Client
Designlab UXA
Individual project
Duration
Jul - Aug 2020
4 weeks, 20 hrs / wk
Contribution
UX design, research, visual identity, copywriting, prototyping and testing, responsive design
The 'Endless Scroll' of not being able to decide on what to watch and the inaccuracy of Netflix's recommendations.
Create a feature that deepens user engagement, by providing Netflix with more data to work with for improved personalization, and minimizing the cold-start problem.
I didn't have a clear vision at the outset of the project, and needed further insights into Netflix users' consumption habits. To get a better idea of who and what to design for, I did some secondary research, user interviews, and surveys.
According to a study in 2018 by Netflix, most consumers still use their television set to watch online content.
To confirm my secondary research, I conducted a survey of my own with 20 participants, with 48% saying they prefer to use streaming platforms on TV, 25% on desktop, 21% on mobile, and 6% on tablet.
I broke down my secondary research into four parts : 1) people's video consumption habits, 2) the video-on-demand market, 3) competitive research, and 4) Netflix's recommendation and personalizations system, with the focus on the fourth.
Taken from a recent Netflix research study on the trends in personalization, Netflix personalizes using the following:
- Ordering / ranking of videos horizontally
- Selection and placement of rows
- Personalized thumbnail images
- Reaching out to members directly about new or upcoming titles
But even as Justin Basilico, Research/Engineer Director at Netlix, says --
personalization is hard: Every person is unique and usually don't single out one type of content to consume
Risk of overpersonalization without providing novelty, freshness, exploration, diversity, etc.
Users want to see different kinds of content depending on the circumstances they are in
Users don't necessarily open Netflix knowing what they want to watch
I gathered opinions and suggestions from a total of 20 Netflix and/or other VOD streaming platforms users through a mix of informal interviews and survey responses.
Ages range from 23-47 y/o with an average age of 29 y/o - all but two participants based in the US.
"They have access to shows I can't get elsewhere"
"We use Netflix to watch 'trendy shows' as conversation starters with family and friends"
In gauging whether the problem of choice was a real problem among users, I asked the participants, on a scale of 1-10, do you struggle with picking what to watch? :
" The content choices presented to me are random and feel like they're based off of other peoples' tastes... Netflix just doesn't seem to understand my personal tastes "
" The recommendations never work for me... not only are they inaccurate, but when I'm just testing a show out, Netflix would think -- 'because you love this' -- and show me a bunch of content within that realm "
It seems that the problem of choice paralysis exists for many. Based on the insights I got from interviews and survey responses, I drafted a user journey map based on a persona who is frustrated of getting a Netflix experience that doesn't feel personalized, and chose to engage with a recommendation feature that only becomes visible after a certain amount of browsing time.
Using this user journey map, I drafted preliminary ideas for a new feature, one that gave users another touchpoint to browse. Given that most users prefer to watch Netflix on TV, I naturally designed for the tvOS :
Before diving in this further, I had to step back. My initial concern is that creating what essentially is a more fleshed out version of an onboarding process, would run the risk of stripping user agency and result in overpersonalization. This quiz-like feature allows users to input preferences based on 3 attributes :
- Time - movie or TV show?
- Mood - Lighthearted? Dramatic?
- Genres - based on mood
Having the results screen show genres related to the mood, giving users the ability to 'refresh' if it wasn't accurate enough, as well as similar recommendations to the main recommendation on screen, would only exacerbate the problem of choice.
While my initial focus was on the user and providing them with modes of interactivity, I realise that ultimately, a user who has been browsing for a while just wants to be recommended content without additional effort. Furthermore, tvOS has limited navigation functionalities that would prove this layout ineffective.
... what does Netflix need to provide users with more accurate recommendations?
This revaluation begged the questions, how might Netflix be able to improve its recommendation? What data is needed to provide users with more accurate recommendations based on their interests? How might Netflix provide an alternate touchpoint for better exploration?
Under these lockdown circumstances, time can feel both precious and endless. Netflix Rapidfire helps give users the time they spent endlessly browsing back by expediting the decision making process. By answering just a few rapid-fire questions, Netflix will be able to better understand the users' preferences, and recommend new content for users to explore.
If one of the biggest complaints about Netflix is the inaccuracy of its recommendations, especially when personalization is one of the pillars of the company, what might Netflix need to strengthen this? By involving user feedback, it will give users the sense that Netflix is clearly using their data in service of them, and provide an extra level of personalization and a new avenue for exploration.
Testing was imperative as designing for TV proved challenging for someone who is more experienced with desktop and mobile interfaces. Split testing was done for each aspect of the design.
While the feature was intended to minimize choice by asking only two questions : 1) What do you have the time for? 2) What are you in the mood for?, the lack of questions actually proved to be dissatisfactory when I tested the first iteration on three users.
"If this was meant to give me more tailored content, then the quiz felt too short and vague to be satisfying"
What happens when users pick "no preference" for TV or movie, and "surprise me" for mood? More questions need to be asked for Netflix to give the user relevant recommendations.
Going back to the 'how might we' statements, the most important information that Netflix could benefit from is what the user likes and dislikes. With some users choosing to ignore the thumbs up/down rating function altogether, enforcing a gamified rating feature would provide Netflix instant and relevant user feedback.
Upon completing the short quiz, users would be presented with a single list of personalized content. I originally wanted to switch up the UI to emulate a TV guide, where users can browse through the list not dissimilar to channel surfing on live TV. I came up with two versions and did split testing to see which version users preferred:
Three of the four participants said they liked having a playlist of options to browse through. A user mentioned that they didn't know what to do with the small arrow on the right when they first saw it, while another user mentioned they didn't like the uncertainty of not know how far right they can go and how many recommendations Netflix would provide.
"I immediately knew what to do with the interface showing the thriller movies because it's so familiar"
The layout had to mimic what users are already familiar with :
A feature rollout for a company at the scale of Netflix is not a simple process. In considering how to measure the success (or failure) of this feature, I integrated a simple call-to-action to rate the feature at the end of the flow, after the user has finished watching. Not only would this provide feedback on Netflix Rapidfire, but it would continue to optimize Netflix's recommendations for the user.
A critical part in this process was taking a step back and referring back to the key problem: inaccurate recommendations. It meant that when Netflix couldn't optimize to make the browsing experience personalized, it may benefit from and occasionally require the users' help.
As an avid Netflix user, it was easy to fall into the trap of designing something based on my own experience. But one of the most important lessons this project has taught me is to test, test, and test some more, and continually iterate on the findings.
Next steps: If I had the opportunity to take this further, I would want to dig into the data to get insights into Netflix's browsing experience. Recommendation systems are complex infrastructures to design for, and there's a reason why personalization still takes so much effort for a company as sizeable as Netflix. This feature can also be continually developed, as most features do - whether that's being able to save a previous 'Rapidfire' search, or include the ability to alter preferences in profile settings.
.png)
