15 September 2014
I am a Netflix power user. Well, I am a power user of their Watch Instantly service. I have been since it came out, and as soon as I could cancel physical DVDs, I did. I just consume all of my media digitally.
I am also a film nerd, so I consume massive amounts of media. Needless to say, I spend a lot of time with the Netflix UI. I think it is a beautiful UI, and I think the overall experience of using Netflix Watch Instantly is a very pleasant one.
As a power user, though, I am always thinking about things to improve, and so I got my UX hat on, and started thinking about the Netflix user experience more closely.
Netflix absolutely nailed the experience of browsing and picking a movie in the physical world.
Browsing movies at the movie store IRL involves looking at eye-level rows of films, moving from left to right, up then down. That is also how you browse Netflix.
Left to right, up then down. It really is brilliant how Netflix replicated an experience we are all used to doing in the physical world, and translated it into digital.
While I think their replication of the experience of browsing rows and rows of movies at a video store is inspired, this experience is merely replicating something that works in the physical world. The problems a user faces browsing a digital library are different.
Browsing titles on Netflix feels to me like I am browsing a video store blindly. I never feel like I am seeing everything that is available for me to watch, because I know there is so much more — but I just can’t see it.
Inevitably, I go to websites like Instant Watcher or What the Hell Should I Watch on Netflix. These sites exist because it is hard to find something to watch on Netflix. They are external solutions to the problems Netflix users face.
I can see that perhaps the library of titles is so large that a user might feel overwhelmed. Thus, the system is designed to guide users to titles they will likely enjoy based on various data points. However, this leads to a vicious cycle. In this model, users are mostly shown titles similar to other films they have liked. But users may not always want to watch similar films. This is boring. Inevitably, users get stuck in a loop of similar options.
A system based on recommendations of things that are similar does not account for outliers, and human taste is complex and often hard to predict. Random, even. Right now, Netflix doesn’t account for random very well. So, how can a system based on recommendations and personal likes account for randomness?
Give your users more power.
I think Spotify is currently doing a great job at mixing personal preferences, recommendations, while allowing for the element of randomness.
Personal preferences are handled by giving users the ability to create their own playlists. Recommendations are handled by both your listening history, but also supported by seeing what your friends are listening to, and allowing users to share playlists with friends. The element of randomness is supported through their radio feature.
The WWE Network is doing a great job right now accounting for the randomness element. Users can choose the content they want to watch, or select “On Now” and watch whatever is being streamed. Works great for use cases when a user just wants to turn something on. This approach can also help with the discovery element by exposing users to content they might not otherwise have been exposed to.
These are currently the issues facing power users:
In this world, your playlists are front and center when you enter Netflix. This model assumes you can build playlists from individual episodes. You can also determine if a playlist should be “shareable” or private just to you.
When a user clicks on a playlist, they can play it, edit it, or share it with friends.
The current view when you enter Netflix would become the “Library.” This is where Netflix will feature New Releases, Recently Added, and the other recommendations you are used to seeing already (i.e. “Because you watched Bob’s Burgers…”).
When a user selects “Randomize,” a user can randomize one of their playlists, or a playlist from the Netflix library (like say, New Releases). A great option for me would be to randomize the whole freakin’ library, but I can see how opening the whole “information firehose” to users could be problematic, so let’s stick with randomizing content from playlists first.
The main takeaway here is that, when you offer a digital service based on recommendations, it is important to give your users control to personalize the content. This helps keep content organized in a way that makes sense for the user, but also allows for the personal recommendation factor, which can be more powerful for users. The ability for users to randomly consume content can also be a very valuable content discovery tool for power users.