The State of the NYC Restroom
We put sensors on the door and watched. Here is the first ground-truth look at how a city actually goes.
The public restroom is the most-used, least-understood room in any building. Managers schedule cleanings by the clock, not by traffic. Owners hear about problems only when someone complains. The data that would fix this has never existed — because no one was counting.
Between December 2025 and March 2026, Yedi counted. We put smart locks on restroom doors in cafés, pizzerias, bakeries and fast-food counters around New York, and logged every time a door opened. In four months we recorded more than 117,000 unlocks across 58 restrooms — the first ground-truth portrait of how heavily this invisible room actually gets used.
Two numbers tell the story. A busy restroom is unlocked ~1,200 times a month — staff, keypad codes, and customers, all day long. And 2,763 of those unlocks were people who paid to get into a room the world assumes is free. Scale and proof-of-demand, in the same dataset.
Here's what those unlocks reveal.
Dec 2025 – Mar 2026
A busy restroom opens ~40 times a day
The paid number — 2,763 — is a sliver of the story. A restroom door opens all day long: staff, cleaners, keypad codes, customers who already have access. Measured directly at the lock, a busy Yedi restroom is unlocked around 1,200 times a month — roughly 40 times a day, from a single door. Across an actively-used fleet, that compounds quickly.
Each mark is a door; filled = actively used. Only the active ones set the pace — dead installs drag any fleet-wide average down, which is why the busy-restroom rate is the honest benchmark.
Where New York goes
The busiest doors read like a map of where the city actually stops — coffee shops, pizza counters, halal carts, bagel spots. A restroom in a busy location runs 1,000 to 1,900 unlocks a month, and one major coffee chain (deep green) fields more top-ranked restrooms than anyone — six of the top twelve, across locations all over the city.
Physical unlocks per month, measured at the lock. One coffee chain's locations shown in deep green. Venue names withheld.
Everything above counts every unlock — staff, keypad codes, and customers alike. The patterns below are drawn from paid unlocks only: the 2,763 times someone chose Yedi and paid to get in. Among payers, the behavior is sharper — and it's a different signal than raw traffic.
Paying customers go on the weekend
Paid visits aren't flat across the week — they swing. Saturday is the busiest day for paying customers by a wide margin, with Sunday close behind, bottoming out mid-week. (Total physical traffic is far steadier — staff and regulars use the door evenly all week; it's the payers who spike on weekends.)
Among paying customers, weekends are 39% of all paid visits and Saturday runs 108% above the quietest weekday. (Physical traffic swings far less — Saturday is only ~35% above the low.)
An afternoon rush, not a morning one
The demand curve builds through late morning, peaks in the early afternoon (around 1–2pm ET), and holds high straight through the dinner hour before falling off. The quietest stretch is 3–7am, when the network nearly flatlines. This is the shape of a commercial district living its day.
Paid visits across the network. The pre-dawn floor is near zero — restrooms sleep when the city does.
The average visit is 3 minutes, 42 seconds
Measured directly from the lock — door-close to door-open — the typical visit runs under four minutes. The median is even shorter, at 2 minutes 52 seconds, meaning most people are quicker than the average and a long tail pulls it up. It's the kind of number a cleaning schedule has never been able to see.
Door-open time from early lock telemetry — 210 de-duplicated sessions since late May. Most stays run 1 to 4 minutes.
Demand doesn't split down the middle
At the one venue running separate instrumented men's and women's rooms — a busy NYC coffee shop — the men's room took 58% of paid visits to the women's 42% over the same window — a 36% gap. A single data point, but a provocative one: the assumption that restroom load is symmetric is exactly the kind of thing operators guess at and never measure.
Two doors, one location, same window. The gap is real and consistent — and invisible without instrumentation.
This is a habit, not a novelty
The easy dismissal of any new system is that people try it once. That's not what the data shows. 22 restrooms have logged 50 or more paid visits each — and together they account for 78% of all paid visits. Demand concentrates in places people return to, week after week.
The busiest day for paying customers was Valentine's Day.
The single busiest day of the year for paying customers. (By raw physical traffic, the busiest days were Saturdays in March.) Whatever New Yorkers were doing on February 14th, a lot of it ran through a restroom.
Facilities have been flying blind. This is what the ground looks like.
Every number here was invisible a year ago — 117,000 unlocks in four months, the weekend swing, the afternoon peak, the 3-minute stay, and the 2,763 of those unlocks that were people paying to walk through a door that used to be free. That's the point. When you can see demand, you can clean on demand: staff the busy rooms, stop over-servicing the quiet ones, and prove the work got done. Vol. 01 is a first look at a few dozen doors. We're just getting started.
Want to see your own restroom's numbers? Book 10 minutes and we'll send you a Yedi sign — on us.
Get a free sign →Methodology & notes
Unlock volume. Unlock counts are physical door-opens recorded at the lock. Between December 5, 2025 and March 30, 2026, Yedi's network logged more than 117,000 unlocks across 58 restrooms. We later downgraded the dedicated logging server to reduce infrastructure cost and archived its data, so the rich physical log covers this window; the current pipeline records access events but not every state change. To characterize ongoing volume we (1) normalized to a per-day rate using the three complete months (January–March 2026), and (2) grossed the result up by 25% to account for opens the sensors did not capture (estimated ~80% capture). On that basis a busy restroom runs ~1,200 unlocks/month and a fleet at that rate ~90,000/month (~1 million/year). These are estimates.
Paid unlocks. The 2,763 figure is an exact, server-side count of successful payment-authorized accesses (January–July 2026) — 2,743 tied to a unique customer payment, against 10 failed attempts (99.6% first-try). Because these are pay-to-enter doors, a paid unlock and a payment are the same event; this is the willingness-to-pay number, and a count, not an estimate.
Which cut is which. The headline volume, the busiest-restrooms ranking, and time-inside (3:42 avg / 2:52 median, from 210 lock sessions) are physical unlocks. The day-of-week, hour-of-day, men's/women's split, concentration, and busiest-day (Valentine's) cuts come from the paid access log and describe paying-customer behavior specifically — physical traffic is steadier and its busiest days were Saturdays in March. Times are Eastern; an April 2026 logging gap excludes that month from time-series views. The men's/women's comparison reflects a single venue with separately instrumented rooms, offered as illustration, not a citywide claim. Individual venue names are withheld; locations are described by category only.