Atly: from social maps to AI-powered discovery
I lead product design at Atly, where we turned a pre-revenue social mapping app into a commercial, AI-powered discovery product, proving the model with a single map community that far exceeded its revenue target.
The product
Atly (formerly Steps) launched publicly in 2023 with $18M in funding to build a community mapping platform, with around 120,000 members curating 6,500+ map communities about the places they love. The knowledge inside those maps was extraordinary. The business wasn't: the product was pre-revenue, and the experience leaned on generic mapping conventions that undersold what the community knew.
The problem
The core product problem: every mapping app rates places. But a 4.5-star pizzeria is a great answer to one person's search and a terrible answer to another's. Atly needed an experience and a business model built around what users actually mean when they search, and it needed to prove people would pay for it.
Role
I joined in 2022 as Design Lead, directing two designers and partnering with the founders, product management, and engineering. As Atly restructured into a small, AI-native team, I became the sole designer. Today I own product design end-to-end, from strategy with the founders to shipped screens.
Team of one
Working this close to the ground means every design decision I make is also a prioritization decision: research, flows, UI, prototyping, and QA all run through one pair of hands, with engineers as daily design partners.
Live product, lean team
A live community product: redesigns couldn't break the workflows of the members and map creators who built Atly's value. A lean team: every exploration had to earn its engineering cost. And a monetization mandate: prove revenue with design and data, not with a marketing budget.
Data economics
The deepest constraint was data economics: the granularity that makes a recommendation trustworthy was originally produced by manual curation, which doesn't scale. The design challenge and the AI challenge were the same challenge.
Signals
Decisions run on a continuous mix of signals: product analytics, user interviews, app-store reviews, and A/B tests. Failed experiments are treated as research; two of the most important findings below came from features we shipped, measured, and killed.
The insight
The defining insight: people don't search for a place, they search with an intent: “romantic pizzeria for a first date,” “café I can work from.” Star ratings answer the wrong question. That insight became the product's core mechanic.
Decisions & tradeoffs
Score the search, not the placeKey decision · shipped
Instead of showing one static rating per place, Atly analyzes reviews from multiple platforms and computes a match score against the user's search intent. The same pizzeria scores 9.2 for “romantic first date” and differently for “quick work lunch.” This is the opposite of what most mapping apps do, where ratings belong to the place no matter what you asked.
The tradeoff: an unfamiliar score that has to earn trust against the universally understood five stars. We answer it by showing the evidence, including the negatives (“moderate noise,” “unfriendly staff”), right under the score. And a heavy data pipeline instead of a cheap ratings field.
Tinder-style place cardsTested · killed
We prototyped swipe-card browsing for places. It failed in testing: choosing where to go is comparative and spatial. People weigh options against each other and against distance, not in a sequence of isolated yes/no calls. We dropped it.
Social-feed engagementTested · killed
We also tried a feed-style engagement model. It optimized for time-in-app, but Atly's users come with a decision to make, not time to fill. Killing the feed sharpened the product's conviction: discovery on Atly is intent-driven, and the interface should get users to a confident answer fast.
One fluid surface instead of deep menusKey decision · shipped
For a younger, mobile-first audience, I moved the interface away from deep menus and heavy page structures toward a single-surface experience: natural-language search, the map, and sliding result sheets in one continuous space. Lightweight and immediate, while keeping the depth of a professional mapping tool.
Proving the business: one map, monetized
The bet
To prove people would pay, we took Atly's strongest community, gluten-free and celiac diners, where a wrong recommendation has real consequences, and built it out at full depth: a safety classification system developed with dieticians and nutritionists, 270,000+ curated locations, and a dedicated data team. It launched as Atly's first subscription map in 2024.
The result, and what’s next
The bet worked: the map far exceeded its revenue goal and validated subscriptions as the model. The next step is the AI thesis: using large-scale analysis of place data to deliver that same level of granularity without the manual curation, for every category of search.
The screens below tell the story in sequence. Onboarding sets the positioning: your kind of places, powered by locals. The place page shows the core mechanic at work, scoring a pizzeria 9.2 out of 10 for the search “romantic pizzeria with cocktails for a first date” and laying out the evidence under the score, negatives included. On the map, every pin carries a match score for the current search instead of a static rating. Editorial collections distill thousands of local reviews into guides, natural-language search and intent chips keep discovery on one fluid surface, and the subscription flow closes the sequence with the paywall that turned the product commercial.
Outcomes
Atly went from pre-revenue exploration to a commercially active product. The first monetized map exceeded its revenue target and proved the subscription model; its learnings now drive the AI roadmap. Everything shown here is shipped and in continuous improvement.
Sources & disclosure
Public record: TechCrunch on Atly's $18M launch (2023) · Gluten-Free Eats launch (2024). Internal revenue figures are confidential and intentionally omitted.
What the killed features taught
The most valuable design work here was the work we threw away. The card browser and the social feed were cheap to test and decisive to kill. Each failure narrowed the product to what it actually is: an intent engine wearing a map.
The pattern I keep
The gluten-free map set the pattern I now design by: prove the value at full depth manually first, then teach the system to scale it. AI didn't change what good design is here; it changed what's affordable to build.