Raw GIS data
turned into decisions.
A PropTech startup was sitting on rich geospatial data with no way to make it useful. The platform needed to let commercial real estate professionals query locations, compare sites, and understand placement intelligence without needing a GIS specialist.
The data existed.
The interface did not.
A PropTech startup had built a dataset of commercial property locations, lease rates, demographic overlays, and foot traffic patterns. The data was valuable. But it was locked in flat files and a specialist GIS tool that only two people on their team could operate. Their sales team could not use it. Their clients could not see it. And new hires spent three to four weeks learning the tool before they could produce any analysis.
GIS tool required specialist training with a steep learning curve.
No way for clients to self-serve location analysis.
Site comparison required manual exports and spreadsheet work.
The data was valuable. We just could not get it out of the tool and in front of the people who needed it.
A location intelligence platform built for people, not specialists.
Natural language location search
Users search for locations in plain English. “Show me high-footfall commercial zones within 500m of a major transit hub in Brooklyn with a lease rate under $80 per square foot.” No GIS training required.
AI site scoring
Every property receives an AI-generated suitability score based on your specific tenant criteria.
Comparable analysis
Select any site and see instantly comparable properties ranked by similarity score.
Client reporting
Site comparison reports generated automatically. What took analysts half a day now takes thirty seconds.
Multi-tenant architecture
Multiple brokerage firms can use the platform with complete data isolation. Each client sees only their data and their AI analyses.
Demographic intelligence overlays
Foot traffic, population density, income brackets, and daytime population overlaid directly on the map. Toggle any layer on or off.
We built the pipeline before we built the interface.
The GIS data came in multiple formats from multiple sources. Property boundaries as shapefiles. Foot traffic as CSV exports. Demographic data via API. Lease data as spreadsheet exports. We built a unified ingestion pipeline that normalised all sources into a single PostgreSQL schema with PostGIS extension for geospatial queries.
The AI does not guess. It scores against your criteria.
We built a criteria-based scoring engine. A broker defines what they look for in a commercial tenant placement: foot traffic threshold, income bracket, proximity to anchors, competitor density. The AI scores every property in the dataset against those criteria and ranks the results. The criteria are saved per client.
The hardest part was making complex queries feel simple.
We spent two weeks on the search interface alone. The goal was for a non-technical broker to express a complex location query in plain language and get accurate results. We used Claude API for intent parsing, combined with a structured query builder that translates the parsed intent into PostGIS queries.
The output needed to be something you could send to a client.
We built a report generation layer that takes the AI analysis and formats it into a clean, professional PDF and web view. Site comparison, demographic summary, foot traffic patterns, and AI-generated narrative explanation. The entire report is generated in under 10 seconds.
The numbers after launch.
“The platform changed what our junior analysts can do on their first week. We had people generating client-ready site reports on day two. Before this, that took months of GIS training.”
Discovery and data audit
We spent the first week understanding the data sources, formats, and quality. Defined the unified schema. Scoped the ingestion pipeline.
UX research and architecture
Broker interviews. Mapped the location research workflow end to end. Defined the natural language search requirements.
Visual design and map UI
High-fidelity map interface, property cards, layer controls, report layout. Interactive Figma prototype.
Data pipeline and schema
PostGIS schema. Ingestion pipeline for all data sources. First working map queries.
Platform engineering
Full stack build. Map rendering. Search interface. Multi-tenant architecture. Client report generation.
AI integration, QA, launch
Claude API integration for search. Scoring engine deployment. Multi-tenant QA. Production launch.
React · Next.js
FrontendNode.js
Backend APIPostgreSQL
DatabasePostGIS
GeospatialSupabase
Hosting · RealtimeClaude API
AI · NL parsingMapbox GL JS
Map renderingFigma
DesignReady to build something like this?
Book a free 15-minute call. We will look at what you are building and tell you honestly what it takes and whether we are the right fit.