Full Transparency

How Haven Analyzes Real Estate

Every claim we make — 847 data points, 4,000+ sources, 10,000 scenario simulations — is explained here. No black boxes.

Our 4,000+ Data Sources

Haven aggregates data across six categories. Below is the breakdown by category with representative source examples. The full source registry is updated quarterly.

Property & Transaction Records

1,200+ sources
  • County assessor and recorder databases (all 50 states)
  • MLS feed aggregation via RETS and RESO Web API
  • Pre-foreclosure and Notice of Default (NOD) filings
  • Probate court records and estate sale filings
  • Absentee owner registries
  • Expired, withdrawn, and cancelled MLS listings

Environmental & Climate Risk

400+ sources
  • FEMA National Flood Insurance Program (NFIP) map data
  • NOAA historical storm and precipitation records
  • EPA Superfund site proximity data
  • USGS earthquake hazard zone maps
  • Wildfire risk indices (CAL FIRE, USFS)
  • Urban heat island temperature data (NASA MODIS satellite)

Neighborhood & Demographic Data

800+ sources
  • US Census Bureau American Community Survey (ACS) 5-year estimates
  • Bureau of Labor Statistics local employment data
  • School district ratings (GreatSchools API, state DOE data)
  • Crime incident data (local PD open data portals, FBI UCR)
  • Walk Score, Transit Score, and Bike Score indexes
  • Business density and local economic activity (NAICS codes)

Market & Rental Economics

600+ sources
  • Rental vacancy rate data by ZIP code (HUD, Census)
  • Median rent trends (Zillow Research, Apartment List public data)
  • Short-term rental occupancy data (AirDNA market-level feeds)
  • Cap rate benchmarks by market and property class
  • Mortgage rate data (Freddie Mac Primary Mortgage Market Survey)
  • HOA fee averages by community and market

Infrastructure & Development Pipeline

600+ sources
  • City planning permit databases (commercial and residential)
  • Transit authority expansion and rezoning plans
  • Department of Transportation road project schedules
  • Broadband availability maps (FCC Form 477)
  • Utility connection and upgrade records
  • School district boundary change filings

Tax & Financial Records

400+ sources
  • Property tax assessment history (county auditor offices)
  • Tax delinquency and lien records
  • Homestead exemption status by parcel
  • Special assessment district maps (Mello-Roos, MUDs, PIDs)
  • Historical appreciation rates by ZIP code (FHFA HPI data)

Total: ~4,000 source feeds. Not all sources are available in all markets. Coverage notes are shown in-app per property.

The 847-Point Property Audit

Each property Haven analyzes is evaluated against a structured schema of 847 data fields across six dimensions. Not every field is populated for every property — availability varies by location and data coverage. The score shown in-app reflects the percentage of available fields where the property meets or exceeds benchmark thresholds.

Property & Title

~180 fields

Deed history, lien status, ownership chain, permit pulls, code violations

Environmental Risk

~140 fields

Flood zone, wildfire, earthquake, Superfund proximity, heat index

Neighborhood

~200 fields

School ratings, crime trend vectors, walkability, business density, demographics

Market Economics

~160 fields

Rental vacancy, cap rate benchmarks, median rent trends, STR data, absorption rate

Infrastructure Pipeline

~80 fields

Planned transit, rezoning filings, road projects, broadband availability, utility plans

Tax & Financial

~87 fields

Assessment history, tax delinquency, special assessments, HOA details, exemptions

10,000-Scenario ROI Simulation

Haven uses Monte Carlo simulation to produce probability-weighted ROI ranges — not a single-point estimate that hides uncertainty. Here is the five-phase process:

1

Base Case Inputs

Haven ingests listing price, estimated rehab cost, local rental comps, current mortgage rates, property tax rate, insurance estimate, vacancy rate assumption, and HOA fees to establish a baseline cash flow model.

2

Variable Ranges

Each input is assigned a probability distribution based on historical variance in that specific market. For example, in Phoenix AZ, vacancy rates have historically ranged from 4% to 11% — this range is modeled, not a single point estimate.

3

Monte Carlo Simulation

10,000 simulations are run by randomly sampling each input variable from its distribution. This produces a full probability-weighted distribution of outcomes rather than a single projection number.

4

Scenario Classification

Results are segmented into three labeled scenarios: Bull (top 20th percentile of outcomes), Likely (median outcome), and Bear (bottom 20th percentile). This gives investors a realistic range, not false precision.

5

Market-Specific Calibration

Distribution parameters are recalibrated quarterly per market using trailing 36-month actuals. A Phoenix model uses Phoenix historical data; an Austin model uses Austin data. No single national assumption applies everywhere.

Important Limitation

Monte Carlo simulations model uncertainty — they do not predict the future. Black swan events (sudden rate shocks, natural disasters, regulatory changes) are not fully capturable in historical distributions. Haven's projections are analytical inputs for decision-making, not guarantees of return. Always consult a licensed real estate professional and financial advisor before making investment decisions.

Methodology questions answered

Where does Haven source its property data?

Haven aggregates data from 4,000+ sources including county assessor and recorder databases, MLS feeds (via RETS/RESO APIs), FEMA flood maps, NOAA climate records, Census Bureau demographic data, local planning permit databases, and rental market indices. The full category breakdown is published on this methodology page.

What are the 847 data points Haven analyzes per property?

The 847 data points span six categories: property and transaction records (title, liens, tax history), environmental and climate risk (flood zone, wildfire, earthquake), neighborhood demographics (school ratings, crime trends, walkability), market economics (rental vacancy, cap rate benchmarks, rent trends), infrastructure pipeline (planned transit, development permits), and tax and financial data (assessment history, HOA, special assessments). Not every data point is available for every property — the count reflects the full data schema; individual property scores are based on available data.

How accurate are Haven's ROI projections?

Haven produces probabilistic range projections (Bull / Likely / Bear) using Monte Carlo simulation across 10,000 scenarios rather than a single point estimate. This is methodologically more honest than single-number projections. The Likely scenario (median outcome) has historically tracked within 8–12% of actual first-year cash-on-cash returns when backtested against closed transactions in our coverage markets. Accuracy varies by market stability and data availability.

How often is Haven's data updated?

Core property and MLS data is refreshed daily. Climate and environmental datasets are updated when source agencies publish new versions (typically quarterly to annually). Demographic and economic data follows Census Bureau release cycles (annually for ACS, every 10 years for full census). Market calibration parameters for ROI simulations are recalibrated quarterly using trailing 36-month transaction actuals.

What is the '98% projection accuracy' claim based on?

This figure refers to the percentage of properties where Haven's Bear scenario floor was not breached in the first year of ownership — meaning the actual outcome fell within the projected Bull-to-Bear range, not below the Bear floor. It does not mean individual point estimates are 98% accurate. We are refining how we communicate this metric to be clearer. The Monte Carlo methodology is described in full on this page.

See the methodology in action

Run a live 847-point analysis and 10,000-scenario projection on any US property — free for 14 days.