Netflix’s House of Cards an Experiment in Data-Driven Television

David Carr from the New York Times wrote a great article this week on how, with House of Cards, Netflix used its mountain of viewer data to give the viewers exactly what they wanted.

Over the years, Netflix has built a proprietary database on what shows its 33 million viewers watch, when they watch them, whether they watch them all the way through, and whether they like them.

Most of the focus with House of Cards is the fact that the show is being released online and all-at-once. What’s more interesting to me it that Netflix knew ahead of time that “House of Cards” would be a hit (full disclosure: Netflix hasn’t released numbers on how many people have watched the show, so we don’t know for sure if it’s a hit. But it’s already the most-watched show on Netflix).

(In his 2008 book Super Crunchers, Ian Ayres looked at how a group of Hollywood outsiders was using using data and Moneyball-style statistics to predict which movies would be blockbusters.)

Netflix is in the unique position now of owning the viewer data and the distribution. It’s like if CBS was the only network with access to Nielsen ratings. Stay tuned. This is going to get interesting.

Google’s Data-driven Redesign

Old and New: Google tested the shade of its logo and whether to use a drop-shadow.

Earlier this month, Google made a dramatic change to the design of their search results pages.

Design is always an iterative process with multiple prototypes and revisions. Usually, selecting which prototypes to use is done by an expert designer. But at Google, the users decided.

In a blog post detailing the re-design, Google explained that Google designers came up with different options, and then they tried each out to see which was the best. In the past, Google has tested small things like what shade of blue to use for link color. (If you’d like to run similar experiments on your site, you can use Google’s free Website Optimizer tool.)

In this case, the “best” meant which design produced the fastest click-throughs. Faster click-throughs mean (1) people are finding what they’re looking for faster and (2) Google can display more search results and AdWords to a person in a day.

This is the design model for the future. Human creativity creates prototypes, which are then rigorously tested in an iterative process using randomized experiments to arrive at the optimal solution.

Two things that make Google’s process successful: randomized experiments over statistically-significant numbers of users and a clear definition of “success” that is the same for the business and the users.