×
Elm City Explorer
An interactive map revealing who really owns New Haven's corporate-controlled properties.
▶ Replay intro
📖 Findings — Data investigations
The Problem
In New Haven, thousands of rental properties are owned by shell LLCs — entities like "NETZ CONSOLIDATED III LLC" or "SHEER OPULENCE LLC" that obscure who actually controls them. Tenants, journalists, and policymakers have no easy way to see who the real landlords are, how many properties they control, or where they're concentrated.
The Solution
Elm City Explorer links 6,953 corporate-owned properties to 421 identified ownership groups by tracing connections through public records. It answers questions like: Who owns the most property in my neighborhood? How many shell LLCs does this landlord operate through? When did they start buying?
Who It's For
JournalistsInvestigate landlord practices and corporate concentration
Housing AdvocatesIdentify major landlords and track acquisitions
City PlannersUnderstand ownership patterns by neighborhood
ResearchersExport data for academic or policy analysis
How It Works
1. Scrape6,953 corporate properties from VGSI assessment database
2. EnrichCT Secretary of State filings: agents, principals, formation dates
3. ClusterGraph-based entity linking via shared principals & agents
4. MapGeocode, assign neighborhoods, compute concentration
5. RefreshRe-scraped weekly via GitHub Actions; changes are surfaced in the Insights report
Entity Resolution
Most corporate landlords operate through dozens of single-purpose LLCs. For example, "Mandy" (841 properties) operates through 100+ entities. We link them by:
Shared Principals Same person listed as principal across multiple LLCs
Shared Agents Same registered agent (excluding ~30 law firms and RA services)
Shared Addresses Same business address (excluding known law firm offices)
Pattern Matching Naming conventions (e.g., "NETZ *", "NHCD *", "BBYM *")
Clustering uses a Union-Find algorithm. Connected components form ownership groups. Coverage: 421 groups covering 4,577 of 6,953 properties (65.8%).
Features
Interactive Map Building footprints colored by ownership group
AI Chatbot Ask natural language questions about the data (powered by Claude)
Neighborhood Analysis 20 neighborhoods ranked by corporate concentration
Group Profiles Researched blurbs on 70+ ownership groups
Network Graphs Visualize shell entity connections
Property Type Filters Residential, commercial, condo, institutional
Comparison Table Sort all 421 groups by value, count, entities
Printable Reports Per-group reports with full property listings
CSV Export Full dataset downloadable for analysis
Guided Tour Short product walkthrough, plus a separate key findings briefing
Data Sources
VGSI (gis.vgsi.com/newhavenct) — Property assessments, sale history, building footprints, zoning, use codes.
CT Secretary of State (CONCORD) — Business entity filings for 3,928 LLCs.
New Haven GIS — Neighborhood boundary polygons.
Data Coverage
A small share of corporate properties appear in the public records but cannot be placed on the map because they could not be geocoded to a parcel location. The exact figure is recomputed each weekly refresh.
Built With
Python
Leaflet.js
Claude API (Sonnet 4)
Vercel
Canvas API
CT SOS CONCORD
VGSI
Data scraped and processed with Python. Entity resolution via Union-Find graph clustering. Frontend is a single HTML file (~2,500 lines) with no build step. AI chatbot uses Vercel serverless functions proxying to Claude Sonnet 4. Deployed on Vercel's free tier.
Coming Soon
The data pipeline (scrape → cluster → map) generalizes to any city with public assessor and secretary-of-state data. We're starting with New Haven and planning to expand to Hartford, Bridgeport, Providence, and Boston.
Limitations
This is a research tool, not a definitive ownership registry. Some connections may be spurious. Some owners may be missing if they don't file with CT SOS. Sale prices of $0 typically indicate related-party transfers. Data was scraped March 2026.