Methodology and press
Where these numbers come from and how to use them
Everything on the crime pages comes from the Minneapolis Police Department's public data. This page says exactly what is counted, how, and where the numbers can mislead.
Source and freshness
The data is the City of Minneapolis open data "Crime_Data" feed, published by MPD and imported here every day. Each daily import looks back 30 days, and every day the site also checks each of the last 24 months against the city's own totals (overall, and by offense type, neighborhood and ward). A month that differs is corrected to match: records MPD posted late are added, revised ones updated, and ones the city withdrew removed. A case the feed lists twice for the same offense counts twice, as it does in the city's totals. The numbers on the site run through November 20, 2026, the date of the newest record in the feed.
MPD posts records days after the fact. Compared with the same weekday a year earlier, records under about ten days old are only partly there: roughly three quarters at a week, about half at three days, a fifth at two. The newest days are shown anyway and fill in as MPD posts them, so the latest stretch can look like a drop that later corrects. Weekly and monthly reports are marked provisional until they have had time to fill in.
What is counted
The unit on the crime pages, the car break-in pages, the neighborhood and ward pages and the homepage charts is an offense, using MPD's own offense count: a case with two victims is two offenses. They all read the same data, so a number matches wherever it appears. The map and the "last 30 days" count on the homepage count pins instead (one per MPD case, plus resident reports), so those can run slightly lower. Dates are the day the offense occurred in Central Time.
Offenses are grouped in plain language from MPD's offense category and offense name. Nothing is guessed beyond that mapping:
- Homicide
- MPD's "Homicide Offenses": murder and non-negligent manslaughter. Counted per victim.
- Aggravated assault
- MPD's "Aggravated Assault": an attack with a weapon or causing serious injury.
- Simple assault
- MPD's "Simple Assault": an attack without a weapon or serious injury.
- Intimidation
- MPD's "Intimidation": threatening someone without a physical attack.
- Robbery
- MPD's "Robbery": taking property from a person by force or threat. Includes carjacking.
- Sex offenses
- MPD's "Sex Offenses", all NIBRS sex offense types.
- Burglary
- MPD's "Burglary/Breaking & Entering": unlawful entry into a building.
- Theft from motor vehicle
- MPD's "Theft From Motor Vehicle": items taken from a car. MPD doesn't record whether a window was smashed or a door was unlocked.
- Motor vehicle theft
- MPD's "Motor Vehicle Theft": the whole vehicle taken.
- Vehicle parts theft
- MPD's "Theft of Motor Vehicle Parts or Accessories": catalytic converters, wheels, stereos and similar.
- Shoplifting
- MPD's "Shoplifting".
- Other theft
- The rest of MPD's "Larceny/Theft Offenses", such as "All Other Larceny" and theft from a building.
- Property damage / vandalism
- MPD's "Destruction/Damage/Vandalism of Property": graffiti, broken windows and other damage. Not specific to vehicles.
- Arson
- MPD's "Arson".
- Fraud
- MPD's "Fraud Offenses": credit card fraud, identity theft, swindles.
- Drug offenses
- MPD's "Drug/Narcotic Offenses".
- Weapon law violations
- MPD's "Weapon Law Violations".
- Other offenses
- Every other NIBRS offense in the feed, including stolen property, counterfeiting, kidnapping and trafficking.
Not counted as crimes. The feed also holds rows that aren't separate offenses. They are shown apart and never added to a crime total, so nothing is counted twice:
- Shots-fired calls
- ShotSpotter detections and 911 calls about gunfire. Counts of calls, not confirmed crimes.
- Gunshot wound victims
- MPD's "Gunshot Wound Victims" count. A separate tally from the offenses above.
- Domestic aggravated assault
- A subset of aggravated assault already counted above. Shown separately, never added in.
- Carjacking
- A subset of robbery already counted above. Shown separately, never added in.
Resident reports are never added to MPD counts
Residents can report a break-in on this site. Those reports are unverified, and some describe an incident MPD also has, because the resident called the police too. The two can't be matched: MPD's records carry no ID we can see and sit at block midpoints, and a resident's date and place are approximate. So every number, rate, ranking and chart here counts MPD data only. Resident reports appear as their own dots on the map and as a separately labeled figure beside the MPD total, never inside it. The share of residents who never told police (a figure the homepage quotes) is the one statistic that comes only from resident reports, and it says so.
How comparisons are made
- Same period. The current year is compared with earlier years over the same calendar window, January 1 through the complete-through date above. A partial year is never set against a full one.
- Months. The month still in progress is drawn as an open marker with no line into it, because a slope toward an unfinished month reads as a drop.
- Projection. The homepage's projected full-year figure divides the year-to-date count by the share of the year that had already happened by this date in earlier years (pooled across them). It is an estimate and is labeled as one.
- Wards. MPD assigns the current (2022) ward boundaries to every record, including older ones. We checked a sample across years and the assignment is consistent, so ward trends compare like with like.
Rates per 1,000 residents
Rates divide offenses by residents counted in the 2020 Census. Resident counts for each neighborhood and ward are built from Census Bureau block counts, assigning each block to the city's neighborhood and ward boundaries by its internal point; the totals add to 429,956 residents. Two cautions: places with many workers and visitors (Downtown, Uptown, industrial areas) look high because their offenses are set against residents only, and a rate is withheld for any neighborhood with fewer than 1,000 residents. The resident counts are a 2020 snapshot and aren't updated each year. The build script is in the repository.
Locations
MPD places every record at the midpoint of its block rather than its exact address, and this site uses those points as given. Resident reports are snapped to the same block midpoints. The "near me" tool counts offenses whose block midpoint is within the chosen radius, so it describes blocks, not addresses.
What is a car-related offense?
"Car-related" on the homepage and the trends pages means four of MPD's offense types, added together. MPD's own words for each:
- Theft from motor vehicle. MPD's "Theft From Motor Vehicle": items taken from a car. MPD doesn't record whether a window was smashed or a door was unlocked.
- Motor vehicle theft. MPD's "Motor Vehicle Theft": the whole vehicle taken.
- Vehicle parts theft. MPD's "Theft of Motor Vehicle Parts or Accessories": catalytic converters, wheels, stereos and similar.
- Property damage / vandalism. MPD's "Destruction/Damage/Vandalism of Property": graffiti, broken windows and other damage. Not specific to vehicles.
Two things to know. Property damage/vandalism is the one type that isn't specific to cars: MPD has no field that separates a smashed car window from a broken garage window or graffiti, so it covers all property. And MPD doesn't record how someone got into a car, which is why this site also collects resident reports.
Left out: burglary of a building, robbery (carjacking is counted there), and every other offense type. They are in the all-crime numbers, and car-related offenses are already inside those totals, so don't add the two. Resident reports of smashed windows and rifled-through cars are shown as separate dots and rows, never added to an MPD count.
Weekly bulletins
The weekly bulletins cover Monday to Sunday and appear once the whole week is inside the data window above. Each week is compared with two baselines: the average of the four weeks before it, and the same week 52 weeks earlier, so weekdays line up. An offense type is marked "above range" when its count is more than two standard deviations over its mean for the prior eight weeks and is at least 10 offenses. With counts this small, about one type in twenty will cross that line by chance in any week, so a mark means look closer, not that something changed. The block table lists block ranges with 3 or more thefts from vehicles, vehicle thefts or parts thefts in the week. It uses the block ranges MPD publishes and nothing finer, and one event with several victims can fill it. The bulletins describe places, times and offense types only. They never name or profile people.
Known limits
- These are reported crimes. Many incidents, especially car break-ins where nothing is taken, are never reported.
- MPD can reclassify or add records after publishing, so a past month can change slightly.
- 2019 is the first year of MPD's current records feed. Its early months (January to April) run well below every later year, so the earliest part of the 2019 line may understate what happened.
- Times of day are as reported and are often rounded to the hour.
- MPD's data doesn't say whether a car window was smashed or a door was unlocked, which is why this site also collects resident reports.
- Counts are totals, and a ranking of places by count mostly reflects where people live, work and gather. Use the per-resident rates, and the caution above, before drawing conclusions about a neighborhood.
- This is an independent, resident-run project. It is not affiliated with the Minneapolis Police Department, and nothing here is an official crime statistic.
How to cite this
Suggested citation (change the date to the day you accessed it):
Glass on the Street. (2026). Minneapolis crime statistics. Retrieved November 22, 2026, from https://glassonthestreet.com/crime. Data: Minneapolis Police Department, via City of Minneapolis open data, through November 20, 2026.
Every page states the date its data runs through. The month-by-month reports at /monthly have a permanent link for each month.
For journalists and researchers
- Downloads and API: CSV files by month, neighborhood and ward, plus a JSON API behind every chart.
- Monthly reports: a short, sourced summary of each month with the headline numbers and biggest movers.
- Every chart has a table view and a CSV download, a PNG download and an embed code.
- Incident-level data, with the exact fields we use, is at the source feed.
- Questions or corrections: hello@glassonthestreet.com
Corrections
Problems found in the data, and what changed, are logged with dates on the data quality page. If a number here is wrong, it will be fixed and listed there.