Most sleep advice ends at “try this.” Ours goes one step further, because you probably already wear a device that records deep sleep every night. That makes you a one-person trial, with the usual caveats about how noisy the data is.
Below are seven tweaks, what the research says about each, and how to check the result on your own numbers. What did your tracker say your deep sleep averaged last week? Keep that figure handy, because it is your baseline.
Why deep sleep matters, and how far to trust your tracker 🧠
Deep sleep, also called slow-wave sleep or N3, is the stage where your brain’s electrical activity slows into large, rolling waves and you are hardest to wake. Adults spend roughly 15 to 20 percent of the night there, and the share shrinks with age. Researchers think the stage helps the brain clear metabolic waste, which is why it gets so much attention in longevity circles.
The best human data comes from the Framingham Heart Study. In a JAMA Neurology analysis of 346 adults aged 60 and older, each person had two overnight sleep-lab studies about five years apart and was then followed for 17 years. Each 1% yearly loss of deep sleep went with a 27% higher dementia risk, based on 52 cases. This is observational. It does not show that losing deep sleep causes dementia, and the participants were older than most of our readers. It is still the main reason we care about the stage.
Now the awkward part. Most consumer trackers estimate sleep stages from movement and heart signals, not brain waves. A 2024 study comparing three wearables against sleep-lab polysomnography in healthy adults found that the Oura ring did not differ significantly from the lab for deep sleep. The Fitbit Sense 2 underestimated it by 15 minutes, and the Apple Watch Series 8 by 43 minutes. A multicenter Korean study of 11 trackers in 75 participants also found big differences between devices in stage accuracy. We did not find funding statements in the summaries we read. For the full intelligence picture behind developments like this one, LongevityHub Pro has you covered. For your own use, the lesson is that your absolute number may be well off, but changes in it over time are more informative.
Keep these rules in mind before you test anything:
Compare yourself to yourself. Use the same device, worn the same way, every night.
Judge any change by a 7-night average, because single nights swing a lot.
Feeling rested counts as a real outcome, and no score overrides it.
If you snore, gasp, or wake unrefreshed after seven to eight hours, ask a doctor about sleep apnea before chasing tracker scores.
Three daytime levers: schedule, exercise, caffeine
Tweak 1: keep a steady wake time. In UK Biobank wearable data covering almost 61,000 people and more than 10 million hours of sleep, the most regular sleepers had up to 48% lower all-cause mortality than the least regular. That is about death risk, not deep sleep, and the authors of a commentary on the field point out that no randomized trial has shown the link is causal. We did not find a trial showing regularity raises deep sleep itself. We still list it first because it costs nothing and makes every other test cleaner.
Tweak 2: exercise, but not right before bed. A meta-analysis of 66 studies found that acute exercise has a small beneficial effect on slow-wave sleep. A separate meta-analysis of evening exercise in healthy adults found a single evening session raised slow-wave sleep by about 1.3 percentage points. The same analysis found that vigorous exercise ending within an hour of bedtime may hurt sleep onset, total sleep time, and efficiency. Small gains, then, with a timing catch.
Tweak 3: set a caffeine cutoff. In a small crossover study in the Journal of Clinical Sleep Medicine, 12 healthy adults took 400 mg of caffeine (two to three cups of coffee) at 0, 3, or 6 hours before bed. Even the 6-hour dose cut total sleep time by more than an hour on an in-home monitor, and participants did not notice the loss. The study measured total sleep, not deep sleep specifically, and we could not confirm who funded it. Still, an afternoon coffee you cannot feel may be showing up in your data.
How to test these three:
Tweak 1: Wake within 30 minutes of the same time for 14 days, weekends included. That window is our suggestion, not a trial result.
Tweak 2: Do 30 minutes of moderate exercise on most days and finish at least an hour before bed. Again, our rule of thumb.
Tweak 3: Stop caffeine at least 6 hours before bedtime and watch your total sleep time as well as deep minutes.
Three evening levers: a warm bath, a cooler room, dimmer screens
Tweak 4: take a warm bath or shower 1 to 2 hours before bed. Your body temperature falls ahead of sleep, and warm water pulls blood toward the skin so you shed heat faster afterward. A Sleep Medicine Reviews meta-analysis screened 5,322 papers, included 17, and pooled 13. It found that water of about 40 to 42.5°C (104 to 109°F), for around 10 minutes, 1 to 2 hours before bed, improved overall sleep quality. Slow-wave sleep was one of the outcomes the authors analyzed. A bath only 15 minutes before bed seems too late, according to the lead author.
Tweak 5: cool the bed, not just the room. In a 2024 Scientific Reports study of 72 people at three centers, a cooling mattress raised nightly N3 by an average of 7.5 minutes over a 7.5-hour night and lowered heart rate by about 2.4 beats per minute. The spread was wide (a standard deviation of 21.6 minutes), so some people gained almost nothing. The effect tracked the gap between core and skin temperature rather than core temperature alone. This was a special mattress, so a cooler bedroom is indirect support at best. We could not confirm the funding.
Tweak 6: dim and warm your screens. A University of Basel study put 72 healthy men in front of displays starting four hours before bedtime. The low-melanopic light (the part of the spectrum that talks to your body clock) shortened the time to fall asleep and blunted melatonin suppression. A polysomnography study from Haifa found that two hours of short-wavelength screen light disrupted sleep continuity and architecture. Neither shows more deep sleep. What you can reasonably expect is faster sleep onset.
What to look for in your data:
Bath: Run it 10 minutes, finish 1 to 2 hours before bed, and watch your time-to-fall-asleep figure.
Cooler bed: Try lighter bedding and a few degrees cooler, then compare deep minutes over a week.
Screens: Switch to dim, warm light for the last two to four hours of the evening.
Expect some changes in sleep latency first, and deep minutes only if you are lucky.
The nightcap trap 🔬
Tweak 7: skip the drink, and don’t trust the score. Here the tracker can mislead you. In a 2024 Sleep study of 30 adults recorded with polysomnography, alcohol before bed increased slow-wave sleep during the first third of the night and cut REM sleep. The effects eased over consecutive nights, which suggests the body adapts. So a tracker might hand you a higher deep-sleep number after a few drinks, which would be a strange reward.
The longer view is less flattering. In a Sleep study of 42 abstinent people with alcohol dependence and 42 controls, the dependent group had less slow-wave sleep, even after long periods without drinking. That group is not the same as social drinkers, so we are not claiming moderate drinking does the same thing. It does show alcohol has long-lasting effects on the stage we are trying to protect. Have you ever noticed a jump in your deep sleep score after a drink?
Practical points:
Alcohol shifts deep sleep earlier in the night and trims REM, so a higher deep number is not a win.
The change fades with repeated nights, which makes tracker trends harder to read.
Test it honestly: two weeks with no alcohol, then two weeks as usual, and compare awake time later in the night.
Never use alcohol as your sleep aid.
How to run the test
Run a two-week baseline with no changes. Write down four numbers: average deep minutes, average total sleep, time to fall asleep, and a 1 to 5 score for how rested you feel. Then change one thing at a time for at least a week and compare 7-night averages. Start with the best-supported ones: a steady wake time, an earlier caffeine cutoff, and exercise.
Treat a swing of a few minutes as noise. Given the device errors above, anything smaller than that is not telling you much. One idea we left out is dinner timing, because the evidence is mixed. A 20-adult crossover found no significant adverse change in sleep architecture when dinner moved from 5 hours to 1 hour before bed.
If a tweak raises your deep minutes and you feel better, keep it. If your numbers rise and you feel the same, keep it only if it is free. Which of the seven will you try first, and what is your baseline?