. For House of Cards, I believe it was the second episode. This is quite powerful information and can help encourage certain behaviors that Netflix wants to see in uses (e.g. binge watching). With the recent string of new original content featuring diversity in it’s cast (Master of None, Jessica Jones, Jane the Virgin) Netflix is using it’s data to understand what users want to see. This reduces the risk of product failure and also makes Netflix a medium that many aspiring stars hope to be featured on.

  2. Good article!

    As for myself…

    I am a fan of Netflix, but on a personal level the machine learning tends to vex me more than anything else. I tend to pick one show or movie and stick with it for a while, which means if I see something I like I just “add” it to my list. I wonder how much that alters the list…

    As for your point, I am not so sure how much this applies to all of its series however. When I heard that Netflix’s data led to House of Card, I misunderstood; I thought they created House of Cards based off of the data, but it bought it instead. For that matter, I remember hearing that The Unbreakable Kimmy Schmidt was supposed to be on NBC until its humor- especially its racial humor- scared the network.

    That said, as much as I can say that Netflix just saw opportunities like Arrested Development Season 4 and took them, I do not know what opportunities it may have missed (in retrospect they may have kicked themselves over not getting Community in the United States), or what data may have misled Netflix (the first season of Hemlock Grove had a mixed reaction at best, and the AV Club gives a good description on how broadcast TV at least lets the writers readjust their writing based on reactions to the first few episodes; see http://www.avclub.com/tvclub/emhemlock-groveem-96734. That said, it did get three seasons). In 20 years, when Netflix is either a well-established institution or buried by Comcast or Internet 5.0 or so, it will be interesting to see more “behind the scenes” information on what Netflix learned from how it used its data. That will be a very interesting book indeed.

  3. Great post! I think Netflix is a fascinating example of a successful data-driven company. After reading your post and a case on Netflix this year, I was reminded how the recommendation system actually also helped Netflix manage it inventory more effectively. Netflix is obviously not going to recommend a DVD that is out of stock and thus undermine the benefit of the recommendation system and convenience of Netflix.

    I’d be curious to hear your thoughts on whether you think the recommendation system will help differentiate the company as it transitions to streaming. Or is there something else that Netflix can do with all the data it’s collected over the years to help the company more successful embrace the streaming world. And how should we think about the difference / role of data in the streaming world? I’m a huge fan of House of Cards and do believe that “content is king” but how far can Netflix play in the content world without abandoning / undermining its value proposition?

  4. I’ve always wondered why software platforms like Netflix and Spotify don’t use their skills at AB testing to actually improve content. Imagine a scenario in which a production company releases two versions of a show premiere to a small audience. The show’s performance metrics could be analyzed can be analyzed to determine which version was more effective in capturing user engagement. Netflix has the ability to determine where users dropped the show (if at all), where users paused, fast forwarded, etc…Imagine being able to test where your shows weaknesses are, allowing the production company to make changes before the show’s actual public premier.

    The same can be done (with less work) for songs on Spotify. Spotify knows when users skip the song, trends in listening pattern, etc..and can help artists improve their content.

Leave a comment