This was a continuation of work we had already been doing for roughly 12 months and was taking a big swing at really trying to kickstart virality in a market.

This doc was not a strategy doc but rather a more tactical doc for the design, data science and engineering team working on the Facebook integration at Netflix. The goal was to provide as much detail as possible so the team could begin implementation.

This project was reasonably straightforward. Many companies had implemented Facebook for registration. What was unique was that we were requiring Facebook to sign up for Netflix and that had large risks, e.g. we could be putting too much friction into the signup process and will lose some momentum with regards to Acquisitions.

We chose these countries to test our Facebook integration because we were worried about testing on our large member base in the United States.

Also, because virality is hard to test (If I share a movie or TV show with someone, the person I'm sharing with also needs to have access to the experience, thus making it hard to create a true Control group for our experiment), we needed to enable this in countries where there was less risk of losing members and that had comparable comparison countries.

In a typical test, we would test everything globally, not country specific.

This was our second attempt at creating a viral experience within Netflix. The first time we tried this was pre-Facebook and we never built the social graph to be large enough. This time we were building off of Facebook's social graph instead of trying to build it on our own.

We wanted to test the extremes: requiring Facebook, Strongly Suggesting and making it completely optional to see the results across distinct approaches.


If this test worked well, the experience of consuming Netflix would have changed in the following ways:

1. A member would have seen what their friends are watching within the Netflix experience. We tested this and it performed well.
2. A non-member would have seen what their friends were watching on Facebook. We tested this and it did NOT perform well. We learned that people neither expected nor wanted what they were watching to be shared to Facebook. Speaking for myself, I want people to believe I watch all intellectual documentaries but in reality I will binge watch Dancing with the Stars :)

By calling this out, I'm acknowledging that I'm trying to push the team to move faster and get this shipped by Nov. 16th, but also making it clear that there is a decent chance that we miss that date by a few days.

This was a straightforward feature to implement and needed Design, Engineering and Data Science to implement. When we first implemented Facebook Connect we had to do a lot of work with our database to enable a new account type where we pulled email from Facebook, but once that was complete this was decently straightforward. The other implementation that took some work was building out the graph of Facebook connections that also use Netflix and serving up recommendations on Netflix based on what they had been watching. This was a popular feature and took some engineering and design work to implement well.

This stands for Social List of Movies, e.g. a list of movies and TV shows someone's friends watched.


I always push myself to try to think of everything knowing that I would not. By pushing myself, I was able to save my engineering counterparts time and allow them to think of even more non-obvious edge-cases I hadn't thought of.

It was really important for us to balance privacy concerns with trying to create a viral experience and we thought hard about how to do so. We tested an option where on every movie or TV show a member watches the default behavior is for it to share to Facebook but they could "unshare" it. This approach worked well but was rather clunky.

We did this to make it as frictionless as possible but it definitely added complexity to our implementation.


These designs were done. My designer delivered the actuall assets separately but I put screenshots in the Spec to help people understand better what we were building.

These were originally designed in English before they went through translations, which happened during engineering implementation.


We let this test run for ~6 months until determining that we weren't seeing enough benefit from virality to justify the friction we were adding to the sign up experience.

We did see quite a bit of people connecting to Facebook, but that's obviously what we expected since we were requiring it (or strongly suggesting it).

We also found that strongly suggesting Facebook was the right approach. It got the vast majority of of new members to choose to create their account with Facebook while providing a release valve for those that were very opposed to the idea.

Tom Willerer

This document outlines the requirements for design, data science, and engineering to implement Facebook integration at Netflix. We were taking a big swing at kickstarting virality in a specific market.

 is off to a great start and our goal is to get at least 30% connected 6 months post launch of our integration in a market. 30% is our first goal, but the real goal is to connect a majority of our members so that they experience a social Netflix and create a 

viral distribution channel for us via Facebook.

 This test will seek to understand the trade-offs of requiring Facebook at sign up which gets us to density very quickly with the hit to new member trials, if any.

By requiring Facebook Connect when someone (everyone!) signs up for Netflix we will have more members sharing what they are watching to Facebook, which will in turn cause more people to see little friend endorsements of Netflix on Facebook and thus more sign ups then could otherwise be achieved by not requiring Facebook Connect.

The markets we considered were Chile, Colombia, Argentina, and Costa Rica. Chile is the most Facebook penetrated followed by Colombia and Argentina with Costa Rica the least Facebook penetrated. From a new start perspective, Argentina is one of our weakest markets, with high levels of piracy and Chile and Colombia are middle of the road. Costa Rica is the smallest of the markets we considered.

The market for strongly encouraging Facebook will be Colombia.

We will have Canada on hold ready to launch 

this is a touch early from initial estimates)

For this test we will pull in a person’s email via Facebook, but not ask them for a password initially. In fact, we won’t ask for their email either, instead they will click a button and we will pull it in from Facebook (with the person’s permission). At the same time we will auto generate a password and send it to the new member via email. Going forward members will be able to login using their Facebook Credentials or Netflix Credentials. [

We will allow members to disconnect their Netflix and Facebook accounts, but still remain a member. To be aggressive without overwhelming members we will ask them to connect two more times via the Connect Layer per Evers’ postcard service. On the PS3 we will just display the Facebook connect 

 list. When they disconnect from Facebook we will NOT prompt on the spot for them to update their password. Instead, if members don’t know their password when they are logged out they will have an option on our login screen that recovers it for them (this is already built). Once disconnected a member can reconnect with a different Facebook account.

For PS3 login, since we are allowing members to enter either Facebook or Netflix credentials, we should authenticate both credentials in parallel.

what do we do if the Facebook email is not unique, i.e. we already have that email in our database? Kick them out of the test – so anyone who tries to Facebook Connect and the email address is already known to us, we toss them from the test, and put them back in the normal Netflix signup flow. [

This highlights a potential fraud deterring benefit we get from requiring FB Connect. Harder to create fake accounts.

Edge Case: in the case where Facebook is unavailable, thus blocking members from signing up, we should, upon detecting an error, flip them out of the test and back in the normal Netflix sign up. 

Edge Case: in the case where someone signs up and attempts to connect a Facebook account that we already have connected we will provide them with this error message and allow them to try connecting a different account. “That Facebook account is already in use. Please use a different Facebook account to sign up for Netflix.”

Error Case: What if parallel email / password call on PS3 fails? In this case we should present the following generic message, “The login information you entered does not match an account in our records.”

Error Case: If not Facebook connected, but try to login with Facebook, present this message, “The Facebook account you are using does not match an account in our records. Please sign in using your Netflix email and password.”

There are many different things we could do to allay privacy concerns, including doing what Spotify recently made available, which is allowing users to go in a “private listening mode”. This mode is not “sticky”, thus each time a user closes the Spotify app it defaults back to sharing.

In my view, this model is confusing and instead we will continue TESTING our unshare/hide model in these markets for the duration of this test. MAY ALSO JUST ROLL OUT UNSHARE/HIDE INSTEAD OF TEST IN REQUIRED MARKET.

We will pull in a members’ email from Facebook and then 

auto-generate a randomized alpha numeric password that should be language agnostic. We will also support click-thru to a page that allows a member to update their password.

The email we will send to members should be focused on letting them know what email we will be using and telling them that we have auto-generated a password for them, with click-thru to update.

Netflix comp coming shortly. Text is as follows:

Welcome to Netflix! You're now Facebook Connected so you can sign into your Netflix account using the Facebook login button.

We've also provided your Netflix sign in information for you to use directly on the Netflix website or through any device that does not support Facebook Connect.

An example with how Mog does this is below:

For this test we will require the publish_actions permission and also require they give us permission to pull in their email via Facebook.

If a member does not accept our request to Facebook Connect we will redirect back to the homepage. This behavior will repeat until / if they accept our request to connect.

in the “Free Trial” section, add a question:

Answer: “It’s easy – to start your free trial, login using your Facebook account. If you don’t have a Facebook account, you can create one for free (hyperlink). <in market 2: If you don’t want to use your Facebook account to login, you can also signup using your email address (hyperlink to regular Netflix register fields).>”

Question: How do I stop sharing my viewing to Facebook?

Answer: You can disconnect from Facebook to stop sharing.

Question: Do I have a specific Netflix password?

Answer: Yes, it was auto-generated and emailed to you when you signed up using Facebook. Go here (add url) to update your password.

For a rejoin, we will require them to connect prior to confirming their credit card. If they had already been Facebook connected, we should ask them to connect again, rather than keeping their previous connection.

For remarketing, we will require them to connect prior to confirming their credit card (same as above). These folks will not have been connected previously so we don’t need to worry about that scenario.

Rejoin is a nice to have for this test, but not necessary. Remarketing we will do for the test.

The non-member homepage on the PC/Mac and PS3 will be reworked to ensure members understand that they will be connecting to Facebook in order to use Netflix. For the PC/Mac we will use the Login Button. Documentation can be found 

. For the PS3 we will use our own UI and the login API (already built). The actual connect screen should go after the initial connect screen and before the credit card screen on both PC/Mac and PS3.

Product Spec: Requiring Facebook Login at Netflix in 2011

Spec: Facebook Required

Sign up to see more