01 · Self-initiated · 2026
Breadcrumbs.
Designed and shipped a food discovery platform built around how people actually share great food.
Client
Self-initiated
Role
Product thinking · IA · UI · Prototype · Ship
Year
2026
Focus
The Observation
People screenshot restaurants on Instagram, save them to Notes, forget about them, and end up Googling "best ramen near me" anyway. Discovery, saving, and going never connect.
The Problem
Food discovery is fragmented across platforms that weren't built for it. The most valuable signal — a personal recommendation tied to a specific dish — has nowhere to live.
The arc
- 01
Thinking
Behaviour before design.
- 02
Doing
Structure before UI.
- 03
Shipping
Live, learning, iterating.
Stage 01
Thinking.
Behaviour before design.
I kept noticing the same thing — people screenshot restaurants on Instagram, save them to Notes, forget about them, and end up Googling "best ramen near me" anyway. The discovery happens in one place, the saving happens in another, and the actual going never happens at all.
The real problem wasn't that people didn't know about great food. It was that the way they captured and shared recommendations was completely broken. There was no structured place for the most valuable thing — the personal recommendation from someone whose taste you actually trust.
The most trusted recommendations happen in conversations. A friend says "get the wagyu ramen, not the tonkotsu." These micro-recommendations are incredibly high-signal — but they evaporate.
That became the core product concept: a trail. A short, dish-specific recommendation tied to a restaurant — like leaving a note for the next person.
Stage 02
Doing.
Structure before UI.
The biggest design decision wasn't visual. It was structural. I needed to define what a trail actually contained, and how trails connected to each other, before designing a single screen.
- →What's the minimum a trail needs to be useful? Restaurant + dish + short note + photo.
- →How do trails connect to restaurants? Many trails per restaurant, surfaced on restaurant pages.
- →How do users discover trails? Recent, trending, cuisine pages, people they follow.
- →How do trails scale? Trail Collections — curated guides like "Melbourne Ramen Trail."
This three-layer model — Trail → Restaurant → Collection — became the backbone of the entire product. Every screen decision that followed was built on it.
Decision 01 — Trails over reviews
Reviews are long, effort-heavy, and hard to scan. I constrained the trail format to make leaving one feel effortless — the friction of sharing had to be lower than the friction of not sharing.
Decision 02 — Dish-level specificity
Most platforms recommend restaurants. Breadcrumbs recommends dishes. "Get the wagyu ramen at Mensho" is more useful than "go to Mensho" — and actionable at the moment of decision.
Decision 03 — Community before algorithm
I designed for human social signals first — following people whose taste you trust. The launch strategy seeded content through a small Melbourne community before opening publicly.
Decision 04 — Suggest a restaurant
Users wanted to add places not in the database. A lightweight contribution flow — name, photo, food photo, receipt as proof — kept quality high without bureaucracy.
Stage 03
Shipping.
Live, learning, iterating.
Breadcrumbs launched as a live web product — not a prototype, not a pitch deck.
- →Trail creation and browsing
- →Restaurant pages with aggregated trails and dish recommendations
- →Trail Collections for curated food guides
- →User profiles and following
- →Suggest-a-restaurant flow
- →In-product feedback collection
- →Founder blog modelling the behaviour we wanted users to adopt
I ran structured feedback sessions designed to surface both usability issues and mental model gaps: "What do you think this product is for?" "Was it easy to find something you'd want to try?" "Did you try leaving a trail? Why or why not?"
- →The trail metaphor landed clearly — users immediately understood it.
- →Browsing by cuisine and trending trails worked well.
- →Trail creation had friction — users wanted clearer guidance on the short note.
- →Dish-level detail was the most appreciated feature — the thing that made it feel different from Google Maps or Yelp.
I iterated: improved onboarding copy, added placeholder examples in the trail form to reduce blank-page anxiety, and prioritised the suggest-a-restaurant flow after repeated requests.
What's next: mobile responsiveness for the context where food decisions actually happen, recruiting Melbourne contributors, IG presence to model the behaviour — and further out, a Restaurant Analytics Dashboard to monetise without compromising user trust.
Outcome
Shipped a live web product with trails, restaurant pages, collections, profiles, and a contribution flow — now iterating on mobile and community growth.
Reflection
Breadcrumbs taught me that the constraints you set for yourself reveal your real design instincts. Without a brief, a PM, or a deadline, every decision is yours — terrifying and clarifying. The most important thing I did was spend time understanding the behaviour before designing anything. The IA came from observation, not convention. And the product is better for it.