view
back

Sello.

Curated restaurant discovery for São Paulo — editorial selection, a community that keeps it honest, and a score that explains where it came from.

TypeOwn product · Live on iOS & Android
RoleProduct design, end-to-end
Year2026
StackFigma, React Native, Supabase
01 Context

Choosing a restaurant shouldn't come down to luck

Every platform is confident, and they all disagree. Instagram calls it the best in the city. TikTok has something more unmissable. On Maps everything is a 4.8. The group chat says "didn't love it, try another."

You end the search with more doubt than you started with. Sello is built on the opposite premise: fewer places, checked, with the reasoning visible.

02 Problem

Why ratings stopped meaning anything

When every restaurant scores 4.8, the score has stopped carrying information.

  • Compressed scales. Public ratings cluster so tightly at the top that they no longer separate good from great.
  • No provenance. A number arrives with no way to see who said it, or when.
  • Volume over judgement. Reach rewards whatever is loudest this week, not what holds up.
  • Nothing decays. A place coasts on a reputation it earned three years ago.
03 System

Four ingredients, one score

The Sello Score is deliberately not a single source. It combines four, so that no one of them can be gamed alone.

01

Curation

Not every restaurant gets in. The team researches, visits and selects — fewer recommendations, but the right ones.

02

Community

The people who actually went keep it current, showing how a place delivers in practice.

03

Recognition

Reviews, awards and industry lists — the trail a good restaurant leaves behind.

04

Sello intelligence

The layer that reads all of the above and turns scattered signals into one number that can be explained.

04 Product

Four things you can do

Each one feeds the next: discovery produces opinions, and opinions improve discovery.

01
Explore
Map, themed guides, and recommendations from people whose taste matches yours.
02
Rate — the veredito
One decisive score instead of a form. Rating a place earns points, which is what keeps the catalogue fresh.
03
Share
Lists and guides built with friends, sent as a link rather than retyped into a chat.
04
Earn
You score when the community saves and uses what you recommended — the incentive rewards being useful, not being loud.
05 Design System

Mesa DS — the system underneath

A product this size does not hold together on screens alone. Sello runs on its own design system, Mesa DS, aligned to Apple's Human Interface Guidelines and carried from Figma into production code as a single source of truth.

Token architecture

Seven token families — colour, typography, spacing, shape, elevation, motion and state — rather than one flat palette. Every colour is defined for light and dark at the same time, so a theme is a property of the token, not a second set of screens to maintain.

  • Semantic over literal. Tokens are named for their job, not their value: surface, surface-2, fill, separator, label-2, label-3. Changing a hue never means renaming a component.
  • One definition, three consumers. Tokens resolve to CSS variables, Tailwind classes and a useThemeColors() hook, so design and code cannot drift apart without the drift being visible.
  • Constraints encoded, not remembered. Minimum tap target lives in the system as a value components read, instead of a rule someone has to recall in review.
The system documents itself

Mesa DS ships with its own documentation surface inside the app — colour swatches, type specimens, spec tables and live example frames. Documentation that sits next to the code is documentation that survives the next release; a spec deck in a folder does not.

Components169Shipped components, including 10 skeleton states.
Colour tokens141Each resolved for light and dark.
Token families7Colour, type, spacing, shape, elevation, motion, state.

The system is what let one designer ship and keep shipping a consumer product on two platforms. It is also where most of the design decisions actually live — the screens are just where you can see them.

06 Challenges

Trade-offs

  • Coverage × trust. Chose the smaller catalogue. A curated app that lists everything is just a directory with better typography.
  • Discovery × booking. Sello deliberately does not take reservations. Owning the decision is a different product from owning the transaction.
  • One city × many. São Paulo first. Curation does not scale by copy-paste — it scales when a city has enough people to keep it honest.
  • Cold start. Editorial curation carries the experience before there is enough community signal to carry it.
07 Takeaways

What I carry forward

  • A score is worth what its provenance is worth. Explaining where it came from beats one more decimal place.
  • Curation is a product decision with a cost. Saying no to restaurants is what makes the yes mean something.
  • Rewarding usefulness instead of volume changes who shows up. Incentives are interface.
  • Shipping to two app stores is a design constraint, not a handover. The work does not end at the prototype.
08 Status

Where it stands

Live on the App Store and Google Play, launched in São Paulo.

Community+1,000Food explorers on the platform.
Catalogue+500Restaurants researched and selected.
Guides+50Themed guides, built by occasion.
Home - categories, the week's pick and the guides the community saves most
Community - the feed of vereditos, and the monthly ranking it feeds
A collaborative guide - places, who has been, and what they said
Profile - been there, want to go, and the lists you follow
Next project Dwello →