Methodology · Machine-Readable
How SpotCrime scores safety
SpotCrime's free city-level Crime Grade is documented here in full: the data source, the metric, the exact cutoffs, the refresh cadence, and the limitations. We also explain the ZIP-level crime score served by CrimeScore, a companion service powered by SpotCrime data. No proprietary black box — if you can't audit a safety score, you shouldn't act on it.
Two scores, two jobs
SpotCrime's free Crime Grade and CrimeScore's ZIP-level percentile answer different questions at different resolutions — and neither pretends to do the other's job. (The CrimeScore score is a separate service, powered by SpotCrime data; it is not part of the SpotCrime API.)
| Crime Grade (SpotCrime) | Crime score (CrimeScore) | |
|---|---|---|
| Question it answers | How does this city compare to other cities? | How does this ZIP rank against every ZIP nationally? |
| Resolution | City / agency level | ZIP / county (FIPS) level |
| Scale | Decile 1–10 + letter A–F (1 = safest) | 0–100 national percentile (higher = more crime) |
| Source data | FBI UCR via Crime Data Explorer (annual) | SpotCrime incident data, scored by CrimeScore |
| Refresh | Twice a year (FBI September release + mid-year pass) | Monthly snapshots |
| Where you get it | /crime-grade pages — free | crimescore.info — powered by SpotCrime data |
The Crime Grade, step by step
The same calculation runs on every one of the 3,700+ city pages. Here is the entire method — there is nothing else behind it.
- 1
Pull the source data
Agency-level violent crime counts and covered populations, pulled directly from the FBI Crime Data Explorer (Uniform Crime Reporting program). No third-party aggregators, no resold data.
- 2
Compute the rate
Violent crime rate per 100,000 residents = (murder + rape + robbery + aggravated assault) ÷ covered population × 100,000, for the most recent fully-reported calendar year.
- 3
Rank against a fixed comparison population
National decile:the city's rate placed against all FBI-reporting U.S. cities with population ≥ 25,000. State decile: the same calculation restricted to cities in the same state with population ≥ 10,000. A city can be a national C and a state A — both are published on its page.
- 4
Bucket into deciles 1–10
The rate is compared against nine published cutoffs (table below). At or below the first cutoff = decile 1 (safest ~10%). Above the last cutoff = decile 10. Cutoffs are recomputed annually from the full national pull — they are not tuned by hand.
- 5
Map the decile to a letter
1–3 = A · 4–5 = B · 6–7 = C · 8 = D · 9–10 = F. The letter is a display convenience; the decile is the score.
National decile cutoffs, published
Violent crime rate per 100,000 residents. A city at or below a cutoff falls in that decile. State deciles use per-state cutoffs recomputed from that state's reporting cities.
| Decile | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| Rate ≤ | 120 | 180 | 260 | 360 | 480 | 620 | 800 | 1050 | 1400 | > 1400 |
| Letter | A | A | A | B | B | C | C | D | D | F |
Reference points: the U.S. national violent crime rate has run roughly 370–400 per 100,000 in recent years — putting the “average American city” experience near deciles 4–5.
What the letters mean
Crime score (via CrimeScore)
Neighborhood crime scores are served by CrimeScore (crimescore.info), a companion service powered by SpotCrime data — they are not part of the SpotCrime API. CrimeScore returns a national crime-intensity percentile plus national and state rank for any U.S. ZIP or county FIPS code, across 13,000+ ZIPs, refreshed in monthly snapshots. Population density is used to smooth and normalize the score.
How the two scores relate. Crime Grade is SpotCrime's live, free, annual, city-level, FBI-sourced decile (A–F). The CrimeScore score is a monthly, ZIP-level national crime percentile served by CrimeScore, powered by SpotCrime data. Crime Grade grades a whole city; the CrimeScore percentile ranks a single ZIP against every scored ZIP nationally. Get a Crime Grade free on spotcrime.io; get a CrimeScore token at crimescore.info.
Access: published_score(region_code) — served by the CrimeScore MCP at mcp.crimescore.info, not the SpotCrime API. Status: Live via CrimeScore — not a SpotCrime API endpoint. More: Crime Data MCP →
Limitations — stated, not buried
Every crime score has failure modes. Ours are these. If a vendor's methodology page doesn't have a section like this, ask why.
Reported, not occurred
The FBI relies on local agency reporting. Rates reflect what was reported to police and forwarded to the FBI, not everything that occurred. Criminologists call the gap the 'dark figure of crime' — victimization surveys consistently find roughly half of violent crime is never reported to police.
Agency participation gaps
Not every agency reports in a given year. During the SRS→NIBRS transition, some large departments missed reporting windows. A missing year for a major city shifts the national comparison pool.
Rates are not personal risk
A citywide rate averages very different neighborhoods. Two addresses in the same city can carry very different risk profiles — which is why finer-grained scores (CrimeScore's ZIP-level percentile, powered by SpotCrime data) complement the live city-level Crime Grade.
Population denominators
Rates use resident population. Cities with large commuter or tourist inflows (small resident base, large daytime population) can show inflated per-capita rates.
A grade is a comparison, not a verdict
An F means 'highest decile among FBI-reporting cities' — it does not mean 'do not go there.' Most neighborhoods in most cities are safe most of the time. Hold two truths: take the data seriously, and respect the people who live in it.
Machine-readable surfaces
This page is the human rendering. Agents and pipelines should pull the structured versions — all generated from the same source module, so they cannot disagree.
Methodology FAQ
- What data source does the SpotCrime Crime Grade use?
- FBI Uniform Crime Reporting via the Crime Data Explorer. Agency-level data, pulled directly. No third-party aggregators, no proprietary score.
- What metric does the Crime Grade use?
- Violent crime rate per 100,000 residents (murder + rape + robbery + aggravated assault), most recent fully-reported year.
- How is the national decile calculated?
- A city's rate is placed against all FBI-reporting U.S. cities with population ≥ 25,000 and bucketed into deciles 1–10. 1 = safest.
- How is the state decile calculated?
- Same calculation restricted to cities in the same state with population ≥ 10,000.
- How do decile scores convert to a letter grade?
- 1–3 = A · 4–5 = B · 6–7 = C · 8 = D · 9–10 = F.
- What are the national decile cutoffs?
- Violent crime rate per 100,000: 120, 180, 260, 360, 480, 620, 800, 1050, 1400. A city at or below the first cutoff is decile 1; above the last cutoff is decile 10. Cutoffs are recomputed annually from the full national pull.
- How often is the Crime Grade refreshed?
- Twice a year. The FBI typically releases the prior year's full data each September.
- What is the difference between Crime Grade and the CrimeScore score?
- Crime Grade is SpotCrime's live, free, annual, city-level decile (1–10, A–F) computed from FBI data. The CrimeScore score is a monthly, ZIP-level national crime-intensity percentile served by CrimeScore (crimescore.info), a companion service powered by SpotCrime data — it is not part of the SpotCrime API. Both are described on this page.
- What are the limits of this methodology?
- The FBI relies on local agency reporting. Rates reflect what was reported, not what occurred. A citywide grade averages very different neighborhoods, and per-capita rates can be skewed in cities with large commuter or tourist populations.
- Is the methodology available machine-readable?
- Yes. The full methodology — cutoffs, populations, metric definitions, refresh cadence, and limitations — is served as JSON at https://spotcrime.io/methodology.json.
Audit us, then build on us
Compare this methodology against any crime-data vendor's. If theirs is more transparent, we'll say so. If it's a black box, that tells you something too.