The 4.1-day average that leaves out the 839 still open
Showing the misleading chart
A monthly review slide we drew from NYC 311 data says elevator complaints now close in four days, down 88% in a year: the average time to close falls from 33.6 days for complaints filed in November 2025 to 4.1 for October 2026. But the average only counts complaints that have closed, and 839 of October’s 1,528 haven’t closed yet. Every month in the year read about four days when it was new. On a fixed 14-day clock, October is the slowest month of the twelve.
01The claim
Elevator complaints in New York now close in four days, down 88% in a year. A monthly review slide of the kind an operations desk might circulate, which we drew from the city’s public 311 file: average days from filing to closure for every 311 complaint of type “Elevator”, grouped by the month it was filed. Complaints filed in November 2025 took 33.6 days on average; complaints filed in October 2026 took 4.1. The average fell in nine of the eleven months between, the four newest months are the four fastest, and the axis starts at zero with every value printed.
02The trick
The average only counts complaints that have closed, and a complaint can only close after it has been filed. The slide was drawn on the last day of the file, 29 October 2026, so the October bar is an average over the 689 complaints that closed within days of filing; the 839 that hadn’t closed yet (status Open or Assigned) are not in the bar at all, and 620 of them are already older than the 4.1 days the bar reports. September is missing 424 of its 2,014 and August 127 of its 2,646, so the right-hand end of the chart is the shape of a deadline rather than a trend. You can watch it happen in the same file. Recompute each month’s average using only the complaints that had closed by the end of the month they were filed in, and every month in the year reads between 3.3 and 5.0 days: November 2025 read 3.5 on 1 December and has climbed to 33.6 since, as its slow complaints closed one by one over the following months. October 2026 is not unusually fast. It is the only month that hasn’t had time to be slow.
03The fix
Give every month the same clock. The redraw plots, for each month, the share of complaints closed within 14 days of filing, counting the ones still open as not closed, so a month with a long queue scores lower instead of dropping its queue from the sum. On that clock the year is roughly level, between 57.4% and 67.7% from November to September, and October — counting only the 736 complaints filed by 15 October, the ones that have had 14 days — is the slowest of the twelve at 49.0%; compared like for like, the first half of every other month scored between 56.8% and 65.9%. Beside it, each month’s average is drawn twice, at its month-end reading and at today’s, so the reader can see every month start near four days and climb. There was real improvement in the year: on the same 30-day clock the share closed rose from 65.1% (November) to 77.8% (July), and within 90 days from 86.1% to 98.0%. The fall in the plain average from 33.6 to 16.8 overstates even that, because July’s slowest complaints haven’t had as long to close, and the fall from 16.8 to 4.1 after July is mostly the clock. A closed-cases-only average is fine for a month that has finished closing; for the newest months, print the still-open count beside it, or don’t print it. (The month-end readings are reconstructed from closing dates and assume they are not revised later; closing dates are recorded to the day, and requests with status Assigned are counted as open. Closure is the status of the 311 request, not a record that an elevator was repaired.)