When you open your favorite app—whether for videos, music, shopping—you don’t see a single endless list of movies. Instead, you’re greeted by a page full of carousels — “Top Picks for You,” “Because You Watched…,” “Award Winners,” “Family Comedies.” This layout is not just prettier — it’s a completely different recommendation problem. Yet, much of the academic research out there still focuses on a single ranked list—a single column of suggestions.
Our recent research, “Full-Page Recommender: A Modular Framework for Multi-Carousel Recommendations,” looks at how to generate these complex, multi-row pages in a way that feels coherent, personal, and engaging. In other words, the challenge is figuring out how to go from one ranked list of “Top Picks for You” to a fully personalized page where each carousel offers contextual information tailored to the viewer.
Top Picks for You
Comedies
Recent Releases
Why Carousels?
1. Reflecting Real User Behavior
When platforms use carousels, users naturally scan pages in an F-shaped pattern—glancing at the top carousels before scrolling down or swiping horizontally. By flattening everything into one long list, traditional evaluation methods miss the way people actually browse.
2. Balancing Multiple Goals
A strong carousel page has to juggle different objectives:
- Relevance: Surfacing what a user is most likely to enjoy.
- Diversity: Keeping the page fresh, so it doesn’t feel repetitive.
- Coherence: Ensuring each row sticks to a theme—think “Romantic Comedies” or “Nature Documentaries.”
- Deduplication: Avoiding the same item appearing in multiple rows.
Pushing relevance too hard can hurt diversity; focusing only on diversity might dilute relevance. Our framework seeks a balanced, principled approach.
Our Three-Part, Modular Approach
We break the problem into three pieces, each manageable on its own:
1. Item Ranking.
We start with a standard recommendation model—matrix factorization, autoencoders, or transformers—that assigns each item a “relevance score” for the user.
2. Collection Generation.
Next, we group items into coherent “collections.” These could be broad genres (Comedy), niche themes (Heist Movies), or editorial picks. In our experiments, we tried:
- Metadata-based grouping (using tags like “comedy” or “historical”)
- Clustering in an embedding space, so items with similar user-behavior patterns get grouped
- Editorially curated lists (e.g., “Editor’s Picks”)
Because items can belong to multiple collections—just like a movie can be both Comedy and Romance—we preserve flexibility.
3. Carousel Layout Optimization.
Finally, we assemble the collections into a vertical list of carousels using a greedy, sequential algorithm. At each step:
- We calculate a combined score for each potential carousel based on item relevance and “freshness” (to avoid duplicates).
- We pick the top-scoring carousel for the next row, then update item freshness (“temperature”) before moving on.
This greedy method approximates a global optimum well enough and runs quickly in real-time.
Why This Matters
Shifting from single lists to full-page carousels isn’t just cosmetic. This approach personalizes not just the items but the context in which they appear. It encourages exploration by offering multiple entry points into the catalog, while still letting platforms balance business goals — whether that’s promoting new releases, highlighting award winners, or reviving older titles.
Although we tested on movie data, the same principles work elsewhere:
- Music apps grouping tracks into Chill Beats, New Releases, or Throwback Hits.
- E-commerce sites mixing Top in Electronics with Customers Also Bought and Seasonal Picks.
- News platforms balancing Breaking News with In-Depth Analysis and Opinion.
Looking Forward
In the paper, we suggest several next steps:
- Joint Optimization: Planning multiple carousels together instead of one-by-one to better approximate the global optimum.
- Automatic Labeling: Leveraging language models to craft catchy, accurate carousel titles.
- Live A/B Testing: Putting these ideas to the test on real platforms.
As multi-carousel layouts become even more prevalent—from video streaming to online shopping—we hope our Full-Page Recommender framework provides a solid foundation for the next wave of discovery engines.
Happy scrolling...