How data can help predict and create video hits
By Ryan Lawler
In the hit-driven content world of cable and broadcast TV, it can be difficult to predict what will be a blockbuster. Network execs have tried for decades to perfect the art of picking and producing TV shows that viewers will love. Even after an extensive review and pilot process, well more than half of all shows fail and are canceled in their first season.
Online providers like Netflix and Amazon are trying hard to buck that trend by using the growing influx of data from their users to evaluate new, original programming and to help choose cost-effective content to license from other creators. What they are showing is that with a solid bit of data in their pockets, they can make well-informed guesses about how content will perform.
Short-circuiting the pilot process
Netflix has long been a licensee of other producers’ content, creating a large on-demand library of rerun TV on the Web. But when it came time to start creating its own original programming, Netflix didn’t start shooting pilots like most traditional networks would. Instead, it used its vast amount of streaming video and DVD rental data to help determine the pieces of content on which to place its bets.
Netflix’s first piece of original programming, House of Cards, was chosen in part because the streaming and DVD provider had high confidence that all the main parts — the story, the lead actor and the director — were well-loved by its users. House of Cards is based on a popular British miniseries that performed well with Netflix DVD users before recently being added to its streaming library. The new U.S. adaptation will star Kevin Spacey with David Fincher producing, both of whom have loyal followings among Netflix users.
Netflix was so confident in the data that it ordered 26 episodes of House of Cards, or a full two seasons’ worth of shows. That is an unprecedented move, especially for a new entrant into the original programming market.
Bringing shows back from the dead
House of Cards was the first piece of original programming that Netflix bet on, but it is not the only one. Due to the massive popularity of Arrested Development on its platform– both in streaming and DVD form — Netflix decided to invest in bringing the show back from the dead, about five years after it was taken off the air by Fox.
When discussing the Arrested Development deal, Netflix Chief Content Officer Ted Sarandos told me by phone it was one of the series that actually became more popular and gained more viewers as time went on.
Because Netflix is an on-demand service, its viewers aren’t tied to a certain time and place for watching their favorite shows or even catching up on some they might have missed. That has led some series, like Arrested Development, Party Down and Battlestar Galactica, to earn more fans over time.
We might see Netflix pick up more canceled series, since it can identify an existing longtail of fans. And its recommendations algorithm can increase the audience by introducing viewers to shows they might have missed the first time they appeared on TV.
From online retailer to online cable substitute
Netflix isn’t the only streaming video provider to leverage its data to help it choose which content to license. Amazon also relies on its vast stores of customer data for its own subscription streaming video service.
Amazon probably has more data on its users than seemingly any company in the e-commerce industry, and the Web retailer giant has long used that data to predict which products users might want to buy. But a lesser-known story is that Amazon is also using its large data sets to help the company predict which TV shows and movies its users will be most likely to watch via its subscription Prime Instant Videos product. And that is enabling Amazon to deliver more-popular videos and negotiate good deals for underappreciated video content.
According to Brad Beale, the head of Digital Video Content Acquisition at Amazon, the company uses data from both its DVD and streaming catalogs to pinpoint content that is worth going after. “One of the coolest things about working here is that we have lots of information around what our customers value. We have information about what DVDs people buy, and information around the VOD platform. We use that information with our content partners to figure out what’s the right content for us to acquire,” Beale told me.
Since introducing Prime Instant Videos a little more than year ago, Amazon has used its data to help it triple the number of titles available through the platform, from 5,000 at launch to more than 15,000 today. And with more content coming online all the time and more users beginning to take advantage of the streaming offering, Amazon gets even more data with which to improve its library.
Finding the right audience
Having the right data might help services like Netflix and Amazon license movies and TV shows they know their audiences will love. But it is not all about content selection. Just as important is the ability to recommend the right content to the right users, something Netflix and Amazon have been doing successfully for years.
The big knock against Netflix’s streaming library is that it has a fairly limited amount of popular new release titles. However, what it lacks in new releases it makes up for in the power of recommendation, suggesting movies and TV shows its users might not already know about.
As it relates to Netflix’s and Amazon’s abilities to build audiences for their streaming content — whether it be for original programming or videos they have licensed — customer data is incredibly important for being able to recommend new content. Sarandos told me it is rather surprising how similar viewing numbers can be between a highly popular Disney title like Cars 2 and some more-under-the-radar Hollywood fare.
While the streaming video industry is still small compared to the traditional cable TV ecosystem — both in terms of subscribers and revenues — the online players have one big advantage over the more established networks: Using a wide range of customer data, they can not only choose content that will likely be popular with users but also boost viewership by introducing them to new things.