How to Run a Fair Sleep Experiment on Yourself

Your watch says you got 48 minutes of deep sleep last night. The night before, it was 71. You had a glass of wine with dinner on one of those nights, so it's tempting to call it. Don't, at least not yet. Two nights tell you almost nothing, because deep sleep moves around for reasons you'll never see.

A small, deliberate sleep experiment can tell you something. You need one change, a clear baseline, enough nights, and a plan for reading the result before you look at it. Here's how to set one up so it's fair to the habit you're testing and honest with you.

Why a sleep experiment on yourself is a real method

Testing something on one person isn't just a quantified self hobby. Clinicians call it an N-of-1 trial, and there's a whole methods literature behind it. A review in the British Journal of Clinical Pharmacology by De Carvalho and colleagues sums up the core idea: each participant "serves as their control," measured at baseline and then during one or more treatment periods.

The review also flags the traps that matter most for sleep. One is carryover, when "treatment effects linger beyond the crossover." A late espresso on Tuesday could still be shaping Wednesday night. Another is autocorrelation, where one measurement influences the next. Sleep is full of it: a short night often brings a rebound night. The authors warn that ignoring this can make ordinary statistics too confident or not confident enough. They recommend setting the number and length of periods in advance, randomizing where you can, and masking when it's feasible.

You can't blind yourself to whether you ate a late dinner. But you can borrow the rest.

Know what your tracker can actually see

Before you test anything, it helps to know how well your device measures the thing you care about. Wearables are good at telling sleep from wake. They're much less sure about stages.

In a 2024 study in Sensors, researchers at Brigham and Women's Hospital had 35 healthy adults aged 20 to 50 spend one night in a sleep lab wearing an Oura Ring Gen3, a Fitbit Sense 2 and an Apple Watch Series 8, all compared against polysomnography (PSG), the lab standard. Every device caught at least 95% of sleep. Deep sleep was harder. Oura correctly flagged 79.5% of PSG deep sleep, Fitbit 61.7% and Apple Watch 50.5%. On average, Apple Watch came in about 43 minutes under PSG for deep sleep, while Oura's average didn't differ significantly. Two caveats: Oura funded the study and the lead author advises the company, and the Apple Watch recorded no data for 6 of the 35 participants.

An earlier 2022 study in Sensors from Central Queensland University tested six devices, including WHOOP 3.0, Oura Gen 2 and Apple Watch S6, on 53 healthy young adults over one lab night. Agreement with PSG across sleep stages ranged from 50% to 65%, with WHOOP at 60% and Oura at 61%. Both picked up 62% of deep sleep. The authors judged all six devices fine for sleep timing and duration but not yet good enough for specific stages. Their group receives research support from WHOOP, which they disclosed.

Both studies used a single night in a lab, so neither can tell you how consistent a device's error is across many nights in your own bed. That's why the instrument has to stay the same. Wear the same device the same way for the whole experiment. Switch from a ring to a watch halfway through and you've changed the ruler.

Choose one habit, and expect surprises

Pick a single change you can keep up for 14 nights. Lab evidence for a habit gives you a reason to test it, not a guaranteed answer.

Caffeine is a good example. In a small randomized, placebo-controlled crossover study in the Journal of Clinical Sleep Medicine, Drake and colleagues gave 12 healthy sleepers a large dose of caffeine at bedtime, 3 hours before, or 6 hours before, and measured sleep at home with a single-channel EEG headband. Even at 6 hours, caffeine cut objectively measured sleep by more than an hour. The drop people wrote in their diaries (41 minutes) didn't reach significance. In other words, they didn't fully notice, which is exactly why measuring beats guessing. Still, it was one big dose in a dozen people, and the authors called their sleep stage findings preliminary.

Alcohol shows why "deep sleep went up" isn't always good news. A 2013 review in Alcoholism: Clinical and Experimental Research by Ebrahim and colleagues found that in healthy volunteers, alcohol tended to increase slow wave sleep in the first half of the night, while sleep in the second half got more disrupted and REM was delayed. So if you test a drink with dinner, the deep sleep number on its own could make it look better than it is. Watch your wake time and how you feel too.

A fair 14-night protocol

Here's a setup that follows the methods research, scaled down to real life:

• Build a baseline first. Track your normal routine for about a month before changing anything, so you know your usual range and not just your average.
• Change one thing only. If you drop late meals and start evening walks in the same fortnight, you won't know which one mattered.
• Write the rule down before night one. "No caffeine after 2pm" is testable. "Less caffeine" isn't. Decide in advance what size of change would count.
• Keep everything else steady. Same device, same wrist or finger, similar bedtime and wake time. Log travel, illness, late work nights and drinks in a diary so you can explain outliers instead of quietly deleting them.
• Allow for carryover. Think about discounting the first night or two of a new habit. The N-of-1 literature suggests downweighting observations early in a period.
• Don't peek and quit. Stopping the moment the numbers look good is a reliable way to fool yourself. Finish all 14 nights.
• Rate how you feel. A quick morning score catches things a stage graph can't.

If you'd rather not run the spreadsheet yourself, this is the job Somnia Deep Sleep is being built for. It reads the stages your Apple Watch, Oura or Whoop already writes to Apple Health, compares a 14-night experiment against your own 30-night baseline, and labels the result Strong signal, Likely or Too early to tell. It's an iPhone app in beta, and your check-ins and experiments stay on the phone.

Reading the result honestly

When the two weeks are up, compare your experiment nights with your baseline, not with your best night ever. Ask three questions. Is the difference bigger than your normal spread? Did it hold across most nights, or does it come from one or two spikes? Does it survive a look at the diary, or does a weekend away explain it?

A shift of a few minutes sits well within what these devices can miss. A change that's large, consistent and matches how you feel deserves more trust. The best check is to repeat it. Switch the habit back, then try it again. Alternating periods like that is what turns a hunch into something closer to a real N-of-1 trial.

What the research can't tell you

The validation studies above each used one lab night with healthy adults. The Brigham authors note that this likely overstates how well devices do at home, and people with insomnia or other sleep disorders weren't included. The algorithms are proprietary and change with updates, so a result for one model may not hold for the next.

The habit studies have limits too. The caffeine trial had 12 people. The alcohol review covered lab studies of healthy volunteers, often at doses and timings that may look nothing like your Friday night. None of this research can tell you how your own body responds, which is the whole reason to test. Even so, a 14-night experiment can't fully separate your habit from stress, weather or the season. And a wearable can't confirm that the "deep sleep" on your screen is the same slow wave sleep a lab would score.

When to stop experimenting and see a doctor

Self-experiments are for fine-tuning habits, not for working around a medical problem. See a doctor if someone tells you that you snore loudly and gasp or stop breathing in your sleep, if daytime sleepiness affects your driving, or if insomnia has gone on for three months or more. The NHS also advises seeing a GP if changing your sleep habits hasn't helped, if you've had trouble sleeping for months, or if it's making daily life hard to cope with.

Sources

1. De Carvalho MCL, de Matos Dourado Simões M, Adler UC, Junior ABV, de Sousa CNS, Sanders LLO. N-of-1 trials in clinical research: Methodological foundations, statistical approaches and implementation challenges. British Journal of Clinical Pharmacology. 2026;92(3):809-821 (published online December 2025). PMC
2. Robbins R, Weaver MD, Sullivan JP, Quan SF, Gilmore K, Shaw S, Benz A, Qadri S, Barger LK, Czeisler CA, Duffy JF. Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults. Sensors. 2024;24(20):6532. PMC
3. Miller DJ, Sargent C, Roach GD. A Validation of Six Wearable Devices for Estimating Sleep, Heart Rate and Heart Rate Variability in Healthy Adults. Sensors. 2022;22(16):6317. PMC
4. Drake C, Roehrs T, Shambroom J, Roth T. Caffeine Effects on Sleep Taken 0, 3, or 6 Hours before Going to Bed. Journal of Clinical Sleep Medicine. 2013;9(11). PMC
5. Ebrahim IO, Shapiro CM, Williams AJ, Fenwick PB. Alcohol and sleep I: effects on normal sleep. Alcoholism: Clinical and Experimental Research. 2013;37(4):539-549. PubMed
6. NHS. Insomnia. National Health Service. nhs.uk

Curious what one habit really does to your deep sleep? Join the Somnia beta and test it on your own nights.

This article is general information, not medical advice.