More Sleep Data ≠ Better Sleep (Yet...)

A new New Year's Eve tradition is emerging in my life.

Export all the data from my devices. Crunch the numbers, slice and dice the data, review the year. Sleep, activity, HRV, mood, exposure to sunlight. All of it. It's my version of watching the ball drop.

This year, I focused on sleep in particular, looking back over the entirety of the quantified-self dataset I’ve been collecting for several years. Over 1,000 nights of data across my Oura Ring, WHOOP, Apple Watch, and Eight Sleep. I can tell you my average bedtime to the minute (12:38 AM). My sleep latency by day of the week. The precise mix of sleep stages. My sleep efficiency to the decimal point.

What's the verdict?

Six additional minutes of sleep duration per night and +2.4 points on Oura’s sleep score (out of 100). Three years of tracking bought me… almost nothing.

Years of dashboards, weekly reviews, and relentlessly learning how to "optimize my sleep performance" bought me six minutes per night.

Data has worked for me and made my life better in many ways. Just… not for sleep.

My resting heart rate has meaningfully dropped over the same period. My HRV is up. My marathon times have improved. The feedback loop is real. Information is power. Data in, adjustment, better output.

So why hasn't it worked for sleep?

I think sleep is different. Sleep is… weirdly hard to improve with the same playbook. And I have some ideas about why. My hunch: you can't perform, decide, or willpower your way to better sleep. I believe the right path is designing an environment where you just don't have to try so hard.

Most things you optimize with data involve doing something. Run more. Lift heavier. Eat cleaner. You get feedback, you adjust, you improve. There's agency in the loop.

Sleep is the opposite. You can't do sleep better. You have to stop doing other things. You improve workouts by doing more. You improve sleep by doing less. 

And yet look at the language these products use. WHOOP literally calls it "Sleep Performance." Oura gives you a score out of 100. Every morning you wake up and get graded on how well you rested… outsourcing your internal subjective sense of how you feel to this piece of plastic on your finger and the numbers you check from it first thing in the morning.

You're not on a Broadway stage. You're not competing in an Olympic event. You're not in a critical boardroom meeting. This is not a performance. You're supposed to be letting go. For me, I'm good under pressure… and pressure helps me perform! But pressure makes it harder to rest.

Did I hit my sleep score goal? Am I getting enough deep sleep? REM? The anxiety of optimization applied to rest. (And do we… actually know what all these things mean?)

There's actually a term for this: orthosomnia. Anxiety caused by obsessing over your sleep data. The thing that's supposed to help you sleep better… quietly making your sleep worse. It’s real, and if you’re even subtly feeling it, you are not alone.

We're treating sleep like a sport when it's actually the recovery from sport. The mindset that works for everything else backfires here. What if the answer isn't performing better, but designing an environment where you don't have to perform at all?

When you try to “optimize” sleep, every day becomes a series of decisions. Think about investing. Active investors obsess over every stock pick, every market movement, every daily fluctuation. Passive investors pick an index fund and forget about it.

Even if active investing has a slight edge in performance (debatable, at best), the time, energy, stress, and mental bandwidth required is enormous. For a lot of people, it's just not worth the cognitive cost.

I wonder if we're “active investing” in our sleep. You wake up, check your score, and find out if you're "up" or "down" today. Red numbers, green numbers. It turns every morning into a market open and sets the tone for how we feel about the day ahead. Literally… could there be a more anxiety-inducing term than “sleep debt”??

And sleep is uniquely brutal here. Exercise essentially has a single decision point: go or don't go. Nutrition decisions often cluster around meals. But circadian regulation is relentless: morning sunlight exposure, afternoon caffeine cutoffs, evening wind-downs, screen time limits, meal and hydration timing. 

It's not one choice. It's dozens of microdecisions, scattered across the day, each one invisible, each one easy to rationalize away. And unlike skipping the gym, you don't feel the consequences until hours later — if you can even trace them back at all.

The mental overhead of constant optimization is exhausting. And the decisions we make at midnight when we're tired are probably worse than just following a simple default anyway.

Defaults beat decisions. That’s the whole point. Smart choice architecture and “automating” daily microdecisions can go a long way. 

I learned about the “last mile problem” during my time at Amazon.

In logistics, the last mile is the final stretch — getting the package from the warehouse to your door. It's often the hardest, most costly part of the whole delivery. Everything else can be optimized, automated, scaled. But that last mile? A tricky grind. For sleep, the last mile is the willpower to actually do the things to sleep better. The hard work of integrating new habits into your life even when you have all the info on what you’re supposed to be doing. Habit change is hard, and life often gets in the way.

The quantified self, especially with sleep, has a last mile problem. We collect the data. We generate the insights. We generally know what we're supposed to do. We're 80% of the way there. And then… it stops. We are data rich, habit poor. And sleep habits have a huge effect on how we sleep.

For many years, the problem hasn’t been knowledge. It’s action. The quantified self world feels like the wild west. We're all collecting crazy amounts of data and just kinda… hoping it turns into behavior change. The wearable companies show us charts and scores. The apps give us "insights." We can dump this stuff in ChatGPT to get some more specific answers. But the hard part is still entirely on us. And nobody's solved that yet.

Data is great at diagnosis. It can tell you exactly what's wrong... But it can't make you do anything. Sleep is an orchestration problem. And right now, we are all our own orchestra conductors, left alone to conduct our circadian rhythms.

If you're tracking your sleep, I have an honest question: has it actually gotten better? Not your dashboards. Your actual sleep. And how rested and ready you feel each day.

If the answer is no, you might be where I am. And where a lot of the people I’ve talked to about their sleep are. But I think we are on the brink of making progress…

This is why I'm building Remi. The personal assistant for effortless, consistent, high-quality sleep.

Not because I've solved all these problems — clearly I haven't! But because I want something that actually helps me throughout the day and at critical moments like midnight, not just something that grades me harshly at 8 AM. And I’ve heard similar feelings and challenges around this from hundreds of people. There’s a real need here, and I intend to work hard to find and build a truly helpful solution. 

The vision is that Remi acts on your behalf so you don't have to stress about the data, know exactly what to do with it, or do all the hard work of habit change yourself. Remi takes these things off your plate and sets better defaults and nudges at the moments that matter.

I'll be sharing more thoughts on the startup, the quantified self, and my learnings from my past year of data throughout January. Learnings from 2025, intentions for 2026, and maybe some answers. Or at least better questions. Thanks for following the journey!