About Public Analyst.ai
Public Analyst.ai is a crime-trend intelligence platform built from public incident data. Every month it publishes per-city briefings, 346 neighborhood pages across 6 live cities, signals leaderboards, and 12-month forecasts — methodology first, no scoreboards.
What it is
The product turns raw police-incident data into tracked signals: spikes, drops, sustained shifts, rare events, and streak breaks against a multi-year baseline. Every signal has a strict numeric threshold described on the methodology page; nothing is editorial judgment in the signal layer.
Above the signal layer sit the briefing prose, neighborhood context, and forecast performance reports — all grounded in the same structured data, so any number on the site can be traced back to its source row in the public dataset.
What it isn't
- Not a scoreboard.Crime totals are misleading without context (population, reporting practice, category definitions). The site indexes on month-over-month change, seasonality, and direction — what's actually moving.
- Not a policy advocacy site.The prose reports the trend; it doesn't tell you what should be done about it.
- Not a real-time feed. Public incident data arrives with a reporting lag. Each monthly briefing covers a fully-settled month — published seven days after month-end so late-reported incidents are included.
Where the data comes from
Each city's incidents are pulled from its public open-data feed and mapped to ten UCR Part 1 categories, aggregated to neighborhood × category × month, and run through the anomaly + forecast engines described on the methodology page. Per-city data sources, NIBRS-migration year, and reporting-lag windows are documented inline on the methodology page.
Population data for per-capita rate calculations comes from the Census ACS 5-year estimates, refreshed annually. Neighborhood boundaries follow each city's official analytical unit — DataSF Analysis Neighborhoods for San Francisco, hand-curated tract groupings for Oakland, the 77 community areas for Chicago, and equivalents in other cities as they ship.
How often it updates
The full pipeline — re-ingest, aggregate, score, forecast, publish — runs on the 7th of every month at 7am Pacific. The buffer past month-end gives each police department's reporting lag time to settle so the briefing reflects a complete month, not a partial one. Lag windows vary by city (SF and Chicago ≈ 7 days, Oakland ≈ 30 days) and are documented per city on the methodology page.
Cities
6 live: Chicago, Los Angeles, New York, Oakland, San Francisco, Seattle.