A plain-English, day-by-day field guide for capable adults whose sleep has slipped — and who are tired of the supplements, the trackers, and the advice that doesn't survive a Tuesday. Built on CBT-I, circadian biology, and the one habit that actually moves the needle.
Read Part I tonight. Run Parts II–IV as your 21-day calendar. The daily log (p. 18) begins on day one. The plan (p. 16) is the map. This guide is educational, not clinical — if you suspect apnea or another sleep disorder, please read Chapter 03 today.
You've tried the apps, the trackers, the magnesium, the mouth tape. Something helped a little for a week. Then it didn't. The reason isn't that you haven't found the right gadget. The reason is that sleep responds to behaviour and environment, in a specific order, and most advice scrambles the order.
This guide is for the high-functioning version of the problem: the people who get the work done, hit the deadlines, look fine on the outside — and lie awake at 3 a.m. wondering if they will ever sleep like a normal person again. That picture has a name in the clinical literature ("chronic insomnia, sub-clinical") and a clean, evidence-based protocol to walk back from. It's called CBT-I, and most of what's below is a translation of it.
Sleep is not a thing you do well or poorly. It is the natural state of a body whose signals are unblocked. The work is removing the blocks.
Start with Chapter 01. The model is simple but most people get it wrong, and the rest of the guide assumes the right one.
"Get eight hours" is the worst-quality advice in the entire wellness genre. Hours are the byproduct, not the goal. The goal is to cycle through the four stages of sleep enough times, in the right order, with enough of each. Once you see the architecture, you'll stop chasing the clock.
N1 · The slip. Five to ten minutes. Drifting off. Easily woken; you might not believe you were asleep. Light noise interrupts it.
N2 · The settle. Light sleep, about half the night. Memory consolidation begins. Body temperature drops a notch. Heart rate slows.
N3 · Deep sleep. The big one. Mostly in the first third of the night. Growth hormone, immune repair, glymphatic clearing (the brain's waste-out cycle). Skip N3 and you wake up "tired" even after eight hours.
REM · Dream sleep. Mostly in the second half. Emotional processing, creative consolidation, fine-grained learning. Cut your sleep short and REM is what you lose first.
You move through N1 → N2 → N3 → back to N2 → REM. One cycle = ~90 minutes. A full night is four to six. More deep sleep early, more REM late. "I went to bed at 2 and woke at 8" is not the same as 10 to 4 — the second skips most of your REM.
Sleep is not a battery you charge. It is a sequence the brain runs to repair four systems.
Modern sleep science explains sleep with two interacting systems:
Process S — sleep pressure. Builds the longer you're awake. Adenosine accumulates. Caffeine blocks the receptors that feel it.
Process C — circadian rhythm. A 24-hour clock that opens and closes the sleep window. Set by light, primarily morning light. Most modern sleep problems are circadian, not pressure.
The fix: load up sleep pressure (Ch. 09), anchor circadian rhythm (Ch. 04, Ch. 07). That's most of the protocol.
Stop tracking your "hours of sleep" for the next 21 days. Track wake time, perceived restoration (1–5), and energy at 3 p.m. (1–5). The log (p. 18) does this. Hours mislead; restoration tells the truth.
"I can't sleep" is five different problems with five different fixes. Naming yours precisely is the difference between fumbling and progressing. Most people have a primary one and a secondary one. Identify both before you start.
Lights out at 11. Still awake at 12:30. Mind churning, body alert. The standard "I lie there and can't switch off" pattern.
Primary fixes: wind-down (Ch. 08), morning light (Ch. 04), worry window (used here too).
Asleep by 11, awake by 3. Maybe pee, maybe not. Mind switches on. Twenty minutes of attempted re-sleep, then another spiral.
Primary fixes: the 3 a.m. protocol (Ch. 11), alcohol audit (Ch. 06), sleep restriction (Ch. 09).
Sleep is fine until almost dawn. Body wakes early, mind alert, no chance of falling back. Often comes with low-grade depression or grief.
Primary fixes: evening light audit (Ch. 04), GP consult re: mood (Ch. 03), sleep restriction (Ch. 09).
You hit the pillow, you slept, the watch says 8 hours. You wake feeling hit by a truck. Often a sign of apnea, fragmented sleep, or alcohol.
Primary fixes: see Ch. 03 (apnea screen), alcohol cut (Ch. 06), bedroom rebuild (Ch. 05).
You're a "night owl" — but increasingly it's making your life worse. Often a circadian delay, not a preference. Most reversible of the five.
Primary fixes: morning light at 7 a.m. (Ch. 04), wake-time anchor (Ch. 07), wind-down (Ch. 08).
You don't have a sleep problem any more. You have a "fear of not sleeping" problem. The original issue is healed; the anxiety persists.
Primary fixes: Anxiety Toolkit guide; the relapse protocol (Ch. 12); compression (Ch. 09).
The same protocol handles all five — but in different orders. Knowing your type tells you which chapters to read first, not which to skip.
Underline your primary type. Circle a secondary. Note the date in the margin. On day 21, come back to this page. The honest before-and-after is more useful than any tracker graph.
Most sleep problems respond beautifully to a behavioural protocol. Some do not — because the real problem is medical. Here is the honest list. If any apply, please see a clinician this week, in parallel with this guide.
01 · You snore — and your partner says you stop breathing. Or you wake gasping. Or you're tired despite eight hours. The apnea pattern. Common, treatable, under-diagnosed. A sleep study is one night.
02 · Your legs feel "creepy-crawly" at night. Restless legs syndrome. Worse on iron-deficient days. Ask your GP for a ferritin level.
03 · You act out dreams. Punching, kicking, shouting in your sleep. REM behaviour disorder. Please see a neurologist — there are specific reasons this matters early.
04 · You sleep nine to ten hours and feel un-rested. If consistent for months, please get bloodwork. Hypothyroidism, anaemia, depression all present this way.
05 · You suspect insomnia driven by depression. Persistent low mood plus early waking suggests treating the mood first. Sleep usually follows. Talk to your GP.
The behavioural protocol and a prescription are not in competition. The strongest finding in the field: CBT-I and short-term medication, used together, beat either alone — then meds taper while the behaviour stays.
Behaviour first, behaviour permanent, medication situational. There is no shame in either. There is shame in white-knuckling for months when you could have asked for help.
Asking for clinical help is not "giving up on the work." It is the work.
If you have run this protocol carefully for four weeks and your sleep has not measurably improved on the log, please see a sleep clinician. That's the threshold where "trying harder" stops being the answer.
Over-the-counter sleep aids (diphenhydramine, doxylamine) work for a week, then stop, and often make sleep worse over months. Use sparingly, if at all.
If any red flag above applies, book an appointment today — even if you're "fine." Sleep clinics are booked weeks out. Start that clock. Then begin the protocol; the two run in parallel beautifully.
More than melatonin, more than magnesium, more than mattresses: light is the variable that sets your circadian clock. Most modern sleep problems are a light problem in disguise — too dim in the morning, too bright in the evening, exactly the opposite of what the system was built for.
Within 30 to 60 minutes of waking, get bright light into your eyes — preferably outside, no sunglasses, no phone in front of your face. The signal goes to a small region of the hypothalamus (the SCN), which tells the rest of the body "the day has started" and starts a 14-to-16-hour countdown to melatonin release tonight.
The doses:
For the last 2 to 3 hours before bed: bright overhead light off. Floor lamps on. Amber bulbs win. Phones on night-shift and dimmed to the floor. Why so dramatic? Because the SCN cannot tell the difference between a phone screen at 11 p.m. and noon sun. Your body genuinely thinks it's still daytime.
You will get further in two weeks with a daily morning walk than with two years of any supplement on the shelf. Take the walk.
Light therapy lamps after 10 a.m. start having the opposite effect — they delay melatonin tonight. If you use one, it goes on before 9 a.m., not later.
Walk outside for 10 minutes within an hour of waking. No phone. Notice — that night — whether sleep onset is faster. Most readers feel it inside three days. The single highest-leverage habit in this guide.
The bedroom should do three things: be cool, be dark, be quiet. Then it should do a fourth: be associated, in your brain, with sleep only. Most sleep environments fail two of the four. The rebuild is cheap, fast, and disproportionately effective.
01 · Cool. Core body temperature drops about 1°F during sleep. The room helps. Target: 65–68°F / 18–20°C. Most people sleep in a room 4–6° too warm and have no idea.
02 · Dark. Pitch dark. Hand-in-front-of-face dark. Even a single LED on the smoke detector reduces melatonin measurably. Solution: blackout curtains, electrical tape over LEDs, no clock on the bedside table.
03 · Quiet. Sudden noise > steady noise. A white-noise machine masks the spikes (traffic, partner, fridge). $30. Larger effect than most supplements.
04 · Single-purpose. The bed is for sleep and sex. Not work, not scrolling, not "just resting." Your brain forms a strong location-association in two weeks; we want the location to mean one thing.
You probably don't need a new mattress. You probably do need a pillow that matches how you sleep — side, back, or stomach. The single most common upgrade with the biggest impact: a body pillow for side sleepers, to keep the spine straight.
The phone has to leave the room. Not because of EMF. Because of availability: the moment you wake, your brain knows the phone is reachable, and it stays slightly alert all night to consider whether to reach for it. Charge it in the kitchen. Buy a $12 analog alarm clock.
If the bedroom is also your office, your living room, and your TV room, your brain has no place that means "sleep." Build the place.
Bedroom TVs are a stable, long-term predictor of poor sleep — and the single hardest habit to break for couples. If you can't remove it: no shows after 9, lights up while watching, off at the latest 60 minutes before bed.
Walk into your bedroom and audit: temperature, darkness, noise, single-purpose. Fix one tonight (electrical tape on LEDs is free). Fix one this week ($30 white-noise machine). Fix one this month (blackout curtains if needed).
Three substances move your sleep more than every supplement on the market combined. The good news: you don't have to quit any of them. You have to time them. The bad news: the timing is more aggressive than you've been told.
Caffeine has a half-life of 5 to 6 hours. After 10 hours, a quarter of your morning coffee is still in your system. That last 25% is enough to fragment N3 sleep — the deep, restorative stage — without you noticing it as "trouble sleeping."
The rule: last caffeinated drink by 2 p.m. No exceptions for "I can sleep right after coffee" — you can fall asleep, you don't sleep well.
One drink helps you fall asleep faster. Two drinks suppress REM and fragment the second half of the night. The standard wine-with-dinner habit costs most people their REM on most nights.
The 21-day move: zero alcohol for the duration. Decide after.
A big meal close to bed makes the body work — digestion, insulin, temperature regulation — at the moment it's trying to do the opposite. Fix is timing, not diet:
Most "supplement stacks" are people trying to solve with biochemistry what they could solve faster with two timing changes.
Move your last caffeine to 2 p.m. by Thursday. Zero alcohol for the next 21 days. Last meal three hours before bed. Track perceived restoration in the log (p. 18). Watch the line move by day five.
The single most powerful lever in the entire protocol — and the one most readers resist hardest. Pick a wake time. Hold it for 21 days. Weekends included. The whole circadian system depends on a fixed anchor at one end. We pick the wake end because it's the only one we can actually control.
You can't will yourself to fall asleep at 10:30. You can absolutely will yourself out of bed at 6:30 — alarm, lights, feet on the floor. Fixed wake time forces the circadian system to find a sleep window that matches. Fixed bedtime, with variable wake, just creates frustration in the dark.
Mornings two and three are the hardest. Sleep pressure is high, you've not yet rebuilt the rhythm, the alarm feels cruel. This is exactly the point at which the protocol works. Most quitting happens here.
The reframe: the bad early mornings are building the sleep pressure (Process S, Ch. 01) that makes the next night easier to fall asleep. You are not "punishing yourself with poor sleep." You are accumulating the chemical signal your body needs.
Fix one variable. Anchor the morning. Watch the rest of the week move toward it on its own.
The "I'll start Monday" version of this protocol never starts. Pick a day in the next three days. Today is also a fine choice.
Set the alarm. Across the room. Same time tomorrow and the next 20 days. Note in the log: did you make it out of bed on the alarm, yes or no. Day 21 is the day you decide what to keep.
Sleep is a ramp, not a switch. The 90 minutes before bed are not "evening" — they are the runway. Most modern lives ask the nervous system to be on at 10:29 and asleep at 10:30, which it cannot do. Build the ramp.
T-90 to T-60 · Close the day. Last work email written. Tomorrow's first task noted. Phone moves to the kitchen. Lights start coming down — overheads off, lamps on.
T-60 to T-30 · Cool the system. Warm shower (counter-intuitively — it dilates blood vessels and the body sheds heat afterward, lowering core temperature). Or: tea. Or: light tidying. Slow movement only.
T-30 to T-0 · Quiet the mind. Paper book. Slow breath. Conversation with the person in your house. Pen-and-paper journal — three lines of "what happened today." Lights down to one lamp. Then to bed.
The highest-leverage move in the wind-down: a 5-minute paper list of "everything I think I need to do tomorrow." Once written, the mind stops rehearsing it overnight. Readers who do this fall asleep 12 to 18 minutes faster within a week.
You are not "wasting" 90 minutes. You are protecting the four hours of deep sleep that follow.
Don't make the wind-down another performance you can fail at. If you have 40 minutes, do 40. If you have 20, do 20. Consistency, not perfection.
Pick three things to remove from the last 90 minutes before bed (email, social, overheads). Pick three to add (lamp, paper book, brain-dump). Run them in order tonight. Note tomorrow morning how you feel.
The clinical core of CBT-I, and the most counter-intuitive idea in sleep: if you can't sleep well, spend less time in bed, not more. For two weeks. Then expand. This technique, used correctly, fixes more chronic insomnia than any drug on the market.
If you're in bed nine hours and sleeping six, you're spending three hours awake in bed — teaching your brain that bed is a place where you lie awake. Compress the window to six and a half. Your sleep efficiency goes from 67% to 92% inside a week. The brain learns: bed means sleep, fast.
The first 3 to 5 nights are hard. You will be tired during the day. This is the point — you are building sleep pressure (Process S) so the body sleeps efficiently when given the chance. By night 6 or 7, most readers fall asleep within 8 minutes.
Less time in bed produces more sleep. Two weeks. Trust the protocol.
Naps during this period sabotage the build-up of sleep pressure. Either no naps for 14 days, or one strict 20-min nap before 2 p.m.
Calculate your sleep window using the log. Hold it for 14 nights. Track efficiency (% of time in bed actually asleep). The graph moves quickly — most readers see >85% by night 7.
Two ten-minute rituals — one at each end of the day — that hold the rest of the protocol in place. Once the wake time, the wind-down, and the light protocol are in motion, the anchors are how the system survives a hard week.
Most "sleep routine" content piles on stuff. The two-anchor structure does the opposite: it identifies the two times of day with the highest leverage and the lowest social friction, and protects ten minutes of each. The middle of the day takes care of itself.
You cannot control the middle of the day. You can almost always control the first ten minutes and the last ten. That's enough.
Don't add a fifth thing to the anchor in week three because you're feeling ambitious. The point is that the anchors are short enough to survive your hardest weeks. Keep them at ten minutes.
Try the morning anchor in full tomorrow. Try the evening anchor tomorrow night. Take three sentences of notes in the log (p. 18). Adjust if needed — but try the exact protocol once before adapting it.
The classic complaint. You wake up clear, alert, certain you'll never get back. The mind sprints. The body braces. The clock taunts. This is a recognisable, specific event — and it has a specific protocol. Run it. Don't improvise at 3 a.m.
Cortisol begins rising in the second half of the night to prepare you to wake. For sensitised sleepers, this normal rise crosses the threshold into alertness around 3–4 a.m. The thoughts that arrive aren't "the truth that hides during the day" — they're the chemistry of cortisol meeting a tired brain in the dark.
The single most useful thought at 3 a.m. is: "These thoughts are not insights. They are chemistry. They will look smaller at 9 a.m. They always do."
Write this on a notecard. Tape it inside your bedside drawer. Read it when you wake. The pre-decision saves you the cognitive load of generating the reframe in a chemical fog.
The thoughts of 3 a.m. are not deeper. They are dimmer. The mind, in this state, mistakes alarm for truth.
If you woke needing the bathroom: do not turn the bright light on. A bedside path-light at floor level. Don't pick up the phone. Don't engage with thinking. Return, lie down, run 4-7-8. The whole event should be under 4 minutes.
If you've been awake more than 30 minutes, do not stay in bed "trying" — that is exactly the trained association we're undoing. Get up, dim room, paper book, return when sleepy. Every time.
Write the reframe on a card tonight. Put it in the bedside drawer. Set the path-light. Memorise 4-7-8. The next time you wake at 3, the protocol runs without thinking. That's the whole point.
Sleep will get worse again. A travel week. A family crisis. A flu. A late deadline. The win at day 21 is not "I never slept badly again." It is "I now have a 48-hour reset I can run." Here it is.
One bad night is a night. Two bad nights is data. Three bad nights is a request from your nervous system to re-engage the protocol — not evidence the protocol failed.
The win isn't that sleep never slips. It's that you have a 48-hour move that brings it back.
Once the reset is in, run a "light week" of the basics: morning walk, fixed wake, wind-down, anchors. Don't run sleep restriction again unless three weeks of basics haven't returned you to baseline.
The biggest relapse trigger isn't travel — it's the slow drift of "just one drink with dinner." Audit monthly. Re-read Ch. 06 and Ch. 07 on the 1st.
Bookmark this page. The next time you have three bad nights, return here. Run the 48-hour reset. Don't improvise — the protocol works because it's pre-decided.
Day by day, written for adults with full calendars. Each day is one small move. Stack the moves; don't restart them.
Most readers' sleep efficiency goes from ~70% to ~90% in 21 days. The protocol becomes background. Re-run the 48-hour reset (Ch. 12) any time three nights slip.
Patterns that quietly make it worse. The fix in each row is intentionally short — it's the move, not the philosophy.
Suppresses REM. You sleep faster, worse, and wake at 4. Fix: zero alcohol for 21 days; decide after.
Resets your circadian clock to a later phase. By Sunday night you can't sleep. Fix: ±30 min, never more.
Trains the brain that bed = wakefulness. Fix: if awake >20 minutes, get out, dim room, return sleepy.
Wrong dose, wrong timing. Use 0.3 mg, 4 hours before bed — and only if circadian re-anchoring is the goal.
Orthosomnia — anxiety caused by the tracker. Fix: track perceived restoration on paper, not wrist data.
No ramp, no chance. Fix: 90-minute wind-down (Ch. 08). Email closes at T-90.
Coffee escalation, then "I'll sleep in Saturday." Cycle continues. Fix: same wake time + morning walk anyway.
Fragmented N3, light leakage, ambient noise spikes. Fix: TV out of bedroom, or off by T-60.
You also keep it for everything else. Fix: $12 analog alarm clock, phone in kitchen.
The Monday version doesn't start. Fix: pick a day in the next 72 hours. Today is also a fine choice.
Each of these is reversible. Most of them in two weeks. None of them require willpower — they require structure.
Two minutes a day, on paper, every morning. The pattern emerges in a week. Tracker apps are not a substitute — the act of writing is part of the work.
You're looking for two things, not "8 hours."
Every Sunday, look back over the seven rows. Ask three questions:
The log makes the invisible visible. Most readers can predict their best/worst nights from the day-before column by week two.
If the log itself starts producing sleep anxiety ("I'd better sleep tonight so my chart looks good"), put it down for three nights. The log is a tool, not a test.
Short answer: at low doses (0.3 mg), occasional use, probably yes. At common over-the-counter doses (5–10 mg), nightly, indefinitely — not enough data, and the dose is way more than the body produces. Use it for circadian re-anchoring, not as a sleep aid.
One disciplined 20-minute nap before 2 p.m. is fine for most. Longer or later naps eat into the night's sleep pressure and ruin the rebuild. During sleep restriction (Ch. 09), no naps for 14 days.
Mostly. Anchor your sleep window on a fixed schedule — most of the protocol still works. The light protocol inverts: blackout curtains and morning-equivalent light when you wake at 5 p.m. Pair with a clinician if possible.
No. This guide is the wrong tool for that phase. Sleep when you can, return to this guide six months after the youngest sleeps through.
Often, yes — days 2 to 5 of sleep restriction (Ch. 09) and the first weekend you hold the wake anchor. The body needs to rebuild sleep pressure. The other side comes fast.
For the 21 days: no. Most cause more anxiety than insight. Use the paper log. After day 21, if you want a tracker, fine — but treat it as ambient data, not a score.
Used short-term, with a doctor, to break a bad streak: useful. Used nightly for months: usually counter-productive. Behaviour first, behaviour permanent. Discuss with your GP.
Two reasons it usually doesn't: people quit on day 4 of sleep restriction (the worst day), or they skip the light protocol (Ch. 04). Go back. Do those two specifically. Most "failed CBT-I" returns surprise results.
Like alcohol — falls into N1 fast, suppresses REM. Sleep looks easy; you wake foggy. For the 21 days, same advice: zero. Decide after.
For most readers with sub-clinical insomnia: 14 to 21 days for efficiency >85%, three to six weeks for the felt sense of "I sleep well." Severe or long-standing cases: longer; please add a clinician.
The honest version: sleep is highly trainable, very fast to improve, and very hard to keep ignoring. Three weeks of structure beats three years of supplements.
Real reader stories, names changed, details lightly altered. Both finished the 21 days. Both still occasionally have a hard night. Both know how to bring it back.
Starting point. Asleep by 11. Awake at 3, every night. 45 minutes of churning. Back to sleep around 4. Up at 7. Scores at day 1: restoration 2 / 5, energy at 3 p.m. 2 / 5, sleep efficiency ~68%.
What she did. Wake anchor at 6:45 (Ch. 07), morning walk (Ch. 04), phone to kitchen (Ch. 05), zero alcohol (Ch. 06). Started sleep restriction on day 7 — 6.5 hour window. The hardest week of the 21 days.
The turn. Day 9. She woke at 3 once more, ran 4-7-8, got out at 3:25, read a paper book in the kitchen, went back at 3:55, slept until alarm. "That was the first night I felt like I had a tool, not a problem."
Result at day 21. Restoration 4 / 5, energy 4 / 5, efficiency 91%. The 3 a.m. wake reduced to roughly once a week — and when it happens, the protocol handles it in 15 minutes.
Starting point. Two glasses of wine most nights. Email until 11:45. Bed at midnight, asleep instantly, awake at 4:30. Six months of this. Scores at day 1: restoration 2 / 5, energy 1.5 / 5, efficiency ~74%.
What he did. Zero alcohol from day 1 — the hardest move and, he said later, the most decisive. T-90 wind-down with email closed (Ch. 08). 6 a.m. walk for 20 minutes (Ch. 04). Wake anchor at 5:45.
The turn. Day 4 was awful — he calls it "the worst night I'd had in a year." Day 5 he slept seven straight hours for the first time in months. "The body had been screaming for the wine to leave. It really did just need three days."
Result at day 21. Restoration 4 / 5, energy 4 / 5, efficiency 93%. After day 21 he reintroduced one glass of wine on Friday and Saturday only. Restoration held. "Two glasses every night was the actual problem. Five months in, I haven't gone back."
Both stories share a structure: anchor the wake time, fix the light, remove the night-time alcohol, build the wind-down. The specifics vary; the protocol doesn't.
Opinionated, ranked, ad-free. The short list of things that actually help.
Please book an appointment — with your GP or a sleep clinic — if any of the following apply:
Asking for clinical help is not a failure of the protocol. It is the most evidence-rich step you can take. This guide and a real clinician can coexist. The best recoveries usually involve both.
Rebuild Your Sleep — 21 Days is a plain-English, day-by-day field guide for capable professionals whose sleep has slipped past "stressed week" into "this is my normal" — and who want a protocol that's behavioural, biological, and non-shaming.
The protocol draws on cognitive-behavioural therapy for insomnia (Carney, Manber, Edinger), the circadian biology literature (Panda, Czeisler), and the practical work of clinicians who run CBT-I (Winter, Walker, Spielman). The synthesis is original to this edition.
This is educational, not clinical. It is designed for healthy adults working with ordinary, treatable sleep difficulties. If you suspect apnea, restless legs, or a sleep disorder, please see Chapter 03 and consult a clinician. This guide and real human support are not in competition; they're at their strongest together.
Set in Source Serif 4 for display and Outfit for body, with JetBrains Mono for labels. Twenty-two pages, Letter format. Designed to be printed and kept on a nightstand.
Same warm palette as its companion guides — cream, dusk blue, sage, honey. Dusk blue marks insight, honey marks risk, sage marks action.
Read it once. Run the 21-day plan (p. 16) as your literal calendar. Score the diagnostic (Ch. 02) on day 1, day 10, and day 21. Trust the trajectory more than any single night's score.
Sleep is not a thing you do well or poorly. It is the natural state of a body whose signals are unblocked. The work is removing the blocks.