I Read 34 Anti-Snoring Device Logs Last Quarter. The Ones That Improved First Weren’t the Wearers.

Last month one of my patients emailed me a photograph instead of her nightly log. It was a sheet of paper that had been pushed under her apartment door — pencil, on the back of an electricity bill, six lines long. She had photographed it on her kitchen table next to a coffee cup, and she wrote in the subject line: “I think this is the result you’ve been asking me for.”

I am a respiratory therapist. I run a Thursday evening device follow-up clinic, which is a polite way of saying that people who buy a neck-worn anti-snoring device on their own send me their exported data and I tell them what it says. I do not wear one. I have never slept in the home of anyone whose log I have read. My entire relationship with these devices is a folder of spreadsheets and the sound of someone sighing on the other end of a phone call.

Last quarter the folder held thirty-four active logs. This post is what I found in nine of them, why the finding surprised me, and the one number in the whole file that I now trust more than any bar chart an app will ever draw for you.

I should say up front: I don’t sell these devices. My clinic hasn’t sold anything since the 2019 brochure reprint, and I have no financial interest in whether you keep yours. What I have is a lot of other people’s nights.

What was actually in the file, and what I ignore

Patients send me one of three things: a CSV exported from the app, a screenshot of a weekly summary, or — more often than you would think — a personal message written at 6 a.m. with the grammar of someone who has not slept. The device itself is the same family of product in almost every case: a soft band around the throat, a small pod that sits under the chin, a phone app that counts snoring minutes and rates your night out of a hundred.

There are three numbers in those exports I will argue about. There are two I ignore completely.

What I trust:

• Contact hours. How long the band was actually on your neck, not how long you were in bed. This is the closest thing the file has to an honesty test, because nobody accidentally wears a band for seven hours.
• Snoring minutes. Crude, microphone-based, and it misses quiet open-mouth breathing entirely — but it is consistent, and consistency is all a trend needs.
• The minute you took it off. Almost every export timestamps the removal. Patients never mention this number. It is the most informative number in the file.

What I ignore: the sleep score out of a hundred, and the breathing-quality bar. I have watched a man’s score drop eleven points on a night when his snoring minutes also dropped by half, because the app decided his “sleep cycles” looked irregular. Those two graphics are a mood ring with a battery. If they help you keep wearing the thing, fine. Do not make decisions with them.

The nine thin walls

The reason I pulled these nine files out of thirty-four is not clinical. It is architectural. All nine people live in older buildings — pre-1960 brick, wood-frame, converted attic, a 1970s slab tower — where the bedroom wall is shared and thin. Four of them had received a complaint. Not a hint, not a joke at the mailbox: something written down, from a neighbor, a landlord, or a building manager.

Nine people is not a study. There is no control group, the data is self-reported, and I am one therapist with a laptop. Treat this as a file review, not evidence.

*Baseline = mean snoring minutes per night across the three nights before first use, from the device’s own microphone. All snoring values are nightly means for that period, not best nights. Baseline for M.R. and P.S. includes one night they described as “a bad one,” which I kept in, because bad nights are the ones the neighbors hear.

The average person in this table cut their snoring time by about 52 percent over twelve weeks. That is a good result for a band with no hose and no prescription. It is also the least interesting thing I found.

The finding I did not expect: the wall improves first

In the four households where someone had complained in writing, I asked for one extra piece of data that no app collects: how many nights a week the neighbor was woken, before and after. Four people went and asked their neighbors a question most of us would rather not ask. Two of them got a number. One got a count kept in a diary, which I did not expect and appreciated enormously.

Look at the two columns on the right. By the middle of week two, three of these four neighbors were already sleeping noticeably better. The wearers themselves were not. At week two, three of the four still rated their own mornings a 2 out of 5, and one rated it a 1. T.A., who improved the most, wrote “I feel like I’m being strangled by a collar” on day eleven, at a point when his neighbor had already stopped hearing him.

The explanation is not mysterious once you say it out loud. A wall does not care how the band feels. It only hears decibels, and the device starts cutting decibels on night one. A person, on the other hand, has to get used to a foreign object under the jaw, a strap that loosens by two centimeters over four hours, a pod that warms up against the throat, and the small plastic click when you turn your head. That adjustment takes three to five weeks.

So the device works on the neighbor’s timeline long before it works on yours. Which means that if you are the one wearing it, the first two weeks are the worst possible time to decide whether it is working — and that is exactly when most people decide.

Case one: M., 44, and the room on the other side of the nursery

M.R. is the penciled note. She is forty-four, works early shifts at a bakery, and lives on the third floor of a 1930s walk-up where the bedroom wall is plaster over brick and the other side of it is a nursery. Her first nine nights were, in her words, “a disaster I paid for.”

Night one she had the band too tight — a red line across the front of her throat that was still there at 6 a.m., and a headache behind her ears by mid-morning. Night two she overcorrected, and the pod slid down until it was resting on her collarbone, where it did nothing except click against her necklace clasp. Nights three through six she took it off between 2 and 3 a.m. every night: contact hours of two hours ten minutes, two hours forty, one hour fifty, three hours five. The app still reported snoring minutes for those nights, because it reports whatever the microphone heard, including the two hours after the band was on the nightstand. This is the single most common way people fool themselves with these exports.

What changed on night seven was not the device. It was that she stopped trying to wear it like a scarf. She moved the pod so the flat edge sat just under the jawline rather than over the windpipe, loosened the strap by one notch, and — this is the detail I now repeat to everyone — started putting it on twenty minutes before getting into bed, while she read, instead of in the dark. Contact hours went from an average of two and a half to six and a half by the end of week two.

Her neighbor’s improvement showed up on day nine, in pencil. M.’s own rating out of five did not reach a 4 until week six.

The thing that nearly made her quit was not the noise and not the strap. It was the twelfth of the month, when she woke at 4:20 with the band pushed up under her chin and a raw patch the size of a coin on the side of her neck. Two nights off, a wash of the band, and a switch to a pillowcase that wasn’t cotton-blend fixed it. She has worn it 78 of 84 nights since.

Case two: six nights in a mill town

T.A. is thirty-seven, an industrial electrician, and in the middle of his twelfth week he was sent to a paper mill town for a six-day shutdown. He slept in a motel where the wall between rooms was, by his estimate, “a door and a promise.” His home numbers at week six were 33 minutes of snoring and seven hours five minutes of contact time. Here is the motel week.

His snoring went up 26 percent on the road and his contact time fell by more than an hour a night, and none of that was the device failing. It was a twelve-centimeter motel pillow tilting his chin toward his chest, a radiator that made him roll away from the wall and onto his back, and one evening of beer. The towel under the pillowcase — four centimeters, folded twice, which he measured with a tape measure because I asked him to and he thought I was joking — recovered most of it by night five.

I have stopped telling people that travel is a stress test for the device. Travel is a stress test for the pillow, the socket, and the evening. The band behaves the same in a motel as it does at home. Everything around the band does not.

Case three: the man who cut his number by more than half and still got heard

A.K., thirty-three, went from 47 minutes to 21 by week ten, the best percentage in the cohort. In week eleven his neighbor knocked on his door on a Tuesday morning and said, politely, that whatever he had done had helped a lot but that he was still audible “around five.”

He recorded three nights on his phone to check. The noise was not spread across the night. It was clustered in the last ninety minutes before his 6:30 alarm, and every one of those nights he woke up on his back with the band pushed up and rotated to the left, the pod no longer sitting under the jaw at all.

This is the honest limit of a neck-worn device, and I would rather write it down than let the app’s weekly average hide it. The device reduces the noise your throat makes. It does not stop you from rolling onto your back at four in the morning, and when you do, a band that has rotated a centimeter and a half is no longer in the right place to do anything. A.K.’s numbers looked good because six of his seven hours were genuinely quieter. His neighbor only ever heard the seventh.

What helped was unglamorous: a folded duvet wedged behind his lower back so that rolling onto his back required waking up enough to move it. Woken nights went from two a week to one in three weeks. Not zero. He kept the device, and he also stopped believing the weekly average was the whole story.

Three questions I now ask at intake

1. Who else can hear you through a wall, and have they ever put it in writing? A written complaint is the only baseline in this whole process that does not flatter you. It is also, in my file, a better predictor of whether you will still be wearing the device at week twelve than any number the app produces.
2. What time do you take it off? Not the snoring count. The removal hour. Everyone who was taking it off before 4 a.m. in weeks one and two was either off the device or down to two nights a week by week nine — except M.R., who fixed the fit. The 4 a.m. removal is a fit problem nine times out of ten, and a fit problem can be solved in an evening with a mirror and some patience.
3. Which side do you sleep on at 4 a.m., not at 11 p.m.? Those are two different people and only the second one matters.

How to read your own export without lying to yourself

• Contact hours first, snoring minutes second. A night with two hours of contact and eight minutes of snoring is not a good night. It is a night the microphone spent mostly alone.
• Weekly means, never single nights. One bad night is weather. Four bad nights in a row is a fit problem, a pillow problem, or a cold.
• Look at your last ninety minutes separately. That is where position changes, strap drift, and your worst noise all live. The nightly total will not show it.
• Photograph the mark on your throat in the morning, for one week. It is an unflattering series of pictures, and it will tell you more about your strap tension than every score in the app. The line should sit low, under the jaw. If it crosses the front of your windpipe, that is why night one felt like a collar.
• Write one sentence at 6 a.m., before coffee. Mine are full of people who feel fine after coffee and terrible before it, and the before-coffee ones were right.

The note

Here is the whole thing, transcribed from the photograph, which is still in the folder:

“To whoever is in 3B — I don’t know what you changed three weeks ago, but I have slept through the night four times this month, which has not happened since 2019. If it was a machine, whatever it cost, it was worth it. If it was a new mattress, I’m sorry for this note. — 3A”

M.R. wrote back and told her it was a band around her neck, and that it had cost less than the two of them probably spent on earplugs. They now nod at each other in the stairwell.

I keep the photograph because it is the only outcome measure in that entire folder that the patient’s neighbor can read. Every other number in those thirty-four files is something the wearer and I argue about. That one is just a person on the other side of a wall, sleeping.

This is a review of nine patient files from one clinic, written by a respiratory therapist. It is not a clinical trial, there is no control group, and nine people cannot prove anything. Snoring varies with weight, alcohol, allergies, position, medication, and the season, and individual results vary widely. A device that reduces snoring noise is not a treatment for sleep apnea, and nothing here is medical advice. If you snore heavily, wake gasping, or fall asleep during the day, talk to a clinician before you buy anything that goes around your neck.