By We looked at 437 nights from 114 people who recorded their sleep with the SleepTrace app between August and November 2026. The app detected snoring in 31.6% of those nights. In the group aged 45 and over it was 57.3% of nights (22 people), against 19.3% under 25 and 22.2% from 25 to 44. The sample is small and people chose to be in it, so read these figures as a snapshot of these nights and not as a statement about everyone.
- Nights recorded437Aug to Nov 2026
- People114median 2 nights each
- Median sleep6 h 36 minmedian, 437 nights
- Median awakenings1median, 437 nights
- Median time to fall asleep13 minmedian, 437 nights
Every figure about sleep on this page comes from the same 437 nights or from a part of them, and the text says which part. The only exception is the section on stopped recordings, which describes the nights that were left out. Under each chart there is a table with the same values, and the full set is in a CSV file at the bottom of the page. Where a group was too small to report, the text says so.
How often a night contains snoring
The app detected snoring in 31.6% of the 437 nights. On those nights the median was 5 minutes of detected snoring. The 75th percentile was 11 minutes and the 90th percentile 21 minutes, so a minority of nights carries far more snoring than the typical one. Snoring for more than 30 minutes was rare: 1.6% of all nights.
Show as table
| Measure | Value |
|---|---|
| Nights with detected snoring | 31.6% |
| Median snoring minutes, nights with snoring | 5 min |
| 75th percentile, nights with snoring | 11 min |
| 90th percentile, nights with snoring | 21 min |
| Nights with more than 30 minutes of snoring | 1.6% |
The minutes come from the app's sound analysis. They are an estimate, not a reading from a medical device. If you want to see what such a recording looks like in the app, the snoring app page walks through it.
Who snores most: by age group
The share of nights with snoring was higher in the oldest group. It was 19.3% of nights under 25, 22.2% from 25 to 44 and 57.3% at 45 and over. On nights with snoring, the median was 3, 2 and 6 minutes in the three groups.
Show as table
| Age group | Nights with snoring | Median snoring, nights with snoring | People | Nights |
|---|---|---|---|---|
| Under 25 | 19.3% | 3 min | 52 | 161 |
| 25 to 44 | 22.2% | 2 min | 37 | 144 |
| 45 and over | 57.3% | 6 min | 22 | 75 |
The 45 and over group is the smallest one: 22 people and 75 nights. A group this size can move a lot with a few more or fewer people, so the real gap may be smaller or larger than shown. We report the higher share and we do not explain it.
Snoring nights and awakenings
Nights without detected snoring had a median of 1 awakening, with a mean of 3.42 across 299 nights. Nights with 1 to 10 minutes of snoring had a median of 3 awakenings and a mean of 4.82 across 102 nights.
Show as table
| Group | Median awakenings | Mean awakenings | Nights |
|---|---|---|---|
| No snoring detected | 1 | 3.42 | 299 |
| 1 to 10 minutes of snoring | 3 | 4.82 | 102 |
People: this query did not return the number of people.
In this data, more snoring and more awakenings show up on the same nights. That is an association. It does not show that snoring wakes people up, and it does not show the opposite. This data does not cover other possible explanations. Groups with more than 10 minutes of snoring have fewer than 30 nights each, so we do not show them. Awakenings are the app's own estimate, like every figure here that comes from sound. For the medical side of the topic, see snoring and sleep apnea.
Falling asleep by age
Across all 437 nights, the median time to fall asleep was 13 minutes, and 18.8% of nights took longer than 30 minutes. By age group the picture differs. The median was 14 minutes under 25, 16 minutes from 25 to 44 and 10 minutes at 45 and over. The share of nights over 30 minutes was 16.1%, 32.6% and 4.0%.
Show as table
| Age group | Median to fall asleep | Nights over 30 min | People | Nights |
|---|---|---|---|---|
| Under 25 | 14 min | 16.1% | 52 | 161 |
| 25 to 44 | 16 min | 32.6% | 37 | 144 |
| 45 and over | 10 min | 4.0% | 22 | 75 |
People per age group come from the onboarding answer; the time-to-fall-asleep query itself did not return them.
The group from 25 to 44 stands out, with about one night in three taking longer than 30 minutes. We have no data on why. Time to fall asleep is an estimate from sound, so single minutes should be read with care. The groups have 161, 144 and 75 nights.
What people feel vs what the app measures
In the morning the app asks how you feel, on five steps from very bad to very good. 297 of the 437 nights have a value. Skipping the question is stored as neutral, so "neutral" below mixes neutral and skipped mornings, and the shares of good and bad mornings are lower bounds. "Good" means good or very good, "bad" means bad or very bad.
The sleep score and the morning
Bad mornings become rarer as the sleep score rises, from 28% below a score of 60 to 5% at 90 and above. Good mornings are about twice as common from a score of 80 up (37.3% and 36.0%) as in the 70 to 79 band (20.0%) and below 60 (21.9%). The 60 to 69 band stands out: its share of good mornings is the lowest, 11.3%. We do not smooth that out. This is not a comparison with a sleep lab. It only asks whether the app's number fits how the morning felt.
Show as table
| Sleep score | Good mornings | Bad mornings | Nights |
|---|---|---|---|
| Under 60 | 21.9% | 28.1% | 32 |
| 60 to 69 | 11.3% | 20.8% | 53 |
| 70 to 79 | 20.0% | 17.1% | 70 |
| 80 to 89 | 37.3% | 13.4% | 67 |
| 90+ | 36.0% | 5.3% | 75 |
People: this query did not return the number of people.
The same picture by mood group. The table shows only the groups "bad", "neutral or skipped" and "good". "Very bad" and "very good" each have fewer than 30 nights on their own, so they only count inside the shares in the charts.
| Measure | Bad | Neutral or skipped | Good |
|---|---|---|---|
| Nights | 37 | 173 | 68 |
| People | 24 | 68 | 28 |
| Mean score | 74.5 | 77.5 | 81.5 |
| Median score | 75 | 77 | 85 |
| Mean sleep (min) | 398 | 394 | 436 |
| Median to fall asleep (min) | 16 | 14 | 12 |
A person can appear in more than one column, because the same person can rate different mornings differently.
Sleep length and the morning
Nights of 7 to 8 hours had the highest share of good mornings in our data, a share more than three times as high as for nights under 6 hours (44.1% against 13.2%). Above 8 hours the share was lower again, 34.7%. That group is small, 49 nights from 21 people, and we do not interpret the dip. Sleep length here is the app's own estimate from sound. The mean score per group is in the table.
Show as table
| Sleep duration | Good mornings | Bad mornings | Mean score | Nights | People |
|---|---|---|---|---|---|
| Under 6 h | 13.2% | 17.1% | 70.3 | 76 | 47 |
| 6 to 7 h | 21.2% | 17.3% | 79.0 | 104 | 46 |
| 7 to 8 h | 44.1% | 11.8% | 83.8 | 68 | 35 |
| 8 h+ | 34.7% | 12.2% | 80.0 | 49 | 21 |
Snoring and the morning
Nights with detected snoring were not rated worse by the person recording: 30.9% good mornings, against 24.5% on nights without snoring (97 against 200 nights). The mean score was 81.5 against 76.3. The app hears snoring in the room and cannot tell whose it is, and the person next to them is not asked. We draw no conclusion.
What we looked at and do not report
Three more things were examined and are left out. The course of sleep over the first weeks with the app: only 14 people in the sample have at least 7 nights, below our limit of 20 people, so we give no figures. Setup answers about trouble falling asleep and about waking at night, compared with the measurement: in each comparison one of the two groups has fewer than 20 people (21 against 14, and 20 against 15), so we report nothing, not even "no difference". Satisfaction with sleep from the setup: only one answer level reaches 20 people, so there is nothing to compare.
How loud the night gets, hour by hour
A phone microphone has no fixed scale, and phones differ. This report gives no sound-pressure levels, only differences. The raw values in the CSV are relative to the microphone's full scale, not calibrated. The chart shows the mean peak level in each hour after falling asleep, measured against the quietest hour (hours 3 and 4 tie and are set to 0).
Show as table
| Hour | dB above quietest hour | Epochs |
|---|---|---|
| 1 | +1.7 | 42,319 |
| 2 | +0.2 | 41,622 |
| 3 | 0 | 40,594 |
| 4 | 0 | 39,776 |
| 5 | +0.3 | 38,159 |
| 6 | +1.2 | 33,659 |
| 7 | +1.4 | 24,979 |
| 8 | +3.0 | 14,061 |
| 9 | +6.5 | 4,886 |
People: this query did not return the number of people.
Hours 2 to 5 are the quietest, all within 0.3 dB of the minimum. The first hour sits 1.7 dB above the quietest one, then the level drops and stays flat. From hour 6 it climbs again, and the ninth hour is 6.5 dB above the minimum. Fewer nights last that long, so the last point rests on 4,886 thirty-second epochs, against 42,319 in the first hour. We did not test why the level rises in the last hours of a recording.
A second measure looks at single moments. The loudest 30-second stretch of a night lies a median of 42.5 dB above that night's mean peak level. In 10% of nights the gap is more than 56.2 dB. This covers 344 nights with at least 2 hours of audio. A large gap says that one moment stood out from the rest of the night. It does not say what the sound was.
Why people stop a recording
Of 960 nights started in the period, 255 (26.6%) were stopped by the user before the end. The app stores a reason when one is given. The most common answer by far was "woke up": 168 times, which is 65.9% of the 255. In 43 cases no reason was given. Another 26 chose "other", 13 could not fall asleep, 3 named noise and 2 needed to charge the phone. Cancelled nights are left out of every other figure on this page.
Show as table
| Reason | Cancelled nights |
|---|---|
| Woke up | 168 |
| No reason given | 43 |
| Other | 26 |
| Could not fall asleep | 13 |
| Noise | 3 |
| Charging needed | 2 |
People: this query did not return the number of people.
Sleep length by age and what the app estimates
The oldest group has the longest mean sleep, 437 minutes, and the shortest time awake during the night, 28 minutes. That describes 22 people. It is not a statement about older adults in general.
| Age group | People | Nights | Sleep, min | Awake, min |
|---|---|---|---|---|
| Under 25 | 52 | 161 | 391 | 39 |
| 25 to 44 | 37 | 144 | 399 | 40 |
| 45 and over | 22 | 75 | 437 | 28 |
Across all 437 nights, the app estimates that 20.2% of sleep time is deep sleep, 13.8% REM and 66.0% light sleep, with 35 minutes awake on average. These are estimates from sound, not EEG. SleepTrace estimates sleep stages from sound and is a wellness app, not a medical device. We do not show the stages by age: in the smaller groups a few heavy users move the numbers too much. How the stages work is explained in how sleep phases work.
Method
Who is in the sample
The data covers nights started between 1 August and 8 November 2026. In that period, 173 people started 960 nights. We removed three groups: 255 nights the user cancelled, 232 nights shorter than 3 hours and 36 nights without a usable duration. That leaves 437 nights from 114 people. Three internal accounts were excluded from the sample.
The nights are not spread evenly. The median is 2 nights per person and the maximum is 50, so one person accounts for 11% of all nights. Of the 437 nights, 355 come from iPhones (102 people), 12 from Android or other devices (5 people) and 70 have no device field (24 people). The people counts add up to more than 114 because a person can appear under more than one device value. Age comes from the onboarding question: 52 people are under 25, 37 are 25 to 44, 22 are 45 or older and 3 gave no age. The morning question asks how you feel: 297 of the 437 nights have a stored value, and a skipped question is stored as neutral.
How the figures were made
The app estimates snoring, awakenings and sleep stages from the sound of the night, with on-device sound analysis. For this report, queries ran on the stored results and returned only sums, medians, percentiles and shares. No single person's night is visible in the output. Age groups are shown only with at least 30 nights and at least 20 people. Breakdowns of nights (snoring buckets, hours after falling asleep) need at least 30 nights, or 500 thirty-second epochs per hour. The reasons for stopped recordings are listed in full, including small counts, because they describe the nights that were left out, not the sample. Where a query did not return the number of people, the table says so. The same limits apply to the mood groups in the table and to the sleep-length groups, whose queries returned the number of people. The sleep score bands have at least 30 nights each; their query did not return the number of people. More on how the app works is on the methodology page, and on about SleepTrace.
Limits
- Snoring, awakenings and sleep stages are estimates from sound. There is no EEG, and we do not compare with sleep-lab results.
- The phone microphone is not calibrated. We show only differences in dB, never sound-pressure levels. The raw values in the CSV are relative to full scale.
- People chose to install a sleep app and to record. The sample is self-selected and says nothing about how the general population sleeps.
- Local time cannot be computed, because the app does not store a time zone. That is why this report has no bedtimes or fall-asleep clock times. Because no time zone is stored, we also report no weekdays.
- Mood is the user's own morning rating; a skipped question counts as neutral, and nights without any value are left out of those figures.
- Watch data covers 27 nights, below our limit of 30, so it is left out.
- The onboarding does not ask about snoring. We therefore cannot compare what people believe about their snoring with what the app recorded.
What we will not claim
No diagnoses. No causes: where two things appear together, we say so and stop there. No comparison with sleep-lab values. No statement about age groups below the size limit, and none about all people or any country.
Reproducibility
The queries return aggregates only. The data was pulled on 8 Nov 2026, and every number on this page matches that pull. The CSV below holds each figure shown here, plus the raw hourly levels behind the loudness chart, with the number of nights for each row and the number of people where the query returns it.
Download and cite
Download the data as CSV. One row per figure: question, group, metric, value, unit, number of nights, number of people (empty where the query does not return it) and the query it came from. Values use a decimal point on both language versions of this page.
To cite this report:
SleepTrace (2026). Sleep Sound Report 2026: aggregated data from 437 recorded nights. Sleep Trace Ltd, Larnaca. https://sleeptrace.app/sleep-sound-report-2026/ (retrieved YYYY-MM-DD)
Aggregated figures and the CSV may be reused with attribution (CC BY 4.0).
SleepTrace is a wellness app, not a medical device. This page is general information, not medical advice. If you stop breathing in your sleep, fall asleep while driving or feel unwell, see a doctor.