How Netflix built a Generative AI powered homepage that boosted engagement
And other recent examples of AI powered product personalization. Case studies from Spotify, DoorDash, Instacart, LinkedIn and more.
🔒DoP Deep goes deeper into the concepts and ideas that are covered in the Weekly Briefing to help you learn lessons from the experiences of top tech companies. If you’d like to upgrade to receive these in-depth pieces of analysis you can upgrade below. New reports are added every month.
The homepage is the first thing users see when they visit your product. And Netflix recently revealed how it plans to fundamentally change theirs using LLMs to boost engagement.
They call this “GenPage” and it’s a glimpse into the future of how all product homepages and personalization features could work.
LLMs have proven their ability to generate responses to prompts and so the teams at Netflix decided to take the principles of how LLMs work and build a deeply personalized homepage from scratch that answers this question:
Given everything we know about this user and this request, what homepage should we generate to maximize user satisfaction?
In this Deep Dive, we’ll take a closer look at what exactly Netflix built, the underlying technologies behind it and how you might be able to use these principles and technologies to add deeply personalized features to your own product.
As well as this, we’ll explore 15+ similar examples of generative AI powered features that drive engagement from other companies including Spotify, DoorDash, Uber, LinkedIn, Instacart and others that you can use as case studies for your own product’s personalization and engagement efforts.
Coming up:
What Netflix built and how it works
The underlying technologies behind it explained simply
Opportunities for introducing generative AI personalization in your product
The full selection of 15+ AI personalization examples and case studies from Airbnb, Apple, LinkedIn, Slack and more
What Netflix built and how you can apply similar AI personalization to your product
GenPage is an experimental new technology that generates a personalized version of the homepage using a single model:
Before GenPage, Netflix built the homepage using a a multi-stage pipeline. This meant different parts of the homepage were built separately. So, for example, the homepage was split into rows and different systems would generate different rows but they wouldn’t know about each other.
You can think of it almost like an assembly line with separate workers, none of whom talk to each other:
Worker 1 decides which rows should exist on your homepage (should we show “Continue Watching”? “Korean TV Shows”? “Because You Watched X”?)
Worker 2 picks which specific titles go inside each of those rows
Worker 3 decides the final order everything appears on the page
The net effect of this is that you could end up with a homepage where each individual row looks fine in isolation but the overall UX is poor. This is something that product teams have traditionally struggled with in recent times - particularly since the shift to component-based design systems where it can sometimes feel as though different parts of an interface are built by entirely different teams (or in this case, models).
How GenPage works instead
Instead of separate workers each handling a piece of the page, GenPage uses one model that builds the whole homepage in a single pass, the same way an LLM builds a response one word at a time.