SAFETY ATLAS VOL. I · 2026
Independent public-safety research for American cities, published annually.
Vol. I · No. 1 · The 2026 American Public-Safety Index Published May 2026 · Methodology v1.0
2026 Edition · The Annual Index

Public-safety
rankings, anchored
to federal data.

FBI UCR & NIBRS Population-adjusted Methodology published

In December 2021, Realtor.com, Redfin, and Trulia stripped crime maps from their listings. The professionals who relied on them have had no replacement they can defend in front of a client. Safety Atlas is built on the FBI's own data, with a published formula, no machine-learning imputation, no opaque algorithm, no auto-renewal traps.

By the
numbers.
50+ DC
States covered
18
Offense categories
log10(n)
Population adjustment
v1.0
Methodology, published
§ 02 · For whom this is written
01 · Relocation

Client-ready answers to the “which neighborhood is safe?” question.

Drop the rankings, the per-domain scores, and the methodology page into your relocation packets. Stop linking to forum posts.

02 · Insurance

Quantitative talking points for homeowners and small commercial risk.

Per-city offense rates indexed to the national figure. On every page; identical methodology in every state.

03 · Editorial

An annual ranking you can cite with a clean reference.

We did the work, the methodology is open, the source data is public. Your byline keeps the credit; we keep the footnote.

04 · Analytic

The underlying data, in Excel, for your own modelling.

Composite score, per-100K rate, raw counts, year-over-year movement. One row per city. Ready for join.

§ 03

Editions & Pricing

One-time purchase · PDF in 5 minutes
SKU · STATE-EDITION
Single-State Edition
$149/ state
One U.S. state. ~15-page PDF + state Excel feed.
  • Top 25 safest + Bottom 25 most dangerous cities
  • Watchlist of cities trending worse YoY
  • Adjusted & raw composite scores, three-domain breakdown
  • State Excel feed: every city × offense × score × YoY
  • Published methodology + citation rights
  • One-click cancel, seven-day refund
Purchase
SKU · REGIONAL-5
Regional Pack of Five
$399
Five states of your choice. Cross-state ranking workbook included.
  • Five Single-State Editions, your pick
  • Shared cover & methodology
  • Cross-state Excel ranking workbook
  • For multi-state relocation, insurance, or news teams
  • Save $346 versus buying singly
Purchase
SKU · STATE-CRIME-BRIEF
State Crime Intelligence Brief
$2,499
One state. NIBRS-grade crime depth. Thirty-plus pages.
  • Top 50 safest + Bottom 50 most dangerous cities
  • Per-neighborhood crime profiles for top-10 cities
  • NIBRS detail: victim demographics, offender profiles, weapon usage
  • Hotspot locations + domestic-violence intimate-partner share
  • Victim–offender relationship breakdown by offense
  • Watchlist of cities trending worse, per-offense leaderboards
  • Twelve-month state-scoped API key, one refresh
Purchase
§ 04

Why us, not them.

A frank comparison · sourced
  Safety Atlas NeighborhoodScout1 CrimeGrade2 AreaVibes3 ATTOM4
Source FBI UCR + NIBRS, only. Proprietary spatial algorithm. FBI + ML imputation. FBI + StatsCan, livability blend. Bundled "community" attribute.
Methodology Published. Auditable. Open formula. Closed; critiqued by independent analysts.5 Closed; ML-imputed by their own docs.2 Closed. Closed.
Population-adjusted Yes — log10(pop) factor. No. No. No. No.
Excel / CSV export Every tier. Enterprise only. Paid data package. None. Enterprise only.
Cancellation One-click. Seven-day refund. Public auto-renewal complaints on record.6 No subscription. Free, ad-supported. Quote & contract.
Entry price $149 / state. $30 / single neighborhood report.1 Free + paid bundles. Free. $95+/mo, mostly enterprise.4

The Realtor.com / Redfin / Trulia crime layer disappeared in 2021.7 The replacements that exist today are opaque, machine-imputed, or gated to enterprise. Safety Atlas is positioned for the working professional in between.

  1. NeighborhoodScout subscription plans, neighborhoodscout.com (May 2026).
  2. CrimeGrade methodology, crimegrade.org/about-crimegrade-data.
  3. AreaVibes methodology, areavibes.com/methodology.
  4. ATTOM pricing profile, datarade.ai/data-providers/attom.
  5. Jeff Asher, "Another cautionary crime-data tale," jasher.substack.com.
  6. NeighborhoodScout customer complaints, ComplaintsBoard.
  7. Redfin, "Neighborhood crime data doesn't belong on real-estate sites," redfin.com/news (Dec 2021); Trulia followed weeks later (Inman).
§ 05 · Beyond the catalogue

Multi-state, redistribution, branded reports, dataset licensing.

Place Intelligence across multiple states, dataset licensing for prop-tech and insurance, branded white-label PDFs for relocation networks, weekly refresh, embeddable widgets. Custom-quoted against the work.

Contact for custom
§ 06

Methodology

Version 1.0 · Anchored to the national rate · No ML imputation

Every offense’s per-100K rate is compared against the national figure. A city at the national average scores fifty.

01 · Composite anchoring

At zero offenses, the score is one hundred. At twice the national rate, the score is zero. No proprietary weights, no curve fitting, no black-box scoring layer.

02 · Three domains, weighted

Personal (50%), property (30%), community (20%). The weights reflect what a reasonable buyer or relocator typically prioritises when reading these numbers.

03 · Population adjustment

A logarithmic factor keeps very small towns from dominating the top of the rankings purely on the basis of low absolute counts.

04 · Source data

FBI Uniform Crime Reporting (UCR) and National Incident-Based Reporting System (NIBRS), aggregated to the city level via agency-to-city mappings. Public domain.

// per-offense score offense_score = max(0, 100 − 50 × (city_per_100k / national_per_100k)) // domain composites personal_safety = mean(HOM, RPE, ROB, ASS, 100, 64A) property_security = mean(BUR, LAR, MVT, ARS, 23D, 23F) community_order = mean(13B, 11A, 280, 290, 520, 35A) // composite (raw) safety_score = 0.50 × personal + 0.30 × property + 0.20 × community // population-adjusted variant pop_factor = clamp(log10(pop) / 5, 0.4, 1.0) crime_penalty = 100 − safety_score adjusted_score = 100 − (crime_penalty × pop_factor)