Blog
My Anti-Snoring Device Is Time-Zone Blind: 21 Nights of Logs Across a Three-Hour Shift
I install museum exhibitions for a living. Ten days in Tucson, Arizona, building a Sonoran desert show — eight hours a day of casework, mount-making, and dust — and then a red-eye home to Brooklyn. Arizona does not observe daylight saving. In the last week of September, that makes it exactly three hours behind New York.
I wore the same anti-snoring device every night for twenty-one nights: three at home before I left, ten while I was away, eight after I landed. I logged four things by hand every morning — lights out, hours in bed, how long it took to fall asleep, and a throat-dryness score from 1 to 5 — and let the device log everything else.
This post is the table. Not much happens in it. That’s the point, because the table is where I found something I had been getting wrong for a year: the device is time-zone blind, and so was I.
Method, Such As It Is
One device, one phone, one Notes file. Every morning, before coffee: app export screenshot, four handwritten fields, done. No wearables, no sleep-lab anything. If a night is missing from the table it’s because I didn’t wear it, and I’d rather show the gap than interpolate over it.
Two definitions, because the rest of this post depends on them:
- Clock-indexed — what the app gives you. Snoring minutes per calendar night.
- Re-indexed — snoring minutes per hour actually in bed, which is the only version you can compare across a trip where some nights were five hours and one was nine.
Twenty-One Nights
| # | Date | Where | Lights out (local) | In bed | Snoring (min) | Min / hr in bed | Dryness (1–5) | Worn | Note |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Sep 24 | Home, Brooklyn | 23:10 | 7h20 | 24 | 3.3 | 2 | Yes | Baseline. Started logging so I’d have something to argue with. |
| 2 | Sep 25 | Home | 22:50 | 7h35 | 27 | 3.6 | 2 | Yes | Second baseline. |
| 3 | Sep 26 | Home | 23:25 | 7h05 | 22 | 3.1 | 2 | Yes | Packed in forty minutes. Didn’t pack the humidifier. |
| 4 | Sep 28 | Tucson, hotel | 00:20 | 5h40 | 31 | 5.5 | 4 | Yes | DFW mechanical delay, 4h20 in a gate area. Local 00:20 = body clock 03:20. |
| 5 | Sep 29 | Tucson, hotel | 23:40 | 6h50 | 34 | 5.0 | 4 | Yes | Install day one: eleven hours on my feet, 190 crates. |
| 6 | Sep 30 | Tucson, hotel | 23:10 | 7h00 | 21 | 3.0 | 5 | Yes | 38°C, 11% relative humidity. Throat raw by 3 a.m. |
| 7 | Oct 1 | Tucson, hotel | 22:50 | 7h20 | 19 | 2.6 | 5 | Yes | $14 drugstore humidifier. First night it didn’t hurt to swallow. |
| 8 | Oct 2 | Tucson, hotel | 23:30 | 6h40 | 26 | 3.9 | 3 | Yes | Ceiling AC aimed at the pillow. Dragged the mattress 60 cm. |
| 9 | Oct 3 | Tucson, hotel | 00:05 | 6h10 | 28 | 4.5 | 3 | Yes | Museum donor dinner ran to eleven. Two beers, against my own advice. |
| 10 | Oct 4 | Tucson, hotel | 22:40 | 7h30 | 18 | 2.4 | 2 | Yes | Nine hours on a scissor lift. Slept like a dropped tool. |
| 11 | Oct 5 | Casita, Avra Valley | 01:10 | 5h50 | — | — | 4 | No | Off night. Chafing behind the jaw after eight straight nights, throat still raw. Roughly 4h50 of actual sleep, woke twice. |
| 12 | Oct 6 | Casita | 23:20 | 6h30 | 29 | 4.5 | 3 | Yes | Load-out until nine. Swamp cooler cycled every four minutes. |
| 13 | Oct 7 | Seat 24D, TUS–PHX–JFK | 00:40 | 4h10 | 26 | 6.2 | 5 | Yes | Worn on the plane. Four wakeups. JFK at 06:10. |
| 14 | Oct 8 | Home | 21:40 | 9h10 | 33 | 3.6 | 4 | Yes | Door at 08:20 after the train. Crashed. Woke at 06:50 feeling like a bad photocopy. |
| 15 | Oct 9 | Home | 01:20 | 6h00 | 38 | 6.3 | 4 | Yes | Clock says 22:20, body says 19:00. Lay there 55 minutes. |
| 16 | Oct 10 | Home | 00:50 | 6h20 | 41 | 6.5 | 3 | Yes | Worst number of the month. I felt fine. That’s the problem. |
| 17 | Oct 11 | Home | 23:40 | 6h50 | 36 | 5.3 | 3 | Yes | Three-hour nap at four in the afternoon. Self-inflicted. |
| 18 | Oct 12 | Home | 22:30 | 7h00 | 29 | 4.1 | 2 | Yes | First lights-out before 23:00 since September 26. |
| 19 | Oct 13 | Home | 23:00 | 7h15 | 26 | 3.6 | 2 | Yes | Woke once, back under in four minutes. |
| 20 | Oct 14 | Home | 22:45 | 7h25 | 25 | 3.4 | 2 | Yes | Unremarkable, in the good way. |
| 21 | Oct 15 | Home | 22:50 | 7h30 | 24 | 3.2 | 2 | Yes | Back to baseline eight nights after landing. Kept logging two more nights (23, 22) and stopped. |
Twenty-one nights, twenty-two records in the app, one of them a zero. Hold that thought.
Where the Number Lied
Three windows, same device, same person, same pillow. Left column is what the app told me. Right column is what the week actually was.
| Window | Nights | Avg snoring (min/night) | Avg snoring (min/hr in bed) | Avg hours in bed | Avg time to fall asleep | Avg dryness | What the app’s weekly view implied | What it was actually like |
|---|---|---|---|---|---|---|---|---|
| Home, before (Sep 24–26) | 3 | 24.3 | 3.3 | 7.33 | 15 min | 2.0 | Baseline | Baseline, and slightly nervous about the trip |
| Away (Sep 28–Oct 7) | 10 | 25.8 | 4.0 | 6.37 | 40 min | 3.5 | “Travel had almost no effect — plus 6%.” | Wrecked. Six-and-a-quarter hours in bed, forty minutes to fall asleep, a throat like sandpaper |
| Home, after (Oct 8–15) | 8 | 31.5 | 4.4 | 7.19 | 22 min | 2.75 | “Something got worse at home — plus 30%.” | The best week of the month. Sleeping 7h11, falling asleep in 22 minutes, throat fine |
The app ranked those weeks in exactly the wrong order. The away week, which flattened me, came out 6% worse than baseline. The recovery week, which felt like getting my life back, came out 30% worse.
Two reasons, and neither is a malfunction:
Minutes per night is a numerator without a denominator. On September 30 I was in bed for seven hours and snored for twenty-one minutes. On October 10 I was in bed for six hours twenty and snored for forty-one. As a per-night figure the second night looks like a crisis. Per hour in bed it’s 3.0 against 6.5 — and the second night’s real problem wasn’t snoring at all, it was that I’d gone to bed at 00:50 with a body clock still on Mountain Time. Per-hour numbers don’t fix everything, but they stop you comparing a five-hour night to a nine-hour one and calling it progress.
And the smaller, nastier one: a night you don’t wear the device is a night the device records as zero. October 5 sits in my app as the quietest night of the trip. I didn’t wear it. My jaw was chafed, my throat hurt, and I slept four hours fifty. If I had trusted the weekly average that week, I would have concluded that the night I skipped was the best night I had — and taken the device off more often.
The Three-Hour Problem, in One Table
Here’s the thing I’d never had to think about before, because I’d never moved my bedtime three hours sideways while logging it.
The app decides what “last night” means using a clock window it learns from your recent pattern. Going west, your body is three hours ahead of the wall, so you go to bed later by the wall and your sleep straddles midnight more often. Going east, the reverse: your sleep window walks backwards into the previous evening.
| Night home | Lights out | Drift vs my 22:50 target | Minutes to fall asleep | What the drift did to the record |
|---|---|---|---|---|
| Oct 8 | 21:40 | −1h10 | 6 | Clean single record; 9h10 in bed, which the app scored as my worst night of the week |
| Oct 9 | 01:20 | +2h30 | 55 | Window opened before I was in bed; the first 40 minutes of the record are me lying awake |
| Oct 10 | 00:50 | +2h00 | 40 | Same distortion, larger; the 41 minutes are real but the 6h20 is the actual story |
| Oct 11 | 23:40 | +0h50 | 25 | Nap at 16:00 put two sleeps in one 24-hour window |
| Oct 12 | 22:30 | −0h20 | 15 | First record of the month that lines up with a normal night |
| Oct 13 | 23:00 | +0h10 | 12 | Lag closed on night six, not night two |
| Oct 14 | 22:45 | −0h05 | 10 | Boring. Finally. |
Six nights to close a three-hour gap, going east, with a nine-hour crash night in the middle that felt helpful and wasn’t. The red-eye is the visible case: 00:40 lights out in a seat, four hours ten of on-and-off sleep, and the app split it into two records because the sleep crossed midnight. That’s how ten nights away became eleven records, and why my “weekly average” that week was computed over a set that included half a flight.
Four Things I Changed
- Re-anchored the log to lights-out instead of the clock. I now write “hours since lights out” as the primary axis, in two-hour buckets, and only convert to clock time at the end. It took one trip to see that every comparison I’d ever made was secretly a comparison of bedtimes.
- One sleep, one record. If lights-out and wake-up land on different calendar dates, I merge them by hand in the notes. Tedious, two minutes, and it’s the difference between a real weekly average and a decorative one.
- Moved the device’s night window before I moved myself. Thirty minutes before local bedtime on the first two nights away, then let it relearn. Going west it wants to start too early; going east, too late. It catches up on its own in about four nights — but those four nights are the ones you’re reading the data from.
- Humidity beat everything else in my bag. Eleven percent relative humidity is not a thing I’d ever thought about. A $14 drugstore humidifier plus a wet towel hung over the AC return grille took my dryness score from 5 to 2–3 in two nights, and the snoring minutes followed. Nothing about the device changed on those nights. The room did.
What This Doesn’t Prove
Three baseline nights is not a baseline, it’s an anecdote with good posture. Ten days of physical labor in dry heat confounds everything — I was more tired than I’ve been all year, which should have made my sleep better and my snoring worse, and I can’t separate those two effects. The trip also had a donor dinner with two beers in it, and beer is the single most reliable way to make any of this data meaningless.
What I’m reasonably confident about is narrower and more useful: the device measured what it measures, accurately, on every night I wore it. What it cannot know is how long I was in bed, how long it took me to get there, or what time zone my body thought it was in. Those three things explained my month better than the snoring minutes did, and they cost nothing to write down.
I still wear it. I just stopped letting a number that doesn’t know what time it is tell me how I slept.
One person’s logs, one device, three weeks. Results vary between people and between nights, and nothing here is medical advice. If snoring comes with pauses in breathing, daytime sleepiness, or morning headaches, that’s a conversation for a clinician, not a blog post — and a wearable device is not a diagnosis.