About the data

Why this site exists

In 2023 a candidate came to Knowi's Oakland office for a job interview. As she sat in her car afterwards, a masked man smashed the window and took her purse. We felt awful about it. We build data software for a living, and the city already publishes every police report, so we did the one thing we could: make that data easy for anyone in Oakland to see, for their own block, without needing an analyst. That is how Oakland Crime Watch started.

Source

The crime data comes from police reports the Oakland Police Department makes public through the City of Oakland's Open Data platform: CrimeWatch Data (ppgh-7dqv). Rows include a category, date and time, case number, block-level address, and police beat. Knowi pulls the feed through the city's API and stores governed datasets that power the site.

You can download the full raw file yourself as CSV from the city — it is the fastest way to check any number here.

Update frequency and the 90-day lag

We refresh from the city every day, but recent totals are incomplete and grow as reports are processed. The city warns that "there may be delays in data due to data processing, incident reporting, or technical issues," and says processing can take up to 90 days after the end of each month. Practically: to see all of March, wait until the end of June. Every page's footer shows the latest incident date in the data and when we last refreshed.

Fair comparison windows. Because of that lag, comparing the newest weeks with the same dates a year ago would show a misleading decline — last year's reports have had more time to arrive. Every year-over-year figure therefore ends at the end of the most recent month that is at least 90 days old (currently Jun 30, 2026) and uses the identical calendar window a year earlier. Raw recent counts are still shown, labelled "reported so far"; they are provisional and may change as reports arrive or classifications are revised, so avoid comparing them with an older, more fully processed period.

How we count

One incident is one OPD case number. The city's feed can publish several rows for one case, so the app deduplicates case numbers before computing any public-page figure. Every case the city publishes is counted, including categories that are not crimes in the everyday sense — recovered vehicles, towed vehicles, warrants, missing persons. Those sit in the "Everything else" family and are never counted as violent, burglary, vehicle or theft.

• Dates. Every chart and count uses the incident date and time OPD recorded, not the date the report was filed or published.
• Category. A case is counted under the category it currently carries in the city's data; when a case has been published under more than one category, the most recent version wins.
• Missing beat or location. A case with no police beat still counts in citywide totals but appears on no beat page; a case with no coordinates appears in lists and counts but not on maps. The city rounds every address to the hundred block, so map locations are blocks, not exact addresses.

How we group categories

OPD files reports under roughly 45 categories. To make trends legible we roll them into five families. The mapping is exact and fixed; nothing is reassigned by hand.

Family	OPD categories included
Violent crime	HOMICIDE ROBBERY FELONY ASSAULT FORCIBLE RAPE
Burglary	BURG - RESIDENTIAL BURG - COMMERCIAL BURG - OTHER
Vehicle crime	STOLEN VEHICLE BURG - AUTO STOLEN AND RECOVERED VEHICLE
Theft & fraud	PETTY THEFT GRAND THEFT EMBEZZLEMENT FRAUD FORGERY & COUNTERFEITING POSSESSION - STOLEN PROPERTY
Everything else	Every category not listed above

The app applies this published mapping again when it builds both recent and ten-year family series. The definitions on this page are the source of truth for figures on public pages.

Violent crime follows OPD's definition (the FBI's Part 1 violent offences): homicide, robbery, felony assault and rape. That is why the violent-crime trend here tracks the direction of the department's own year-end reports. One correction is applied first: the city files coroner cases, described as SC UNEXPLAINED DEATH, under the HOMICIDE category, and they outnumber actual homicides ten to one. Cases whose charges are only an unexplained death are relabelled UNEXPLAINED DEATH and counted under Everything else. Official homicide totals published by OPD can still differ slightly from this feed, which lags and is classified at the time of the report.

Beats, areas and neighborhoods

OPD divides the city into 5 police areas and 57 community policing beats. Boundaries come from the city's Police Beats dataset, simplified slightly for the maps. Neighborhood names are the labels OPD uses for each beat; many beats cover several neighborhoods and a few names are approximate. City contacts on beat pages are the Community Resource Officer and Neighborhood Services Coordinator the city lists for that beat.

Rates

Per 1,000 residents divides a year's total by an annual population estimate for Oakland; the current year uses the latest estimate available. Per 1,000 residents at beat level divides a beat's count for the last 12 months by its 2020 Census population (census blocks assigned to the beat that contains them). Neither is a victimisation rate; large commercial or transit beats attract reports from people who do not live there, so downtown and the Coliseum area rate high. Rates use residents only; people who work in or visit an area are not counted, so a rate is not a measure of any one person's risk.

Limits

• Addresses are rounded to the hundred block by the city; map markers are approximate.
• Timestamps are as recorded by OPD. For property crime discovered later, the time is often when the report was filed.
• Categories reflect the initial classification and may be updated after investigation.
• A rise in reports can reflect more crime, more reporting, or a change in how OPD records — the data cannot separate these on its own.

How we count 311 requests

The 311 pages and the city-services section of each beat page use a second city feed, City of Oakland Open Data — OAK 311 service requests: requests made by phone, app, web or email since 2013. We pull it every morning, and every figure runs through the last complete day; the date is shown on each page. A 311 request is a request for a city service, such as clearing dumped debris or fixing a street light. It is not a crime report.

• A status is the city's record. "Marked closed" means the city closed the request in its system. The feed does not say whether the problem was fixed, and a request can stay open in the system after the work is done.
• Requests counts every request opened in the last 12 months, including those later cancelled. Several residents often report the same problem, so requests outnumber problems. More requests from an area can mean more problems or more residents who report them.
• Still marked open counts requests opened in the last 12 months whose status is OPEN, PENDING, WOCREATE, WAITING ON CUSTOMER. Requests opened more than a year ago are not counted, whatever their status, so a request drops out of this figure when it turns a year old even if nothing has changed.
• Marked closed within 30 days and median days until marked closed are measured on requests opened in the 12 months that ended 30 days ago, so each has had at least 30 days. Only requests marked CLOSED or still open are included. Every other status has no close date in the feed and is left out: CANCEL, REFERRED, EVALUATED - NO FURTHER ACTION, UNFUNDED, GONE ON ARRIVAL, REQUEST COMPLETE. The page on how long requests stay open shows how many requests fall in each group. A request that is still open counts as not yet closed. The median is the day by which half of those requests had been marked closed; where fewer than half have, the site says "most still open" and gives no number. Requests opened 30 to 60 days ago have had little time, so the median leans slightly long.
• Beats are not compared on closure times. Different kinds of request close on very different schedules, and beats file different kinds, so a beat's overall median mostly reflects what its residents ask for.
• Request types are the city's own category codes with plain-language labels. The police beat and council district on each request are the city's.
• Addresses in the city's 311 feed are exact. This site rounds them to the hundred block, and rounds map positions to about a block. Request numbers are shown so a request can be followed up with the city.
• Rates divide a beat's 12-month requests by its 2020 Census residents.
• Older history. Requests opened more than 400 days ago are no longer refreshed from the city, so their status on this site is as of the day they passed that age.

Corrections

If a figure looks wrong, write to hello@knowi.com with the page and the date you viewed it. Data through Oct 29, 2026.

Credits

Crime data: City of Oakland / Oakland Police Department. Beat boundaries: City of Oakland GIS. Photography, via Wikimedia Commons:

• Downtown Oakland skyline seen across Lake Merritt on a clear day — Daniel Ramirez, CC BY 2.0
• Telegraph Avenue in Temescal, Oakland, with a Welcome to Temescal banner across the street — Mx. Granger, CC0
• Oakland at dusk seen from the hills, city lights coming on and the bay beyond — Joe Parks, CC BY 2.0
• Port of Oakland at sunset, boats in the marina and cranes on the horizon — Tony Webster, CC BY 2.0
• The Fox Theater marquee on Telegraph Avenue in Uptown Oakland — Andrew Robles andrewroblesphoto, CC0
• College Avenue in Rockridge, Oakland, shops and street trees — Mx. Granger, CC0

Built and maintained by Knowi, Oakland, on Knowi Apps.