Streaming Personalization and Recommendations

,

Whole Page Optimization is a set of Machine Learning driven mechanisms that work in concert to determine what content a customer is most likely to engage with, then generate a page, optimizing the content on screen based on that probability.

The goal of Whole Page Optimization (WPO) is to be able to generate a content discovery experience in which the content presented is tailored to the tastes of the viewer so that subscribers will find something to watch each time they log in.

  • Max WPO example 6
  • Max WPO example 5
  • Max WPO example 4
  • Max WPO example 3
  • Max WPO example 2
  • Max WPO example 1

The Assignment

The challenge that every streaming platform faces is “how do we present relevant content to our customers while also presenting content that the business has invested in”.

Users and Audience

Whole Page Optimization is designed to tailor the home page experience for every profile that is authorized the platform to use personal data signals- in this case most users globally with sensitivity for kids and locations in which there are regulations around using personal data.

Process

My team is responsible for collaborating with design research, outlining design strategy, representing the customer’s POV in highly technical conversations, articulating how the mechanisms work to generate the desired outcome and checking the quality of the output in UAT.

Starting with a lit review, we sifted through a variety of existing research to understand user’s mindsets and what we understood about how people browse for content on our streaming platforms. What we learned was that people had 5 major mindsets that need to be considered when constructing the page. Additionally we learned that people needed a break from a sea title grid work – in other words some variety and visual tempo reduces cognitive load as people move vertically down the page.

Wireframes were produced in collaboration with content teams in order to quickly communicate our ideas for how the page might be constructed and to gain alignment across teams and with stakeholders. We served as advisors during visual design sprints where we collaborated with A11Y, component design and design systems teams.

In order to power the components and similarly the vertical construction of the page, we collaborated with data science and ML engineering for each component to shape how each works- Recommend For You, Top 10, and Because You Watch ass examples. In order to communicate how the mechanisms work, we worked with the motion team to compose a narrative, storyboard and animate the stories, resulting in widespread understanding of how the algos work to create a personalized experience for each person.

We launched the features in the US, continue to iterate on algorithms, expanding to global support with additional launches in LATAM and EMEA through A/B testing and continued UX research.

Outcomes and Lessons

  • Drove 2% increase in engagement
  • 7% reduction in high-risk churn
  • Launched in 75+ Countries on 25+ platforms
  • Introduced motion-based storytelling for technical solutions
  • Persisted through 4 product managers
  • Gained a deep understanding of the value of different user signals can be used to generate desired outcomes.