A research page is frozen the day it is published.
Every other page on this site re-queries its figures each time the site is published. A research page does not. Its numbers are pulled once, written to a file, and served unchanged until someone republishes the page deliberately. 9 of the 17 figures this site tracks against the current record sit on the two studies below, and each one carries the SQL that produced it, the date it was published, and the size of a change that would count as drift.
The reason is what a citation is for. A reader who quotes a number in a filing, an article or a brief needs that number to still say the same thing when someone checks it next month. A figure that is re-queried on every publish moves under the reader, and a figure that moves is not quotable however recent it is. So the two kinds of page on this site do opposite jobs on purpose: a guide and the borough record are as current as the last publish, and a study holds still.
See what is open on one building.
NYC heat violations spike every October.
Heat season starts October 1, and the violation record turns with it rather than after it. September records 5,070 heat and hot water violations against October's 10,419, and January runs 5.8 times September's volume. The page carries the month-by-month shape, the same September-to-October comparison repeated for every year separately, and a control group of violation categories with no seasonal rule, which do not move. 4 of its figures are in the drift register.
The standard is 24 hours. The measured median is not.
New York City gives a landlord 24 hours to restore heat on an immediately hazardous violation. Of the 148,225 Class C heat and hot water violations on HPD's record, 72,598 have ever been certified corrected, which is 49.0%. The median certified case took 9 days, and 47,488 have stood open more than a year with nothing filed. The page measures the gap year by year and shows the certification rate has not drifted with volume. 5 of its figures are in the drift register.
What freezing costs, and what happens to a stale figure
A frozen number goes out of date. The record it was drawn from keeps moving, and the honest version of that is to measure the gap rather than hide it. Every figure on both studies is entered in a register held in this site's own source, alongside the query that produced it and a threshold for how far it may move before the gap matters. A weekly job re-checks all 17 entries in that register against the record as it stands that day and reports what has moved.
That job reports. It does not rewrite either the page or the register, because a published figure that has moved is a decision for a person: republish the study with a new date, or leave it and say what it was measured on. Both studies state their own as-of date on the page, and the first of these figures was published August 22, 2026. The underlying pull covers inspections recorded July 30, 2023 through August 27, 2026.
What a study here is not
Both pages are aggregates over HPD's violation record, counted by violation. No building, owner, tenant or address is named on either one, and none is named here. A median is the middle case: half the certified violations in that group cleared faster and half slower. Every percentage is stated with the group it is a percentage of, because the share ever certified and the share still open past a year are different questions with different denominators.
HPD's own file lags the inspector by days to weeks, so the most recent cases are under-counted in any measurement taken from it, including these. Neither study claims to explain why a number is what it is. Each one measures what the record says against a standard the city set, and stops there. The queries behind any figure are available on request at press@violationscout.com.
Where this sits
A status name on its own is decoded in the status glossary. How long things take and how often they happen is the guides, re-queried every publish. The current record by borough and category is the borough grid, and citywide it is the record in time. How HPD, DOB and OATH/ECB divide the work is on the agencies page. What this site holds and how it reads it is the method.
Aggregates only; no building or owner is named on this page. Source: NYC Open Data, our own working copy, pulled August 29, 2026. Citywide, ViolationScout.
Every figure above has a query behind it. Cite the page and its as-of date. Ask and we will send the query.
Ask for the query