Meli Play — Case Study

01 Signal

Usage dropped 4% in 30 days

Analytics showed a measurable decline against the previous quarter — and pointed to where the leak was, never why. On a content platform, recurrence compounds. We opened research before opening Figma.

02 Insight

People kept WhatsApp chats with themselves

Four interviews, ages 21–60, converged on the same workaround: users were leaving Meli Play to ask friends what to watch, and saving titles in side conversations to remember later.

"Tenho uma conversa de WhatsApp comigo mesma pra anotar todos os filmes que o pessoal do trabalho me recomenda."
Carla · 21
"A tecnologia não é minha praia. Gostaria de ter um lugar com os conteúdos destacados pela minha família."
Roberto · 60

The catalog wasn't the problem. The social layer around it was missing.

03 Gap

Every player recommends to the user. None lets the user's people recommend.

A sweep of the four incumbents — each owns a distinct wedge, all miss the same layer.

NETFLIX

Algorithmic recs based on viewing history.

Disney+

IP-led, tight integration with Disney's ecosystem.

Prime Video

"Watch Party" for synchronous remote viewing.

Max

Same-day theatrical releases on select titles.

04 Solution

Build what the catalog was missing

Five features around a single idea: the user's people become the strongest recommendation engine the app has. Every feature anchors to a direct quote from research.

Favoritos tab. Minha Lista + titles you liked — reachable in 2 taps from anywhere.
Assistir Novamente. A rail inside Para Você that surfaces finished content automatically.
Recomendados pelos Amigos. A feed populated by what friends share inside the app — kills the WhatsApp workaround.
Native sharing. Primary action on every title screen, so the social flywheel actually starts.
Reactions (Gostei / Amei / Não é pra mim). Explicit feedback that corrects the recommender over time.

05 Impact

What we projected

Reversal +4% Recovers the measured monthly drop that kicked off the project.

Engagement +22% Weekly active sessions driven by the social feed loop.

Retention +15% 30-day retention among users interacting with friend recs.

06 Takeaways

What I carry forward

A 4% drop is a research prompt, not a UI brief. Starting with data kept the team honest.
The social layer is cheap to build, expensive to fake. People notice when a rec comes from a friend vs. an algorithm.
When every incumbent misses the same thing, it's not a gap — it's a product.

Process

4 interviews (21–60) · CSD matrix · benchmark · hi-fi in Figma · usability testing.

Meli Play — Home with personalized rails