How the numbers on Where 311 complaints pile up were made, with every filter and formula. The build log lists each judgment call and who made it.
New York City's 311 Service Requests from 2020 to Present (dataset erm2-nwe9) on NYC Open Data, queried through its Socrata Open Data API (SODA) endpoint https://data.cityofnewyork.us/resource/erm2-nwe9.csv on Oct. 14, 2026. At that moment the dataset held 22,788,256 service requests from Jan. 1, 2020, to Oct. 13, 2026; the newest was logged at 3 a.m. that day. Each service request is counted as one complaint. Times are 311's created_date, which is New York local time.
incident_address, incident_zip and complaint_type and counting rows. The quarterly counts sum to 22,788,256, matching the dataset's own row count exactly. The build stops if they don't match.A location is a street address with a house number, plus its five-digit ZIP code. Before matching, every address is uppercased, stripped of extra spaces (older records store some addresses with doubled spaces, such as 655 EAST 230 STREET) and has six abbreviations spelled out: BLVD, AVE, ST at the end, RD at the end, PL at the end and PKWY. That rewrite touched 22,924 complaints under 4,790 spellings.
An entry counts as a street address when it starts with a house number followed by a street name. Entries with no house number, and bare numbered streets such as 1 AVENUE or 85 STREET, name a stretch of street, a park or an airport rather than one place. They are set aside: 1,243,802 complaints under 26,327 such entries, led by JOHN F KENNEDY AIRPORT in 11430 (8,209), JFK in 11430 (6,975), FLUSHING MEADOWS CORONA PARK in 11368 (4,900), LA GUARDIA AIRPORT in 11369 (3,797), RIVERSIDE PARK in 10025 (3,525), CENTRAL PARK in 10000 (2,878). Another 944,278 complaints have no address at all. Together that is 2,188,080 complaints, 9.6% of the total, that count toward every total but are never ranked as a location.
That leaves 1,301,096 street addresses with at least one complaint. Addresses with no complaints are not in the data, so they are not counted.
Addresses are sorted by total complaints, largest first, with ties broken alphabetically by address. The top X% is the first round(N × X) addresses, where N = 1,301,096. A slice's share is the complaints at those addresses divided by all 22,788,256 complaints, including the ones with no street address. Using only complaints with a street address as the base would raise the top 5% figure from 51.6% to 57.1%.
| Slice | Addresses | Fewest complaints in slice | Complaints | Share of all |
|---|---|---|---|---|
| Top 0.01% | 130 | 2,627 | 1,147,933 | 5% |
| Top 0.1% | 1,301 | 855 | 2,619,702 | 11.5% |
| Top 1% | 13,011 | 205 | 6,887,658 | 30.2% |
| Top 5% | 65,055 | 51 | 11,763,957 | 51.6% |
| Top 10% | 130,110 | 27 | 14,126,775 | 62% |
| Top 20% | 260,219 | 13 | 16,555,250 | 72.6% |
| Top 50% | 650,548 | 4 | 19,468,561 | 85.4% |
The median address has four complaints; 349,306 addresses have exactly one. The 10 busiest addresses hold 2.5% of all complaints and the 100 busiest 4.7%. With the 100 busiest removed from both the top and the total, the top 5% still holds 49.3%.
The slider stops every 0.01 percentage point up to 0.1%, every 0.1 point up to 10% and every point after that. Below 0.1%, the slice is less than one square in the address grid, so that square is drawn partly filled.
311 uses 261 complaint type names in this period once spellings that differ only in capitalization are merged (the most common spelling is kept). The page groups them into six categories. The grouping is a judgment call; the full list:
| Category | Complaints | Complaint types in it |
|---|---|---|
| Apartment and building conditions | 4,682,723 | Appliance, Boilers, Door/window, Electric, Elevator, Flooring/stairs, General, Heat/hot water, Indoor sewage, Non-residential heat, Outside building, Paint/plaster, Plumbing, Safety, Unsanitary condition, Water leak |
| Noise | 5,186,968 | Every complaint type whose name starts with "Noise" |
| Parking and vehicles | 4,759,893 | Abandoned vehicle, Blocked driveway, Derelict vehicle, Derelict vehicles, Illegal parking |
| Police and street order | 1,223,022 | Animal-abuse, Bike/roller/skate chronic, Disorderly youth, Drinking, Drug activity, Encampment, Homeless person assistance, Homeless street condition, Illegal fireworks, Mass gathering complaint, Non-emergency police matter, Panhandling, Posting advertisement, Traffic, Urinating in public, Vendor enforcement |
| Streets, sanitation and trees | 5,212,433 | Broken parking meter, Bus stop shelter complaint, Commercial disposal complaint, Curb condition, Damaged tree, Dead animal, Dead/dying tree, Dirty condition, Dirty conditions, Dumpster complaint, Electronics waste appointment, Graffiti, Highway condition, Illegal dumping, Illegal posting, Illegal tree damage, Litter basket / request, Litter basket complaint, Litter basket request, Lot condition, Missed collection, Missed collection (all materials), New tree request, Obstruction, Overflowing litter baskets, Overgrown tree/branches, Request large bulky item collection, Residential disposal complaint, Rodent, Root/sewer/sidewalk condition, Sanitation condition, Sanitation worker or vehicle complaint, Sewer, Sewer maintenance, Sidewalk condition, Snow, Snow or ice, Standing water, Storm, Street condition, Street light condition, Street sign - damaged, Street sign - dangling, Street sign - missing, Street sweeping complaint, Traffic signal condition, Unsanitary pigeon condition, Vacant lot, Water conservation, Water maintenance, Water system, Wood pile remaining |
| Everything else | 1,723,217 | Every complaint type not listed above |
The chart of complaint types shows the 70 types with at least 40,000 complaints. For each, its position is the type's complaints at addresses in the slice divided by all complaints of that type, including those with no street address. Types that 311 mostly logs without an address, such as traffic signal problems, sit near zero at every slice for that reason.
The map shows the top 5% (65,055 addresses). Each address takes the coordinates from the quarter in which it had the most located complaints. Averaging across quarters would have placed some addresses between two geocoded spots: for 2,290 of the 65,055, the quarterly points sit more than 100 meters apart. One address has no coordinates and is not drawn. A dot's color is the category with the most complaints at that address; its area grows with the square root of its complaint count.
For each of the 22 addresses pulled record by record, complaints are sorted by created time and these are measured:
The timing describes when 311 logged a request. A phone that saves requests offline and sends them later could log several in the same second, so same-second pairs are not treated as evidence on their own. Nothing in the public data identifies who filed a complaint or how, beyond the channel field, so the page describes the timing and does not name a cause.
A third pull counted every complaint by complaint type, status and resolution_description, the note the agency writes when it closes a complaint: 28 quarterly queries citywide, and 33 queries covering every spelling of the 1,301 busiest addresses (the top 0.1%). Both reconcile with the counts above: 1,147,933 complaints at the top 130 addresses and 2,619,702 at the top 1,301.
The status field does not help: 99.3% of complaints at the 130 busiest addresses and 97.8% everywhere else are marked closed. The city used 1,258 different closing notes in the period. Each note was sorted into one outcome by regular-expression rules applied in this order, first match wins; anything not closed is "still open":
| Outcome | A note goes here if it matches (case ignored) |
|---|---|
| Closed as a duplicate | duplicate|earlier complaint|already (been )?(reported|filed|submitted)|previously filed|similar complaint|same (location|condition|issue)|under another service request |
| Agency could not get in or get details | unable to gain|not able to gain|could not gain|couldn.t gain|unable to complete|insufficient|can ?not be processed|could not be processed|unable to access|could not access|not able to (inspect|access)|no one (was )?(home|available)|did not provide |
| Agency found nothing to act on | no evidence|no violations? (were|was) (issued|found|observed)|did not violate|were gone|not necessary|no further action|could not find|couldn.t find|found no|no condition|not found|was found to be|temporary condition|no work is necessary|meets (its )?standards|valid permit|did not observe|didn.t observe|no action (is|was) (needed|necessary|required)|not observed|unfounded|no encampment|no longer (there|present)|gone on arrival|was not able to (find|locate)|could not locate|unable to (find|locate)|no violation|no criminal violation|did not find|not warranted |
| Fixed by the time the agency checked | (was|were|been|has been|had been) (corrected|restored|resolved)|condition was corrected|claimed the vehicle|vehicle was moved|has been moved|indicated that the condition |
| Closed with no clear outcome | keep the location under observation |
| Agency took action or found a violation | took action|summons|arrest|violations? (were|was|has been|have been) (issued|previously issued)|notice of violation|found violations|repaired|corrected the (condition|problem)|completed the request|completed the requested work|collected|cleaned|removed|towed|hearing was scheduled|issued an? |ticket|fined|abated|fixed the|made repairs|work was completed|the problem has been addressed|installed|replaced|restored service|took enforcement|shut the|performed the work|addressed the issue|issue you reported was addressed|offered services |
| Referred elsewhere or not the agency's job | jurisdiction|referred|forwarded|transferred|has requested the |
| Closed with no clear outcome | everything else, including blank notes |
Every note, its outcome and its count are in data/outcome-rules.csv. Citywide, 15.7% of complaints got a note with no clear outcome.
Comparison. "Everywhere else" is every complaint outside the top 0.1% of addresses, found by subtracting the top addresses' counts from the citywide counts, type by type. The expected share of an outcome at the busiest addresses is the sum, over complaint types, of that type's count at the busiest addresses times the outcome's share for that type everywhere else, divided by the busiest addresses' total. This removes the effect of the busiest addresses having a different mix of complaint types; it does not remove any other difference.
Per address. The same calculation, run for each of the 1,301 busiest addresses, gives its own expected action share. 103 of the top 130 fall below theirs and 66 fall below half of it.
Batch closings. For the 22 addresses pulled record by record, complaints are grouped by closed_date to the second; a batch is 10 or more complaints closed in the same second. The share of complaints in batches is 24.8% at 655 East 230th Street, 25% at 661 East 230th Street and 13.3% at 672 East 231st Street, and zero at 1465 Washington Avenue, 177 Sands Street and 1869 83rd Street. Heat complaints are often closed in large batches when one building-wide complaint covers many apartments, so batches are not compared for heat.
Not known. The notes are the agencies' own words. The data has no record of visits, so it cannot show whether each complaint got its own response, whether the condition was real or what "took action" consisted of.
scripts/fetch_311.py pulls the quarterly aggregates. scripts/build_data.py builds data/summary.json, data/points.json, data/top-5-percent-addresses.csv and build/checks.json.scripts/fetch_records.py and scripts/analyze_cadence.py pull the individual records and build data/cadence.json.scripts/build_docs.py writes this page and the build log from those files.