Walk into any retail site selection meeting. Someone pulls up the crime heatmap. The map is built on FBI Uniform Crime Reporting data, aggregated annually, published 8–12 months late, and biased toward jurisdictions that report consistently. It is the wrong tool for the job, and it has been wrong for a decade.
311 data — the non-emergency municipal call stream — is updated daily, captures behavior rather than legal categorization, and tells you what residents actually experience. Locus ingests 198K 311 records per month across 47 reporting cities. The story it tells diverges meaningfully from the FBI one.
Two cities, two stories
Take a Bronx hex cell and a Bushwick hex cell. By annual UCR data, both look mid-tier. Comparable Part 1 offense counts, comparable trajectory. By 311 data over the last 24 months, the divergence is enormous: Bushwick's disorder complaints are down 31% YoY; the Bronx cell is up 14%. Both have similar 'crime rates' on paper. Only one is gentrifying.
Why 311 leads UCR
Three reasons 311 is the earlier signal. First, behavioral threshold: residents call 311 long before something crosses into a Part 1 offense category. Graffiti tagged on a wall today doesn't show up in UCR ever; the broken-window dynamics it implies show up two years later in the offense data. Second, frequency: 311 publishes daily; UCR publishes annually with a year's lag. Third, demographic composition: gentrifying neighborhoods see 311 complaint volume drop because the population calling 311 is leaving faster than the new population picks it up.
The third reason is also the failure mode. A 311 drop can mean 'neighborhood improving' or 'residents giving up.' Distinguishing the two requires looking at category mix — graffiti and noise complaints declining is improvement; service-request complaints (housing code, garbage) declining is abandonment.
Category structure matters
| Category | What it captures | Direction = improving |
|---|---|---|
| Noise complaints | Resident composition + density | Down |
| Graffiti / vandalism | Disorder & broken-window dynamics | Down |
| Abandoned vehicles | Street-level activity & enforcement | Down |
| Illegal dumping | Infrastructure quality | Down |
| Housing code violations | Building owner accountability | Up (resident pressure) |
| Tree / sidewalk requests | Civic engagement | Up (engaged residents) |
| Sanitation missed pickups | Service quality | Down |
Improving neighborhoods don't show across-the-board 311 decline. They show declining disorder complaints AND rising service-request engagement. That divergence — quieter streets, more residents calling about tree wells — is the strongest single signal Locus reads from this dataset.
What this is good for
Retail siting (especially food & beverage), apartment underwriting in mixed neighborhoods, and any analysis where 'is this block actually getting better or just statistically stable' is the operational question. It is less useful for office siting (where commuter inflow matters more) or industrial (where 311 is dominated by truck-route complaints).