Sleep optimization: what the evidence says to do, in order

The short version

The best-supported intervention for bad sleep is behavioral, and it has a name: cognitive behavioral therapy for insomnia (CBT-I). Both the American College of Physicians and the American Academy of Sleep Medicine put it first, ahead of medication (PMID 27136449, PMID 33164742). It works on your sleep window and on your association between bed and sleep, not on a checklist of habits.

The goal is not to add hours to the night. It is to make the sleep you get more efficient, which is where the measured effect actually is. Fix the window first, then light, then caffeine, then alcohol. The supplements and the gadgets are the last thing to consider, and for most people they are the thing to skip.

What to fix first, and why the order matters

The internet treats sleep advice as a flat list: dim the lights, stop drinking coffee, buy a weighted blanket. That list has a name, sleep hygiene, and the AASM conditionally recommends that clinicians not use it as a single-component therapy for chronic insomnia (PMID 33164742). The advice is not wrong. It is just weak on its own, and it is usually the only thing most people are told.

The ordering below is by evidence quality and by mechanism, not by how easy each change is. Nobody has run a trial that randomises a caffeine cut-off against an evening light protocol at matched adherence, so any page claiming to rank these by measured effect size is making it up. What can be said is where the guidelines put the behavioral core, and what each lever is actually proven to do.

The sleep window comes before everything

Two components do most of the work: stimulus control and sleep restriction. Stimulus control means the bed is for sleep, and you do not go to bed until you are actually sleepy. Sleep restriction means the time you allow for sleep starts close to the time you actually sleep, then widens as efficiency improves. The AASM conditionally recommends each as a stand-alone therapy and strongly recommends the full multicomponent package (PMID 33164742).

The numbers behind that recommendation come from a meta-analysis of 87 randomised trials covering 3,724 patients. On the Insomnia Severity Index the effect size was g = 0.98, and on sleep efficiency 0.71. Wake after sleep onset improved by g = 0.63 and sleep onset latency by 0.57 (PMID 28392168).

Then there is the number almost nobody quotes: total sleep time moved by g = 0.16, the smallest effect in the whole set (PMID 28392168). The protocol makes your sleep less broken. It does not hand you an extra hour. If you start this expecting a longer night, you will quit in week two, which is exactly when sleep restriction feels worst and before the payoff shows up.

Light: bright day, dim evening, dark room

Light is the strongest signal your internal clock takes from the outside world, and there is now an international standard for measuring the part of it that matters. The consensus recommendations are expressed in melanopic equivalent daylight illuminance: above 250 lux during the day, under 10 lux in the three hours before habitual sleep, and under 1 lux in the sleep environment (PMID 35298459).

The evening figure is measured at the eye, in a vertical plane, at about 1.2 metres of height, which is roughly where your face is when you are sitting up. Daylight is the easy way to hit the daytime target. Dimming the room and getting the last light out of the bedroom is the easy way to hit the other two. Light hardware questions belong to the blue light glasses guide and the sleep products guide, not here.

Caffeine and alcohol, in that order

Caffeine has a longer tail than it feels like it does. In a randomised home study, 400 mg of caffeine taken 0, 3 or 6 hours before habitual bedtime each produced significant sleep disturbance against placebo, and the authors concluded that six hours is the minimum sensible cut-off (PMID 24235903). The interesting part is that the six-hour dose was still measurably disruptive, so "I can sleep after coffee" is not the test.

Alcohol is a different problem, because it works in both directions. A review of the polysomnographic literature found that at every dose it shortens sleep onset latency and consolidates the first half of the night, then increases disruption in the second half (PMID 23347102). The first REM period is delayed at all doses.

That is the trade a nightcap makes: a faster onset, bought with a fragmented back half and a 3am wake-up. If you are tracking anything overnight, this is also the lever that shows up most clearly in the data, because it raises heart rate and suppresses variability while you sleep.

Temperature and movement

The temperature effect people describe is real, but it is about your skin, not the thermostat. Using a thermosuit, researchers raised skin temperature by 0.4 °C without changing core temperature; that suppressed nocturnal wakefulness and shifted sleep toward deeper stages (PMID 18192289). In elderly participants it almost doubled the proportion of slow wave sleep. Read that as support for the principle, a cool room and a warm surface, not for any specific cooling pad or chiller, because no consumer product was tested.

Exercise helps modestly. A meta-analysis of 66 studies found regular exercise had small effects on total sleep time and sleep efficiency, and a moderate effect on sleep quality (PMID 25596964). The surprise is what did not moderate the effect: intensity, and whether the exercise was aerobic or anaerobic. Consistency and timing mattered more than how hard you go.

What a tracker can and cannot tell you

Consumer trackers are much better at knowing you are asleep than at knowing you are awake. Against laboratory polysomnography across 34 adults and three nights, epoch-by-epoch sensitivity was at least 0.93 for every device, while specificity ranged from 0.18 to 0.54 (PMID 33378539). Low specificity means missed wakefulness, which means the device overstates how well you slept.

It gets worse in the situation that matters. The same study found devices performed worse on nights with poorer or disrupted sleep, so the data is least trustworthy exactly when you are most likely to look at it. Sleep stage breakdowns were inconsistent between devices. And no consumer sleep score has been validated against a health outcome in any literature reviewed here, so treat the score as an interface, not a measurement.

What to skip

Melatonin and valerian first. The AASM's pharmacologic guideline suggests clinicians not use melatonin, valerian or tryptophan for chronic insomnia, and every drug recommendation in that guideline is graded weak (PMID 27998379). That is a lower bar than "it does nothing", and it is still not a reason to buy a bottle instead of fixing the window.

Sleep hygiene courses and apps next. The review that examined each hygiene recommendation found that while the individual components are associated with sleep, their direct effects are largely untested in the general population, and that sleep hygiene is ineffective as a clinical treatment (PMID 25454674). You already know the list. Knowing it harder is not the intervention.

Then the devices sold as substitutes, and the score-chasing that comes with them. If the specific question is which magnesium form to take, the magnesium guide covers it. If it is which blackout curtain or lamp to buy, the sleep products guide carries the picks and their prices. Neither belongs in the middle of a protocol decision.

The honest ceiling

The best behavioral protocol makes your sleep more efficient and less interrupted, and it moves total sleep time very little (g = 0.16, PMID 28392168). It is also not the same thing as sleeping more. Habitual short sleep is associated with higher all-cause mortality (RR 1.12) and so is long sleep (RR 1.30) across 1.38 million participants, and the long-sleep finding is best read as a marker of underlying illness rather than an instruction to stay in bed (PMID 20469800).

The other ceiling is access. Face-to-face CBT-I of four or more sessions outperformed self-help and shorter courses in the same meta-analysis (PMID 28392168), which means the app is a starting point and not an equivalent. If you have chronic insomnia and a clinician who offers CBT-I, that is the version with the evidence behind it. If you do not, the protocol above is the free approximation, and it is still better than the gadget.

Where the rest of this lives

This page is the protocol. It sells nothing and quotes no price. The shopping questions sit in three other guides that each verify prices against the vendor page on the day they ship.

• Blackout, earplugs, light and warmth, with picks and prices: the sleep optimization products guide.
• Which supplement form, what dose, and the 350 mg ceiling: the magnesium and sleep guide.
• What the overnight number on your ring means and what moves it: HRV explained.
• Which device, and what it costs over years: the wearables guide.
• Evening eyewear, graded against the trials: the blue light glasses guide.

Sources

• Qaseem et al. 2016. Management of chronic insomnia disorder in adults: ACP clinical practice guideline. Ann Intern Med. (PMID 27136449)
• Edinger et al. 2021. Behavioral and psychological treatments for chronic insomnia disorder: AASM clinical practice guideline. J Clin Sleep Med. (PMID 33164742)
• Edinger et al. 2021. Same guideline, systematic review, meta-analysis and GRADE assessment. J Clin Sleep Med. (PMID 33164741)
• van Straten et al. 2018. Cognitive and behavioral therapies in the treatment of insomnia: a meta-analysis. Sleep Med Rev. (PMID 28392168)
• Irish et al. 2015. The role of sleep hygiene in promoting public health: a review of empirical evidence. Sleep Med Rev. (PMID 25454674)
• Drake et al. 2013. Caffeine effects on sleep taken 0, 3, or 6 hours before going to bed. J Clin Sleep Med. (PMID 24235903)
• Ebrahim et al. 2013. Alcohol and sleep I: effects on normal sleep. Alcohol Clin Exp Res. (PMID 23347102)
• Raymann et al. 2008. Skin deep: enhanced sleep depth by cutaneous temperature manipulation. Brain. (PMID 18192289)
• Kredlow et al. 2015. The effects of physical activity on sleep: a meta-analytic review. J Behav Med. (PMID 25596964)
• Brown et al. 2022. Recommendations for daytime, evening and nighttime indoor light exposure. PLoS Biol. (PMID 35298459)
• Milner and Cote 2009. Benefits of napping in healthy adults. J Sleep Res. (PMID 19645971)
• Chinoy et al. 2021. Performance of seven consumer sleep-tracking devices compared with polysomnography. Sleep. (PMID 33378539)
• Cappuccio et al. 2010. Sleep duration and all-cause mortality: systematic review and meta-analysis. Sleep. (PMID 20469800)
• Sateia et al. 2017. Clinical practice guideline for the pharmacologic treatment of chronic insomnia. J Clin Sleep Med. (PMID 27998379)

Frequently asked questions