Personalization Strategy
As a founding member of the Data-Driven Experiences team, I worked with our design director to built out the design strategy for machine learning-based features within the HBO platform. This vision grew into a product-wide strategy meant to balance our product’s human-touch and algorithmically generated content experience.
This resulted in the establishment of our product organization’s personalization and recommender system strategy, design process, and a roadmap for HBO Max to becoming a truly personalized ecosystem.
This project is the foundation for all personalization features built at HBO Max.
Context
HBO Max started with no personalization. This created a lack of trust in our platform and increased the gap between us and our competitors.
HBO Max had no personalization | With no algorithmically generated recommendation systems we were unable to come close to competitor parity or meet user’s expectations.
All pages were manually curated | And picking individual content for millions of users took time and energy from our editorial teams and cost the company millions in paid hours.
No overarching personalization vision | The organization had multiple personalization projects and initiatives with no overarching vision or collaborative design process.
No single content management system | With no real backend to feed our recommender systems we had disjointed and bloated tools with no connective tissue or defined logic on curating content or organizing it.
Research
Working with our research team we conducted a survey of 3,167 customers and ran in depth interviews with 16 users measuring attitudes toward homepage personalization.
Perceptions are shaped by the home screen experience | User perceptions over personalization are anchored in their home screen experience and are shaped by four key aspects of the experience: personalized recommendations, content organization, interfaces, and the ability to customize the experience and influence recommendations.
Perceptions of personalization influences repeat app engagement | When asked about the impact of feeling that an experience is designed with them in mind, participants across all groups generally expressed that an app or service feeling “designed with me in mind” tended to make them want to use more of the app.
Personalization can backfire | When reviewing competitors in this space we consistently heard that users are frustrated when platforms “pigeon hole” them into certain categories. Users can often feel trapped into what an algorithm think they like with no exit strategy.
Talking to more and more users we really began to understand that it’s not all about surfacing quality content but when and where? How can we reach people at the right time, in the right moment and not forget about the user’s complex and ever changing context?
Cross-functional Workshops
Global buy in was critical for this project. In order to achieve an agreed upon path I included members of the editorial team, engineering, product and design in large scale collaborative workshops focused on reviewing our personalization strategy and asking the question;
The cross functional workshops were hugely successful in gaining buy in and developing an understanding that we don’t want to lose our human-curated approach but how might we improve that experience? We were able to push forward with the agreement of moving forward and a trust in our team as leaders in this space.
Design Tenants
The cross functional workshops were hugely successful in gaining buy in and developing an understanding that we don’t want to lose our human-curated approach but how might we improve that experience?
We mean to step out of the bias against algorithmically generated content as inhumane, dangerous and untrustworthy into a tool we can develop using our own guidelines and imbedding our own opinions. Using ML as a tool and not as a crutch is the core of the evolving discipline of Human Centered Machine Learning.
“Models are opinions embedded in mathematics”
What Opinions Do We Want to Embed in Our Models?
How do we actually get there?
Keeping our core tenants in mind the generative experience strategy is includes a series of steps to allow for a growing personalization over time with the ability to change and develop with our users instead of pigeon holing them into what “we think” they are.
Design Principles
What Would That Look Like?
A dynamic relationship that builds over time evolving as a customer’s taste evolves.
Starting with Homepage Curation
The Results
This framework of personalization was presented to and signed off on by the product executive team. All personalization and recommender system projects at HBO Max and Max have been developed using this framework.