Most sleep content sells you supplements. The evidence-ranked order of effect sizes puts supplements in fifth place. This protocol sequences the high-leverage levers first because that's where most of the gain lives.
What is the right order to fix broken sleep?
Sleep is a structured biological process with measurable stages and predictable responses to interventions. See Sleep Architecture for the physiology primer. The protocol below is the evidence-weighted sequencing of interventions, one tier at a time, with 1-2 weeks of holding each new habit before stacking the next.
How do you anchor your circadian phase in weeks 1 to 2?
Intervention 1: consistent timing. Bedtime +/- 30 min night-to-night. Wake time +/- 30 min across weekdays and weekends. "Social jetlag" , large weekend wake-time drift , is an independent metabolic stressor ( Besedovsky et al. 2019 ).
This is the single hardest habit to install. It is also the lever with the largest behavioral effect size outside temperature. If you can only change one thing, change this.
Intervention 2: bedroom 16-19°C (60-67°F). Cool rooms shorten sleep onset latency and increase slow-wave sleep. Mechanism: core body temperature naturally drops ~1°C during sleep onset; a cool room facilitates the drop.
Practical: cooling mattress pad is cheaper than AC upgrade. Thermostat set to 17°C at bedtime, rising to 20°C 30 min before wake time, works well.
What to expect: week 1 feels worse (habit installation is friction). Week 2 starts showing clear sleep onset improvements for most people.
When should you cut caffeine and alcohol for better sleep?
Intervention 3: caffeine cutoff 6-8 hours pre-bed. Half-life averages ~5 hours with significant genetic variability (CYP1A2 Haskell et al. 2008, n=27 fast vs slow). For a 200 mg coffee at 3pm, 100 mg is still circulating at 8pm.
Rule of thumb: if bedtime is 10pm, cutoff is 2pm. Some readers need 10am cutoff; a 2-week experiment with no caffeine after 10am tells you which bucket you're in.
Intervention 4: no alcohol within 4 hours of bed. Alcohol shortens sleep onset (feels sedating) but fragments REM and increases awakenings in the second half of the night. For resistance trainees, even moderate drinking attenuates muscle protein synthesis post-workout. On training days, zero is the target.
What to expect: if caffeine or alcohol was a hidden factor, changes show up by day 3-5. Total sleep time often increases by 15-45 minutes.
How does morning and evening light affect sleep?
Intervention 5: morning bright outdoor light 10+ minutes. Outdoor light is 10x brighter than indoor (1,000-10,000 lux vs 100-300). Morning light within 60 min of waking anchors circadian phase via suprachiasmatic nucleus input. Step outside with coffee for 10 min. Overcast counts; cloudy is still 1,000+ lux.
Intervention 6: evening <50 lux, 2 hours pre-bed. Bright indoor lighting in the evening suppresses melatonin and delays circadian phase. Dim the lights 2 hours pre-bed. Warm color temperature + reduced intensity beats "blue-blocking glasses" alone.
What to expect: circadian-related insomnia (can't fall asleep) improves within 1 week for most. The phase shift takes ~10 days to fully entrench.
Week 5: supplements, narrowly
If sleep is still broken, add supplements , not before. In order:
| Phase | Dose | Notes |
|---|---|---|
| Magnesium glycinate | 300-400 mg 60 min pre-bed | Small effect in low-Mg adults; larger if deficient. Glycinate is the sleep-studied form. |
| Melatonin (physiologic) | 0.3 mg 30-60 min pre-bed | Physiologic dose. Higher doses (3-10 mg) are pharmacologic and often worsen sleep quality. |
| Glycine | 3 g pre-bed | Modest effect on subjective sleep quality and onset latency in small trials. |
| L-theanine | 200 mg pre-bed | Anxiolytic more than sedative. Stacks with magnesium if anxiety-driven insomnia. |
| Apigenin | 50 mg pre-bed | Popularized by Huberman; evidence base is thin. Try only if above doesn't work. |
Start with magnesium alone. If no effect at 2 weeks, add melatonin at the physiologic dose. Never start at the pharmacologic dose: 5-10 mg melatonin doesn't outperform 0.3 mg for sleep quality and often causes grogginess and vivid dreams.
See the melatonin compound entry and magnesium glycinate entry for why lower doses beat higher ones (receptor saturation curves).
Weeks 6-8: hold, measure, refine
- Measure with your wearable (Oura, Whoop, Apple Watch). Trends matter more than absolute minutes. Total sleep time is reliable within 5-15 min; stage classification is noisier.
- Drop any supplement that isn't demonstrably helping after 2 weeks. The default should be fewer interventions, not more.
- If you are still sleeping poorly at week 8 after hitting all the above: see a clinician or sleep specialist. Undiagnosed sleep apnea, restless legs, delayed sleep phase disorder, and true chronic insomnia all mimic generic "bad sleep" and require different interventions.
When should you try CBT-I for insomnia?
Cognitive Behavioral Therapy for Insomnia is the most-evidenced treatment for chronic insomnia, outperforming sleep medications at long-term follow-up. Ong 2014 (n=54) showed mindfulness-based variants produce durable improvement ( Ong et al. 2014 ). Apps (Somryst, CBT-i Coach) deliver structured CBT-I at low cost.
Consider CBT-I if:
- Insomnia persists ≥ 3 months despite behavioral fixes.
- You lie awake for ≥ 30 min multiple nights/week.
- You wake in the night and can't re-sleep for ≥ 20 min multiple nights/week.
CBT-I is not a panacea; it's the most evidence-backed non-drug intervention available.
What to drop from the supplement aisle
- Sleep gummies with 5+ mg melatonin: the dose is pharmacologic, often counterproductive.
- "Sleep stacks" with ashwagandha + theanine + magnesium + GABA + melatonin: shotgun stacks stack side effects more reliably than efficacy.
- Blue-blocking glasses used alone: small effect beyond dimming lights.
- Mouth tape: useful for confirmed mouth-breathers/snorers; not a general sleep-quality intervention.
Counter-view
Matthew Walker's "Why We Sleep" ( Walker 2017 ) argues short sleep is catastrophically bad for health; Alexey Guzey's 2019 critique flagged overstatements. The conservative read: short sleep is bad, specific magnitude of badness varies by claim. Huberman's sleep cocktail stack goes further than the evidence strictly supports; start at the base protocol, add his extras only if the base fails.
