Skip to content
01Case studySW · 01 of 10

An AR property-viewing app for buyers who can't be in the room.

A Swiss-DACH agency was selling unbuilt apartments to remote buyers, and floor plans plus 360° photos weren't closing deals. We built iOS and Android apps that drop a full-scale apartment into any room. Three years, three markets, 28,000 viewings.

ClientConfidential
Year2022 — 2024
Duration3 yrs
StackSwift · Kotlin · RealityKit · ARCore · Node · Postgres · AWS
Hero image for apartment-arFIG 01 · HERO

Selling space you can't yet walk through.

Apartment AR's clients were Zurich- and Vienna-based developers selling pre-construction units to international buyers. The buyer pool was overwhelmingly remote — Saudi families, London relocations, Hong Kong investors — who couldn't fly in for a 40-minute viewing of an unbuilt apartment.

The agency's salespeople had renders, 360° photos, and CAD floor plans. None of it answered the buyer's real question: how big does this actually feel? Sales cycles dragged past nine months, with multiple in-person trips, and roughly one in three buyers walked away after the first physical viewing because the unit felt smaller than the renders looked.

The brief, from the agency's CEO, on day one: I want a tool my salesperson can put in a buyer's hand from anywhere in the world, and the buyer should know in five minutes whether this apartment is the right size for them.

A true-to-scale AR apartment, dropped into any room.

We shipped paired iOS and Android apps that let an agent and a buyer enter a co-viewing session, with the apartment placed in true 1:1 scale in whatever physical space the buyer was standing in — a hotel room, a living room, an empty office. The agent steered a tablet view of the same scene from Zurich, with annotations and material swaps the buyer could see in real time.

The hard problem was never the AR. ARKit and ARCore had matured enough by 2022 that the placement was solvable. The hard problem was the session: keeping two devices on opposite sides of the world in sync, low-latency, with the agent able to point at a kitchen counter and have the buyer's screen show exactly that surface highlighted, all over patchy hotel Wi-Fi.

Six core capabilities shipped in the first year: AR placement at full scale, two-way co-viewing with annotations, runtime material swaps, floor stitching between rooms, pre-staged furniture sets, and an in-call quote builder. Each one survived its first contact with a real salesperson; that was the bar.

F · 01AR placement
Drop a full-scale apartment into any room with one tap. ARKit on iOS, ARCore on Android, shared scene graph in between.
F · 02Co-viewing
Agent and buyer in the same session, with cursor, annotations, and voice routed through a single low-latency channel.
F · 03Material swaps
Walls, floors, finishes, and lighting swap at runtime — the agent can show a buyer four kitchen variants in 90 seconds.
F · 04Floor stitching
Move between rooms without re-anchoring. The buyer walks through the apartment by walking around their hotel room.
F · 05Furniture sets
Pre-staged furnishing packages let the buyer see the unit as a family home, a pied-à-terre, or a short-let, without leaving the call.
F · 06Quote builder
Generate and email a fitted price quote with the buyer's material choices baked in, before the call ends.

Friday demos for three years.

Production from sprint two. The first version was deliberately ugly and deliberately complete — one apartment, one agent, one demo buyer on the founder's iPad. We took it to the agency's sales floor on a Friday in week three and watched a salesperson use it badly. We rewrote the agent UI the following Monday.

From there it ran on a six-week cadence: ship to TestFlight, watch six real viewings, fix the three biggest things that hurt the sale. The agency stopped doing internal training sessions on the app after month four because the product was simpler to use than to explain.

We expanded into the broader DACH market in late 2022, then the UAE in 2023, then the UK in 2024. Each market added local quirks — different unit-of-measurement defaults, different lighting expectations for premium, different opinions on what a kitchen should look like — and each addition shipped behind a feature flag for two weeks before going live.

From one agency to a regional standard.

Sales cycle compressed from nine months to under four. The agency stopped flying buyers in for the first viewing — that step happened in the buyer's living room, on the app — and reserved physical visits for serious buyers only. Three of the four developer clients added explicit Apartment AR viewing required before site visit to their sales process.

Across three years we shipped 142 versions, supported 28,000 buyer viewings, and handed off zero on-call pages to the client's internal team — that part stayed with us. The product is still in production today, on the same architecture we drew on a whiteboard in early 2022, with the same two engineers leading it.

Apartment AR was acquired by a larger Swiss real-estate group in late 2024. The app, the pod, and the engagement continue under the new owners.

PULL QUOTE / 04
Somyx didn't build us a feature. They built us the part of the sales process that used to be a plane ticket.
CEODACH real-estate agency
Outcome
Versions shipped
142
Buyer viewings
28,000
Sales cycle compression
9 mo → < 4 mo
On-call pages handed off
0
NEXTCase study 04SW · 04 of 10
Automotive SaaS2017 — 2019

A multi-tenant chatbot platform for car dealerships, with three databases earning their place.