“Also recommended for Sarah…” – Thanks, Netflix.

From improving subtitles to choosing which movies to license, to even creating personalized trailers for their proprietary shows, Netflix creatively uses data capture and analysis to satisfy, retain, and attract customers.

When I first sat down to write this post, I planned to focus on Netflix’s success in utilizing user data to provide recommendations for users on what to watch. But the more research I did, the more I found out that this is just one of many examples of how Netflix uses data analysis in creative ways. From improving subtitles to choosing which movies to license, to even creating personalized trailers for their proprietary shows, Netflix is all over it.

Did you know that Netflix collects all of this data about you every time you log on? (source: Kissmetrics)

  • When you pause, rewind, or fast forward
  • What day you watch content
  • Date watched
  • Time watched
  • Where you watch (zip code)
  • What device you use to watch (iPad, etc)… and for what (TV shows vs movies)
  • When you pause and leave a show
  • Ratings you give (about 4 million per day)
  • Searches (about 3 million per day)
  • Browsing and scrolling behavior

Netflix recommendation system

It’s like that best friend who knows you love comedies and that last week you watched Zoolander and Groundhog Day. They don’t think you’ve seen School of Rock, so how about checking it out?

Netflix’s proprietary recommendation system utilizes user data to provide personalized recommendations on other movies and shows to watch. When Netflix users first sign up, they are asked to fill out a short survey with their preferences and rankings on movies. Netflix then uses this, along with data collected about past viewings, to recommend movies and shows to users. This is a huge value add to the user – not only do they get recommendations on movies based on their preferences, but they also get exposure to films (sometimes older) that they otherwise wouldn’t have even heard about. And hey, it works. Over 75% of viewing activity is driven by these recommendations.

Licensing movies

How, you may ask, does data help Netflix decide which movies to license for their collection? Netflix’s goal when deciding is to figure out what users will enjoy the most. For example, say a huge box office hit was just released, and would come at a very hefty price tag to Netflix. For the same price, Netflix could license 5 or 6 other movies by the same actors and/or genres, and get a total user enjoyment of even more than what it would be with the box office hit. Jenny McCabe, Director of Global Media Relations, says that to do this,

“We look for those titles that deliver the biggest viewership relative to the licensing cost. This also means that we’ll forgo or choose not to renew some titles that aren’t watched enough relative to their cost. We always use our in depth knowledge (aka analytics and data) about what our members love to watch to decide what’s available on Netflix….If you keep watching, we’ll keep adding more of what you love.”

Personalized trailers on proprietary shows

Netflix has some incredible shows of their own – House of Cards, Orange is the New Black, and the list keeps growing. What you maybe didn’t realize is how deep Netflix gets in using data to make these shows appeal to you, specifically you. For the launch of House of Cards, Netflix made ten difference trailers, and your viewing behavior dictated which of these trailers that you saw. You love watching Kevin Spacey films? You without a doubt say the trailer that featured him prevalently throughout. Well then of course you will want to binge watch this new series featuring your favorite actor, Kevin.

Conclusion

Netflix’s creative use of data goes hand in hand with their success thus far, and ultimately drives their user satisfaction and retention that will drive this growth going forward. Netflix will soon (if they are not there already) be able to compete with – and defeat – the HBOs of the world because they are able to capture and creatively utilize so many data points from their user.

Previous:

A/B Testing for All!

Next:

import.io & Web Scraping for the Average Joe or Jill

Student comments on “Also recommended for Sarah…” – Thanks, Netflix.

  1. Netflix’s data approach is certainly one of its major benefits. Particularly as a way to sort through the large library of movies available, its true value add is limiting and tailoring customer choice. I am excited to see where Netflix will go next. Instead of modeling based on the movies/shows you watch and then recommending other similar ones based on their own knowledge and what others have watched, it seems there is a lot more potential. Other data variables could be added to improve sophistication such as time of day and day of week and if you are alone or with friends or on a date etc. This could increase the value add even more.

    I also wonder if the Netflix approach to data could be used for other products, such as music?

  2. Interesting post – I too had primarily considered Netflix’s use of data surrounding the recommendation engine, so it’s cool to see how else they use it. I worry a bit about two things. First, how defensible is their position? Many new and traditional players are getting into the SVOD space and it’s impossible to compete on licensed content alone. Their two sources of differentiation are their use of data and their original content. Other players can compete on the data front: HBO Go can gather just as much data from their users and curate the content. That leaves only original content. I think using data is a great way to figure out which shows Netflix should go with based on what their existing users already enjoy, but I worry that this will be at the expense of truly innovative artistic productions.

  3. Thanks for a great post — I had no idea how much data Netflix was collecting on me when I watch, but I’m not necessarily surprised. I wonder if something else they might move into is connecting Netflix accounts between friends (or adding “friends” on Netflix), and then combining viewer data to start suggesting movie / TV nights as social gatherings based on what you and your friends have in common. That might also impact the way in which they think about releasing new content as well as pitching viewership to content creators in order to license their content. Netflix might also be able to do this internally within the different profiles in a single account (I believe you can create up to four profiles under one paying membership).

  4. I think Netflix got the balance right between art and algorithm, when it comes to a show such as House of Cards, in particular. Amazon, for instance, which forayed into producing its own content using a heavily data driven approach, was far off the mark in achieving this balance between art and data, and all but one of its shows have flopped after a few episodes. A few questions come to mind regarding Netflix’s use of data, and the use of data in artistic content creation in general –
    How do they balance data with human judgment in the creative process? I too, like Salem, worry about what the overemphasis of data will do to the artistic process. What will become of truly innovative art, if producing of new content depends on determining which existing ‘variables’ of content are most pleasing to the average viewer? A lot of art that later becomes popularized in our culture was not necessarily popular with the masses at the beginning. Take the film “Birman,” for instance. I doubt this type of film would have been produced using an algorithmic approach, as it wouldn’t prove “popular” enough when evaluated against an algorithm’s recipe for success. A film like Birman, however, does end up having a tremendous impact on popular culture, and artistic production, because of the opinion of just a few “expert” human reviewers, rather than the data from the masses. This “expert” opinion may not align with what the masses may have initially thought, but serves to influence and inform culture in an important way, as their “review” trickles down. I wonder what will happen to the more unusual/”out there” types of art if content generation were to come to rely too heavily on data and algorithms, at the expense of the human touch.

Leave a comment