Netflix

An update of the streaming giant's recommendation system.

Timeline

March 2022 - June 2022

My Role

Lead UX/UI Designer — Research, Testing, Visual Design, Interaction Design, Prototyping

This project was the concluding work for my master's degree. The module gave me the freedom to formulate and fulfil a creative brief based of my choosing.

Overview

​The project was first conceptualised when using Netflix in my university house with friends. We had been scrolling through the platform for what seemed an age, attempting to find a suitable title to watch. Eventually, we gave up and switched to a competitor platform which instigated a thought process of mine: "why isn't there a more effective way of finding desired content on Netflix besides relying on its 'smart' algorithms and AI?"

Following this thought, I decided to carry out further primary research to gauge how prevalent this issue was amongst other users. As it turned out, many users encounter the same dilemma on a day-to-day basis.

Goals

To assist Netflix in retaining their position as the top and most innovative smart TV streaming service by:

Updating the functionality and interface by enhancing the search and recommendation systems.

Increasing viewing engagement, not scrolling engagement.

Fulfilling Netflix's slogan "see what's next" on a deeper level.

Skills Developed

Adobe XD

Photoshop

Wireframing

TV UI Design

Prototyping

Usability testing

Research, Flows and Journey Maps

THE BRIEF

Users were being held back.

The Challenges

Frequent on-demand streaming service users are familiar with the endless scrolling experience.

Not finding the right show/film in good time created frustration and a negatively associated experience.

Netflix had scarcely updated its recommendation features since 2005, it was time for a new and more effective system.

The Purpose

To create new features for Netflix to alleviate these challenges and create increased viewing engagement.

To help Netflix ward off competitors and retain their position as the top and most innovative on-demand streaming service.

The Audience

All Netflix users, particularly the entertainment fanatics amongst them.

Frequent streaming consumers.

Medium to high digital competence.

Non-Netflix users were also targeted.

The Important Stuff

The new features were designed to the format of in-home TV viewing.

The new features align with Netflix’s brand identity, style and tone of voice, creating seamless integration into the current system and making way for easy understanding and use.

The final executions have been presented through a fully functioning prototype using Adobe XD.

CATEGORY RESEARCH

Understanding the industry.

The Video On-Demand Streaming Industry

Video on-demand streaming is a form of digital TV and movie entertainment consumption. On-demand streaming holds advantages over regular scheduled TV such as the abilities to pause, play, fast-forward, rewind and re-watch at each user's convenience.

2020 Video On-Demand Market Shares

Industry Trends

Live streaming

Faster internet speeds, more capable media devices and the want for ad free content have driven the demand for live streaming.

User subscription figures anticipated to grow exponentially

The convenience benefits streaming offers over terrestrial TV is beginning to appeal to older generations, converting them to the industry

Increased user personalisation

Users are seeking further interface personalisation to make their streaming experiences more accurate and pleasing.

The rise of new players

The market leader, Netflix, lost 31% of its market share to new players in the market (2019). This demonstrated the increased competition within the industry and the need for innovative strategy and development.

BRAND RESEARCH

Getting to know Netflix.

Strategies

Due to the entertainment providing nature of Netflix, the service implement a customer-centric strategy that drives strategic decisions. Here, enhancement of the user's experience lies at the heart of the brand.

Netflix also implement a growth strategy, with the aim of expanding and growing into different markets through its services. For example, in 2013, Netflix expanded into content production.v

Furthermore, with the largest market share, Netflix also implement a market leader strategy. With this position comes competition, therefore, Netflix must be wary of brand value.  Value can be protected through consistent and competitive pricing and performance in terms of operation and content offered.

Semiotic Analysis

Logo

Signifier: Black and standout red

Signifies: A striking combination that conveys power, strength, passion, expertise and a premium cinematic feel

Colours

Signifier: Various shades of black for backgrounds, white text, colourful feature artwork

Signifies: Excitement, entertainment, authenticity, engagement

Typeface

Signifier: Gotham Bold and Gotham Book
Signifies: Clarity, boldness, functionality, trendiness, freedom

Key Brand Information

RECOMMENDATION SYSTEMS

Decoding the algorithms.

About Recommendation Systems

In relation to a video on-demand streaming service, a recommendation system is put in place in an attempt to personalise each user experience for maximal function and satisfaction. By utilising artificial intelligence (AI) and machine learning algorithms, the main aim of a recommendation system is to deal with information overload by clearly providing predicted content for each user's convenience.


All online platforms that possess an intention to push products to consumers utilise recommendation systems; the systems are put in place to guide consumers further down the consumption journey to the point they repurchase or reuse a service.

Netflix utilises AI and machine learning from user data to construct their content recommendations.


The data collected is from user behaviour figures such as favourite movies/shows, genres, actors, plots and ratings, this is known as matrix factorisation, a class of collaborative filtering. The system uses the data to find a relationship between the contents' and the users' entities, making a prediction based on associations.


Additionally, collaborative filtering is utilised further by basing recommendations on other similar users' viewing figures.


These data sets are then used to predict what each user may like to watch, resulting in a more personalised experience for each user.


However, user taste and preference is impossible to predict correctly 100% of the time due to their constantly evolving nature, so Netflix counters this by favouring their original content and the latest releases.

Netflix utilises AI and machine learning from user data to construct their content recommendations.


The data collected is from user behaviour figures such as favourite movies/shows, genres, actors, plots and ratings, this is known as matrix factorisation, a class of collaborative filtering. The system uses the data to find a relationship between the contents' and the users' entities, making a prediction based on associations.


Additionally, collaborative filtering is utilised further by basing recommendations on other similar users' viewing figures.


These data sets are then used to predict what each user may like to watch, resulting in a more personalised experience for each user.


However, user taste and preference is impossible to predict correctly 100% of the time due to their constantly evolving nature, so Netflix counters this by favouring their original content and the latest releases.

Similarly to Netflix, APV's recommendation system is based on AI and machine learning through collaborative filtering and matrix factorisation.


APV differentiates when it comes to their "safe bets" system, where popular and well-known content is recommended amongst the other filtering methods.


Unlike Netflix, APV states that it does not favour own produced content, although, they do favour the latest releases.

Similarly to Netflix, APV's recommendation system is based on AI and machine learning through collaborative filtering and matrix factorisation.


APV differentiates when it comes to their "safe bets" system, where popular and well-known content is recommended amongst the other filtering methods.


Unlike Netflix, APV states that it does not favour own produced content, although, they do favour the latest releases.

In the case of Disney+, AI and machine learning are also utilised through collaborative filtering and matrix factorisation. However, they differ from Netflix and APV through their focus on positive and negative individual preferences such as movies you prefer and movies you scroll past and ignore.

USER TESTING

Evaluating Netflix's current usability.

Recommendation Accuracy

To test the accuracy of the current recommendation systems that Netflix have in place, the number of features desired (out of 40 on each recommendation row) to be watched at some point in time was recorded. 8 participants from a variety of ages, identities and locations were used for the research.

The research concluded that on average, the recommendation system is not accurate enough for user needs, being less than 25% accurate in all key recommendation rows. It is imperative that this accuracy is increased. 

Recommendation Accuracy Testing Results

Time

The time taken for users to find desired content on streaming platforms plays a great role in user satisfaction; when the process becomes longer, more time is wasted and more frustration is caused, resulting in a negatively associated experience. With this in mind, to further test the accuracy of the current recommendation systems, the time taken for each user to find a desired title was recorded. Lower times (under 3 minutes) are positive and higher times (over 5 minutes) are negative. 

The results demonstrated that Netflix's current recommendation system not only causes its users to misspend their valuable time, but also creates risks of frustration, app switching and negative experiences.


A key objective of this campaign is to decrease these times.

Time Testing Results

CONCLUSIONS

Users are craving evolution.

Identification of Platform Issues

Through the primary and secondary research carried out, three main issues concerning Netflix's poor user user experience have been uncovered.

Key Platform Issues

Research and Testing Conclusions

The testing has concluded that for the large majority of users, the Netflix recommendation system is majorly flawed. The system fails to deliver accurate suggestions in a timely fashion, which if done correctly, holds the ability to satisfy user needs beyond expectations. As a result, users are becoming frustrated, irate and discouraged, which in turn is causing users to give up and/or consider switching to competitors.


On a personal note, from the research completed and through my own personal experience, I believe that no matter how "efficient" AI and machine learning are, when it comes to the unpredictability of preferences, personal taste and recommendations, nothing performs greater than human input and interaction.

Key Takeaway

These results created a sense of urgency for innovation and growth in relation to Netflix's recommendation system; by doing so, Netflix customers will become more greatly satisfied with their all-round streaming experience on the platform, warding off negative associations, threats from competitors and subsequent losses in market share.

THE DESIGN PROCESS

Empathise, Define, Ideate, Deliver.

User Persona

User Journey Map

Ideation

Wireframes

FINAL EXECUTION

The future of Netflix is highly personalised.

Updated Landing Page

The updates to the platform begin at the landing page, where a total redesign can be seen. Upon the application launch, users are presented with three options: browse, search or, the first of the new features, advanced search. Clicking each will take the user to their designated screens.

Landing Page

Advanced Search

Updated Home Page

Further updates to the platform include a slight but noticeable new look for the home screen and title cards. Firstly, the recommendation row titles have been updated to offer more specific and tailored content to each user based on, for example, favourite actors, moods, time of day and decade ranges.


Secondly, friend recommended content is marked with yellow stars; this feature ties in with Netflix's new social features. This new feature allows users to effortlessly spot content that friends have recommended, acting as a back up to any unsuccessful content searches and ultimately improving the search and recommendation processes. A new row has also been created specifically for this feature.

Home Page

New Title Card Features

The features within title cards have also received an update. Here, users can now, if recommended by a friend, see who has recommended the item of content (so they know if the recommendation can be trusted).


Secondly, in addition to the current like and dislike components, users can recommend and share the title to their friends through the social feed, direct message or other social medias, as well as having the capacity to add the title to one of their albums or create a new one.

Title Card

Personal Profiles

A key update to the system is the new profile feature. Firstly, users can now create bespoke albums for their and their friends content finding needs. Secondly, integrated into each user's profile is a social feed, where users can view content that friends have shared, liked/disliked or recommended. This feature enables users to discover content based on the preferences of those closest to them, increasing the accuracy of results.


Each profile features a shortcut bar as seen on the example profile and at the top here. Here, users can access likes/dislikes, recommendations, friends and previous posts.

Profile

Likes and Dislikes

Recommendations

Friends

Top 10 Album

Netflix

An update of the streaming giant's recommendation system.

Timeline

March 2022 - June 2022

My Role

Lead UX/UI Designer — Research, Testing, Visual Design, Interaction Design, Prototyping

This project was the concluding work for my master's degree. The module gave me the freedom to formulate and fulfil a creative brief based of my choosing.

Overview

​The project was first conceptualised when using Netflix in my university house with friends. We had been scrolling through the platform for what seemed an age, attempting to find a suitable title to watch. Eventually, we gave up and switched to a competitor platform which instigated a thought process of mine: "why isn't there a more effective way of finding desired content on Netflix besides relying on its 'smart' algorithms and AI?"

Following this thought, I decided to carry out further primary research to gauge how prevalent this issue was amongst other users. As it turned out, many users encounter the same dilemma on a day-to-day basis.

Goals

To assist Netflix in retaining their position as the top and most innovative smart TV streaming service by:

Updating the functionality and interface by enhancing the search and recommendation systems.

Increasing viewing engagement, not scrolling engagement.

Fulfilling Netflix's slogan "see what's next" on a deeper level.

Skills Developed

Adobe XD

Photoshop

Wireframing

TV UI Design

Prototyping

Usability testing

Research, Flows and Journey Maps

THE BRIEF

Users were being held back.

The Challenges

Frequent on-demand streaming service users are familiar with the endless scrolling experience.

Not finding the right show/film in good time created frustration and a negatively associated experience.

Netflix had scarcely updated its recommendation features since 2005, it was time for a new and more effective system.

The Purpose

To create new features for Netflix to alleviate these challenges and create increased viewing engagement.

To help Netflix ward off competitors and retain their position as the top and most innovative on-demand streaming service.

The Audience

All Netflix users, particularly the entertainment fanatics amongst them.

Frequent streaming consumers.

Medium to high digital competence.

Non-Netflix users were also targeted.

The Important Stuff

The new features were designed to the format of in-home TV viewing.

The new features align with Netflix’s brand identity, style and tone of voice, creating seamless integration into the current system and making way for easy understanding and use.

The final executions have been presented through a fully functioning prototype using Adobe XD.

CATEGORY RESEARCH

Understanding the industry.

The Video On-Demand Streaming Industry

Video on-demand streaming is a form of digital TV and movie entertainment consumption. On-demand streaming holds advantages over regular scheduled TV such as the abilities to pause, play, fast-forward, rewind and re-watch at each user's convenience.

2020 Video On-Demand Market Shares

Industry Trends

Live streaming

Faster internet speeds, more capable media devices and the want for ad free content have driven the demand for live streaming.

User subscription figures anticipated to grow exponentially

The convenience benefits streaming offers over terrestrial TV is beginning to appeal to older generations, converting them to the industry

Increased user personalisation

Users are seeking further interface personalisation to make their streaming experiences more accurate and pleasing.

The rise of new players

The market leader, Netflix, lost 31% of its market share to new players in the market (2019). This demonstrated the increased competition within the industry and the need for innovative strategy and development.

BRAND RESEARCH

Getting to know Netflix.

Strategies

Due to the entertainment providing nature of Netflix, the service implement a customer-centric strategy that drives strategic decisions. Here, enhancement of the user's experience lies at the heart of the brand.

Netflix also implement a growth strategy, with the aim of expanding and growing into different markets through its services. For example, in 2013, Netflix expanded into content production.v

Furthermore, with the largest market share, Netflix also implement a market leader strategy. With this position comes competition, therefore, Netflix must be wary of brand value.  Value can be protected through consistent and competitive pricing and performance in terms of operation and content offered.

Semiotic Analysis

Logo

Signifier: Black and standout red

Signifies: A striking combination that conveys power, strength, passion, expertise and a premium cinematic feel

Colours

Signifier: Various shades of black for backgrounds, white text, colourful feature artwork

Signifies: Excitement, entertainment, authenticity, engagement

Typeface

Signifier: Gotham Bold and Gotham Book
Signifies: Clarity, boldness, functionality, trendiness, freedom

Key Brand Information

RECOMMENDATION SYSTEMS

Decoding the algorithms.

About Recommendation Systems

In relation to a video on-demand streaming service, a recommendation system is put in place in an attempt to personalise each user experience for maximal function and satisfaction. By utilising artificial intelligence (AI) and machine learning algorithms, the main aim of a recommendation system is to deal with information overload by clearly providing predicted content for each user's convenience.


All online platforms that possess an intention to push products to consumers utilise recommendation systems; the systems are put in place to guide consumers further down the consumption journey to the point they repurchase or reuse a service.

Netflix utilises AI and machine learning from user data to construct their content recommendations.


The data collected is from user behaviour figures such as favourite movies/shows, genres, actors, plots and ratings, this is known as matrix factorisation, a class of collaborative filtering. The system uses the data to find a relationship between the contents' and the users' entities, making a prediction based on associations.


Additionally, collaborative filtering is utilised further by basing recommendations on other similar users' viewing figures.


These data sets are then used to predict what each user may like to watch, resulting in a more personalised experience for each user.


However, user taste and preference is impossible to predict correctly 100% of the time due to their constantly evolving nature, so Netflix counters this by favouring their original content and the latest releases.

Similarly to Netflix, APV's recommendation system is based on AI and machine learning through collaborative filtering and matrix factorisation.


APV differentiates when it comes to their "safe bets" system, where popular and well-known content is recommended amongst the other filtering methods.


Unlike Netflix, APV states that it does not favour own produced content, although, they do favour the latest releases.

In the case of Disney+, AI and machine learning are also utilised through collaborative filtering and matrix factorisation. However, they differ from Netflix and APV through their focus on positive and negative individual preferences such as movies you prefer and movies you scroll past and ignore.

USER TESTING

Evaluating Netflix's current usability.

Recommendation Accuracy

To test the accuracy of the current recommendation systems that Netflix have in place, the number of features desired (out of 40 on each recommendation row) to be watched at some point in time was recorded. 8 participants from a variety of ages, identities and locations were used for the research.

The research concluded that on average, the recommendation system is not accurate enough for user needs, being less than 25% accurate in all key recommendation rows. It is imperative that this accuracy is increased. 

Recommendation Accuracy Testing Results

Time

The time taken for users to find desired content on streaming platforms plays a great role in user satisfaction; when the process becomes longer, more time is wasted and more frustration is caused, resulting in a negatively associated experience. With this in mind, to further test the accuracy of the current recommendation systems, the time taken for each user to find a desired title was recorded. Lower times (under 3 minutes) are positive and higher times (over 5 minutes) are negative. 

The results demonstrated that Netflix's current recommendation system not only causes its users to misspend their valuable time, but also creates risks of frustration, app switching and negative experiences.


A key objective of this campaign is to decrease these times.

Time Testing Results

CONCLUSIONS

Users are craving evolution.

Identification of Platform Issues

Through the primary and secondary research carried out, three main issues concerning Netflix's poor user user experience have been uncovered.

Key Platform Issues

Research and Testing Conclusions

The testing has concluded that for the large majority of users, the Netflix recommendation system is majorly flawed. The system fails to deliver accurate suggestions in a timely fashion, which if done correctly, holds the ability to satisfy user needs beyond expectations. As a result, users are becoming frustrated, irate and discouraged, which in turn is causing users to give up and/or consider switching to competitors.


On a personal note, from the research completed and through my own personal experience, I believe that no matter how "efficient" AI and machine learning are, when it comes to the unpredictability of preferences, personal taste and recommendations, nothing performs greater than human input and interaction.

Key Takeaway

These results created a sense of urgency for innovation and growth in relation to Netflix's recommendation system; by doing so, Netflix customers will become more greatly satisfied with their all-round streaming experience on the platform, warding off negative associations, threats from competitors and subsequent losses in market share.

THE DESIGN PROCESS

Empathise, Define, Ideate, Deliver.

User Persona

User Journey Map

Ideation

Wireframes

FINAL EXECUTION

The future of Netflix is highly personalised.

Updated Landing Page

The updates to the platform begin at the landing page, where a total redesign can be seen. Upon the application launch, users are presented with three options: browse, search or, the first of the new features, advanced search. Clicking each will take the user to their designated screens.

Landing Page

Advanced Search

Updated Home Page

Further updates to the platform include a slight but noticeable new look for the home screen and title cards. Firstly, the recommendation row titles have been updated to offer more specific and tailored content to each user based on, for example, favourite actors, moods, time of day and decade ranges.


Secondly, friend recommended content is marked with yellow stars; this feature ties in with Netflix's new social features. This new feature allows users to effortlessly spot content that friends have recommended, acting as a back up to any unsuccessful content searches and ultimately improving the search and recommendation processes. A new row has also been created specifically for this feature.

Home Page

New Title Card Features

The features within title cards have also received an update. Here, users can now, if recommended by a friend, see who has recommended the item of content (so they know if the recommendation can be trusted).


Secondly, in addition to the current like and dislike components, users can recommend and share the title to their friends through the social feed, direct message or other social medias, as well as having the capacity to add the title to one of their albums or create a new one.

Title Card

Personal Profiles

A key update to the system is the new profile feature. Firstly, users can now create bespoke albums for their and their friends content finding needs. Secondly, integrated into each user's profile is a social feed, where users can view content that friends have shared, liked/disliked or recommended. This feature enables users to discover content based on the preferences of those closest to them, increasing the accuracy of results.


Each profile features a shortcut bar as seen on the example profile and at the top here. Here, users can access likes/dislikes, recommendations, friends and previous posts.

Profile

Likes and Dislikes

Recommendations

Friends

Top 10 Album