The Experiments and Data Behind Fixing My Deep Sleep

Cross-posted from: https://rohitmalekar.in/deepsleep

Deep sleep has been the single most influential lever in shifting my mornings, from starting the day weary and drained to waking up fresh, rested, and hopeful. It's the phase of sleep most tied to physical repair, immune health, and cognitive recovery. Over the past year or so, I’ve been experimenting with small, deliberate changes to improve deep sleep.

Rather than tackling everything at once, I layered one habit at a time: eating earlier, going to bed earlier, exercising consistently, and finally stabilizing bedtime routines. Each experiment built on the last, and the data tells a compelling story of how these habits stack and compound. None of these are novel insights, but seeing them play out in my own data is a great form of reinforcement to keep going.

I used Amazfit (Zepp) and Strava to track the underlying data and combined the exports to hack together these insights.

Experiment 1: Early Dinner & No Late-Night Snacks (60 days)

I started with food timing: dinner done by 7:30 PM and eliminated post-dinner snacking.

Before this change (blue dots), most nights fell short of 1 hour of deep sleep, a benchmark often associated with restful, restorative sleep cycles. After the change (green dots), the pattern flipped and nights above 1 hour became dominant.

What struck me was how immediate and reliable this was. This wasn’t a gradual adjustment. It was more like flipping a switch. It reinforced a simple truth: late-night eating forces the body to prioritize digestion over deep repair.

This was my first "algorithmic" result—no ambiguity, no guesswork. The takeaway? Give your body a head start before sleep, and it will use the night for recovery, not metabolism.

Experiment 2: Earlier Bedtimes Unlock More Deep Sleep (6 months)

Once eating habits stabilized, I turned to bedtime. I suspected that timing mattered, even if total sleep hours stayed the same.

The boxplot confirmed it:

• 10–11 PM bedtimes consistently yielded the highest deep sleep.
• Pushing sleep past 1 AM slashed deep sleep, regardless of duration.

Even on nights when I slept 7–8 hours starting at 2 AM, deep sleep never fully recovered. The takeaway? You can’t out-sleep a late bedtime. The circadian window for deep sleep doesn’t simply shift forward.

This reframed how I see sleep. It’s not just a bank account where you “deposit” hours. Timing matters because deep sleep is front-loaded early in the night. Miss the window, and it’s gone.

Experiment 3: Exercise’s Lagged Effect on Deep Sleep (6 months)

Next came exercise, but I wanted to know: does one workout help tonight’s sleep, or is the effect cumulative?

By looking at 3-day rolling exercise calories, the answer became clear:

• Correlation was stronger than a single-day lag.
• Deep sleep rose steadily with sustained activity, not one-off bursts.
• High-intensity weeks (>600 cal/day avg) produced ~20% more deep sleep than low-activity weeks.

This explained why "exercise helps sleep" sometimes feels inconsistent. A single run won’t undo a sedentary week. Deep sleep seems to track not with the last workout but with an ongoing signal: "This body is in use, prepare it for more."

Experiment 4: Bedtime Consistency Compounds the Gains (6 months)

Finally, I looked at bedtime consistency, measured as the standard deviation of bedtime over a rolling 7 days.

Pairing bedtime regularity with exercise intensity revealed the strongest outcomes:

• Moderate consistency in bedtime routine + high exercise: Deep sleep averaged 1.49 hours—the best results.
• Even highly consistent bedtimes with low intensity exercise held deep sleep above 1 hour.
• But erratic bedtimes erased gains, even when I exercised hard.

This was the “keystone” realization: you can’t out-train bad sleep habits. Exercise supports deep sleep, but only if you also give it a predictable window to work within.

The Habit Stack That Worked

Looking back, each experiment naturally built on the previous one:

1. Better eating habits first → Created lighter evenings and easier digestion.
2. Earlier bedtimes next → Synced sleep with circadian deep-sleep windows.
3. Consistent exercise added → Primed the body for recovery, raising baseline deep sleep.
4. Bedtime consistency reinforced it all → Locked in a rhythm where gains stopped eroding.

The order mattered. Had I tried exercising hard while eating late and sleeping irregularly, it would’ve failed. Instead, each layer supported the next, creating momentum.

Final Thoughts

Six months in, I see deep sleep less as an isolated “metric” and more as the trailing indicator of balanced rhythms: eat earlier, wind down earlier, move regularly, and keep bedtimes steady.

The biggest shift for me is that the mornings just feel different. Waking up isn’t a negotiation anymore; it’s clear-headed and without the fog of incomplete rest. Deep sleep isn’t won with one hack—it’s a cascade effect. Start with the simplest lever, stick with it, and let the improvements compound.

A few caveats:

• Physiological differences, genetics, and lifestyle context (stress, work patterns, etc.) mean the same interventions might yield different results for you.
• Sleep stage tracking can be noisy, especially distinguishing light vs deep sleep. Trends are more reliable than absolute values.
• I haven't tracked external factors (stress, caffeine, screen time) that can strongly influence sleep quality.