I Tracked My Sleep for 90 Days: Here's What the Data Actually Shows

Source: Dev.to

Three months ago I started an experiment. I wore a sleep-tracking wearable to bed every night and kept a handwritten sleep diary on my nightstand. I logged when I got into bed, when I fell asleep, any nighttime waking, and how I felt on a 1-10 scale each morning. Ninety days later, the patterns were clear, and a lot of them contradicted what I assumed going in. The single most impactful change across the entire 90 days was not sleeping longer, not taking supplements, and not buying a better mattress. It was waking up at the same time every day, including weekends. For the first three weeks I kept my normal pattern, which meant waking up at 6:30 on weekdays and sleeping until 8:30 or 9 on Saturday and Sunday. My average subjective energy score during that period was 5.2 out of 10. Starting in week four, I locked my wake time to 6:30 every single day. Within about ten days, my average energy score climbed to 7.1, and it stayed there for the remaining weeks of the experiment. Your circadian rhythm is anchored primarily by your wake time, not your bedtime. Shift that anchor by two hours on weekends and you give yourself jet lag every Monday morning. Sleep researchers call it “social jet lag,” and by one estimate it affects two-thirds of adults in industrialized countries. Bedtime consistency mattered too, but it was clearly secondary. A fixed wake time with bedtime varying by 30-45 minutes still produced good results. A variable wake time with a consistent bedtime did not. You have probably heard that sleep cycles last 90 minutes, and that you should plan your sleep in multiples of that number. I believed this for years. The data told a different story. My wearable tracked sleep stages, and while I take the exact classifications with a grain of salt, the cycle-length data was consistent. My average cycle was closer to 80 minutes, not 90. Published research shows cycles range from 70 to 120 minutes and vary not just between people but within a single night. Your first cycle is typically shorter with more deep sleep. Later cycles are longer with more REM. Rigid “sleep in 90-minute blocks” planning is based on an approximation that might not match your biology. A better approach: aim for 7-9 hours and pay attention to how you feel waking at different times within that range. You will find your own natural cycle length within a few weeks. During weeks five through eight I experimented with room temperature using a programmable thermostat, setting different temperatures on different nights. The data was unambiguous. Nights between 65 and 67 degrees Fahrenheit (18-19 Celsius) consistently showed longer deep sleep phases and higher morning energy scores. Above 70F, deep sleep dropped noticeably. Below 63F, I woke up more frequently. Your core body temperature needs to drop about 2-3 degrees Fahrenheit to initiate and maintain sleep. A cool room facilitates that drop. A warm room fights it. I had always kept my bedroom around 72F because it felt comfortable while awake. But the temperature that feels comfortable when you are conscious is not the temperature that produces the best sleep. I am a coffee drinker. Two cups in the morning, and I used to have a third around 2pm when the afternoon slump hit. During the experiment I tested cutting that afternoon cup. Caffeine has a half-life of roughly 5 to 6 hours, which means that a 2pm coffee is still approximately 25% active in your system at midnight. That does not sound like much, but 25% of a cup of coffee is enough to measurably reduce deep sleep and increase nighttime awakenings, even if you feel like you fall asleep just fine. When I dropped the afternoon coffee in week six, my sleep onset time (the gap between getting into bed and falling asleep) decreased by an average of 12 minutes, and my deep sleep percentage went up by about 8%. I replaced the afternoon coffee with a short walk outside. The daylight exposure probably helped my circadian rhythm as a bonus. This one surprised me the most. I spent two weeks using blue-light blocking glasses and enabling night mode on all my devices in the evening. Then I spent two weeks using my devices normally with full brightness. The difference in my sleep metrics was negligible. The studies that originally linked blue light to sleep disruption used light intensities far higher than what a phone or laptop produces. A 2019 study from Brigham Young University found that screen brightness matters more than color spectrum. A dim screen with blue light is less disruptive than a bright screen with blue light filtered out. What did matter was a consistent pre-sleep routine. The nights where I followed the same sequence in the 30-45 minutes before bed produced better sleep regardless of screen use. The routine was simple: make tomorrow’s to-do list, read for 15 minutes, brush teeth, lights out. The content mattered less than the consistency. My brain learned the sequence as a shutdown signal. On weekends where I had two drinks in the evening, REM sleep dropped by roughly 40%. I fell asleep faster, which is the deceptive part, but the sleep architecture was measurably worse. Less REM means less memory consolidation, less emotional processing, and a higher chance of waking up unrested despite enough total hours. Even one drink showed a visible effect. This matches what sleep researchers like Matthew Walker have published for years. Not a moral judgment. Just what the numbers said. Naps before 2pm lasting 20 minutes or less were genuinely restorative. Afternoon energy scores were consistently higher on nap days. But anything longer than 20 minutes produced sleep inertia, that groggy feeling that lasts 30 minutes or more after waking. Naps after 3pm made it harder to fall asleep at my target bedtime. The ideal nap was around 1pm, 15-20 minutes, alarm set. Without an alarm I would reliably sleep 40-50 minutes and wake up worse than before. After 90 days, three changes stuck: fixed wake time every day, bedroom at 66F, and

no caffeine after noon. Everything else was either marginal or unsustainable. Those three changes took my average morning energy from 5.2 to 7.4 out of 10. Not perfect, but a meaningful improvement in how every single day starts. If you want to experiment with your own sleep timing without doing all the math on cycles and wake windows, I built a sleep calculator that helps you find optimal bed and wake times based on when you need to get up. It is a decent starting point before you figure out your own patterns. The real lesson from 90 days of tracking was that sleep is not one big thing you fix. It is a handful of small, boring habits you get consistent about. The data does not lie, even when it contradicts what you assumed. I’m Michael Lip. I build free tools at zovo.one. 350+ tools, all private, all free.