The 82% fall in fly citations that happens every winter

Dark navy review slide, kicker “Supervisor training desk · Fly-control refresher · Six-month review · New York City”, headlined “Fly citations fell 82% after supervisors took the refresher — 19.9% to 3.5%”. A subhead says the refresher, covering drain-line cleaning, door sweeps, and waste-room and grease-trap routines, went out to kitchen supervisors in October 2025, and that the chart shows the share of New York City health inspections that cited filth flies every month from the last month before the course through February, straight from the city’s inspection data, every routine initial inspection counted, on a zero-based axis. A badge reads “Demonstration · drawn by us from public data”. The main panel, titled “Initial inspections citing filth flies (violation 04N) · share of routine initial inspections, by month”, is a gold line on an axis from 0 to 25%, with a dashed vertical marker labelled “Refresher rolls out · Oct 2025” just after the first point: September 2025 19.9%, October 18.3%, November 17.4%, December 9.5%, January 2026 5.8%, February 3.5%. To the right, a big gold figure “−82%” is captioned as fly citations falling from 19.9% of initial inspections in September to 3.5% in February, five months of decline in a row; a read-out says every month since the refresher went out came in lower than the month before, and that by February fewer than one inspection in 28 found flies against one in five the month before the course; a recommendation says to make the refresher mandatory for every new kitchen supervisor and roll it into the spring onboarding cycle; and a table lists each month’s initial inspections, the number citing flies and the share, from September 2025 (1,139 inspections, 227 citing flies) to February 2026 (1,134 and 40). The footer says the figures are the share of routine (cycle) initial inspections with at least one violation 04N citation, from the city’s open data, and that the slide is a demonstration in which the desk, the refresher and the recommendation are invented and every inspection count is real. — Showing the misleading chart

Cream-background chart headlined “Fly citations collapse every winter, course or no course”. A subhead says the slide’s six months, in gold, are one turn of a cycle that runs every year: flies are a warm-weather pest, so citations climb through summer and autumn and collapse by February; the same September-to-February fall was 76% the winter before and 63% the one before that, with no refresher; and six months after the slide, August 2026 reached 24.1%, higher than any month in 2024 or 2025. The main panel is one blue line of monthly shares from January 2023 to September 2026 on a zero-based axis to 30%, with September-to-February bands shaded. It is a sawtooth: peaks of about 20% to 27% each summer and autumn, troughs of about 3.5% to 7% each late winter and spring. Each winter carries a label: “−63%, Sep → Feb, no course” over 2023–24, “−76%, Sep → Feb, no course” over 2024–25, and, in gold, “−82%, Sep → Feb, the slide” over 2025–26, where the slide’s six months are drawn as a thick gold segment. The line then climbs again to 24.1% in August 2026. Months with fewer than 400 initial inspections, mostly in 2023, are drawn as hollow dots. A side panel, “Same month, a year earlier”, plots lollipops of the change in percentage points for each month from April 2025 to September 2026 against the same month a year before, with a dashed gold line between September and October 2025 marking when the course went out. Above the line, before the course, April 2025 is up 0.6 points and May to September 2025 are already 2.1 to 5.7 points lower than in 2024 (June 18.7 to 13.0). Below it, blue lollipops run 1.1 to 3.9 points lower from October to June (for example February 5.2 to 3.5), then rust lollipops point the other way for July (17.1 to 20.5), August (18.3 to 24.1) and September (19.9 to 22.4). A card says that month for month the winter after the course ran 3.7 points below the one before, 9.6% of October–March inspections against 13.4%, but that April–September 2025 had already run 2.4 points below 2024 before any course, and the next summer ran higher, so what is left is about a point. Another card says a test of the course needs a comparison that shares the calendar: kitchens whose supervisors took it against kitchens whose supervisors had not yet, a random half going first, over the same months and again in July. The footer notes that the city’s file keeps only restaurants still active on the day it is pulled, going back three years before each one’s latest inspection, and drops older records of restaurants inspected recently, so older months hold far fewer inspections and are not a like-for-like base, and that the −63% starts from a thin month of 216 inspections. — Showing the misleading chart

Be the first to star this exhibit

A training-review slide we drew from New York City’s restaurant inspection data: after a fly-control refresher goes out to kitchen supervisors in October 2025, the share of initial inspections citing filth flies falls from 19.9% to 3.5% by February. The same September-to-February fall was 76% the winter before and 63% the winter before that, with no course, and by August 2026 the rate was back up to 24.1%.

NYC Department of Health and Mental Hygiene, “DOHMH New York City Restaurant Inspection Results” (NYC Open Data, 43nn-pn8j) — initial cycle inspections citing violation 04N, January 2023 to September 2026, retrieved 6 October 2026, 2026Added October 6, 2026

01The claim

The refresher worked. A six-month review slide of the kind a training desk might circulate, which we drew from New York City’s public inspection records: in October 2025 a fly-control refresher (drain lines, door sweeps, waste rooms) goes out to kitchen supervisors, and the share of the city’s routine initial restaurant inspections that cite filth flies falls every month afterwards, from 19.9% in September to 3.5% in February. That is down 82%, on a zero-based axis, with every month’s inspection count printed beside it. Recommendation: make it mandatory for every new supervisor.

02The trick

Every number on the slide is right. What goes wrong is the calendar. Flies are a warm-weather pest, so fly citations in New York climb through summer and autumn and collapse over the winter, every year, whatever anyone teaches. It isn’t inspectors easing off in the cold: in the same inspections, mouse citations don’t fall over the winter, while flies and roaches do. A course that goes out in October and is judged in February is being credited with the season. The slide’s window starts near the top of the cycle and ends at the bottom, and a before-and-after comparison with nothing beside it can’t tell a course from a calendar. The city’s file shows the same fall without a course: September to February dropped 76% in 2024–25 (22.3% to 5.2%) and 63% in 2023–24 (24.1% to 9.0%). The slide’s 82% is a relative fall, 16.4 percentage points, and it looks a little steeper than the others mostly because February 2026 was the lowest month on the chart. Six months after the slide, in August 2026, 24.1% of routine initial inspections cited flies, more than in September 2025, the month before the course went out.

03The fix

Hold the season still before reading the change. The cheapest way is to compare each month with the same month a year earlier. On that view the winter after the course did run lower: 9.6% of October–March routine initial inspections cited flies, against 13.4% the winter before, a gap of 3.7 points. But the gap was there before the course existed. April–September 2025 had already run 2.4 points below the same months of 2024, and July, August and September 2026 came in 2.4 to 5.8 points above their twins. What is left for the course is about a point, and the city’s file can’t vouch for even that. It keeps only restaurants that are still open, and it drops older records of restaurants inspected recently, so earlier months hold far fewer inspections and are not a like-for-like base. A real test of the course needs a comparison that lives through the same months: kitchens whose supervisors took it against kitchens whose supervisors had not yet, ideally with a random half going first, judged winter and summer alike.