Your Data Knows Before You Do

After months of tracking my mood and sleep, I found patterns I never would have caught through introspection alone. Some of them changed how I manage my condition.

I used to think I knew myself pretty well.

I’ve spent years in therapy. I journal. I reflect constantly — probably too much, honestly. I can tell you my triggers, my patterns, my warning signs. Or at least I thought I could.

Then I started tracking my mood and sleep with actual numbers, consistently, for months. And the data told me things I had completely wrong.

The correlation I expected vs. the one I got

Here’s what I assumed going in: sleep and mood are tightly linked. Sleep badly, feel bad. Sleep well, feel good. Simple, obvious, everyone knows this.

Turns out, the correlation between my sleep duration and my mood the next day is weak. Like, surprisingly weak. Around 0.3 on a scale where 1.0 would be perfect correlation. If you showed that number to a statistician they’d say “barely there.”

But wait — I know sleep matters. I’ve experienced it. Every major crash I’ve had was preceded by bad sleep. How can the correlation be weak?

Because sleep doesn’t predict mood directly. It predicts capacity. My energy-to-mood correlation is strong — around 0.68. And my sleep quality (not just duration) to mood correlation is moderate. But raw hours of sleep? Weak direct link.

What this means practically: I can sleep 6 hours and be fine, if those 6 hours were solid. I can sleep 9 hours and feel terrible, if it was fragmented and restless. The number I was tracking — hours in bed — was the wrong metric. What mattered was sleep quality, and its downstream effect on energy, and energy’s effect on mood.

I never would have figured that out through introspection. I would’ve kept saying “I need 8 hours” and wondering why some 8-hour nights still led to bad mornings. The data showed me I was solving the wrong equation.

The thing about sequences

One of the more useful things I’ve learned from tracking is that mental health events aren’t random — they’re sequential. And the sequence matters more than any individual metric.

Here’s a pattern I’ve identified in my own data:

Sleep starts fragmenting (quality drops even if hours stay the same)
Energy gets weird — either unusually high or weirdly flat
Mood follows, usually with a 24-48 hour delay
Stability drops — I start reacting to things instead of responding
Something happens — a conflict, a mistake, a decision I wouldn’t normally make

By step 5, it feels like the event caused the problem. “I had a terrible day because this happened.” But the data shows the cascade started at step 1, usually 3-4 days before the “event.” The event was just the thing that finally broke through the weakened defenses.

This is why daily tracking matters. Not because any single day’s numbers are revelatory, but because the sequence of days tells a story that you can’t see in real time. You’re living inside the sequence. You need the data to see it from outside.

Memory is a terrible historian

Ask me how last month was and I’ll give you a general impression. “Pretty good, mostly stable, a couple of rough days.” That’s my honest best recollection.

Then I look at the actual data and it tells a different story. Maybe there were four rough days, not two. Maybe the “pretty good” stretch had a slow downward trend that I didn’t register because each day was only slightly worse than the last. Maybe the energy spike I remember as “a productive week” was actually the early phase of hypomania, and if I’d caught it then, I could’ve avoided what came after.

This is a well-documented psychological phenomenon — we remember peaks and endpoints, not averages. It’s called the peak-end rule, and it makes us unreliable narrators of our own mental health.

I’ve caught myself doing this with my psychiatrist. “How have the last two weeks been?” And I’ll say “fine, mostly” because the last couple of days were okay and that colors everything. Meanwhile the data shows a clear mood dip in the middle of that period that I’ve already forgotten about.

This isn’t about being dishonest. It’s about the basic architecture of human memory being poorly suited to the task of monitoring a chronic condition. We’re just not built for it. That’s what data is for.

What the tags taught me

One thing I track alongside the numbers is tags — simple labels for what was going on that day. Work stress. Relationship stuff. Exercise. Social time. Travel. Nothing fancy, just context.

After a few months, I looked at which tags were associated with my highest moods and which with my lowest. Some were obvious: days tagged “exercise” averaged higher mood. Days tagged “work stress” averaged lower. No surprises there.

But some were not obvious at all. Days I tagged “reflection” or “self-awareness” — days where I spent time thinking about my mental state — averaged higher mood than days I tagged “productive.” That’s backwards from what I would have predicted. I assumed productive days were my best days. The data said otherwise. My best days were the ones where I had space to think, not the ones where I got the most done.

That changed how I structure my weeks. I stopped trying to optimize for maximum output and started protecting time for nothing. Just thinking. Walking. Reviewing logs. And the data confirmed that shift was working.

The scary pattern

I’m going to be real about something. Tracking data about your mental health can also show you things you’d rather not see.

There was a period in my data where I could clearly see a depressive episode building over about two weeks. The mood scores were declining, the sleep was getting longer (hypersomnia is a depression signal for me), the energy was flat. And I could see it in the chart, clear as day.

The unsettling part wasn’t the episode itself — I’ve been through those before. The unsettling part was realizing that the data saw it before I did. I was in the middle of it, thinking “I’m just tired, it’s been a long week,” while the chart was essentially screaming that this was a pattern, not a circumstance.

On one hand, that’s exactly what the data is supposed to do. On the other hand, seeing your own decline mapped out in a graph is a specific kind of uncomfortable that I wasn’t prepared for.

But here’s where I’ve landed on it: uncomfortable and informed is better than comfortable and blindsided. If the data shows me something’s coming, I can adjust. I can talk to my doctor. I can protect my sleep. I can clear the schedule. I can’t do any of that if I don’t see it.

You don’t need a lot of data

One thing I want to push back on is the idea that you need months of tracking before this is useful. You don’t.

Even two weeks of consistent daily tracking — mood, sleep, energy — will start showing you things. It won’t be statistically rigorous and it won’t catch long-cycle patterns, but it’ll be more than you had before. And for a lot of people, the simple act of putting numbers on their daily experience is itself a revelation.

You don’t need to be a data person. You don’t need to understand correlations or statistical significance. You just need to write down how you feel, how you slept, and how much energy you have, every day, for a couple of weeks. Then look at it together.

I guarantee you’ll see something you didn’t know.

The punchline

Your brain is doing its best. It’s also lying to you. Not maliciously — just structurally. It forgets the bad days, smooths over the patterns, and presents you with a narrative that’s simpler and more flattering than reality.

Data doesn’t do that. Data just sits there, accurately, being uncomfortable and useful.

That’s why I built Steadyline around data first, feelings second. Not because feelings don’t matter — they’re the whole point. But because understanding your feelings requires seeing them clearly, and clearly is exactly what your brain won’t do on its own.

I’m a healthcare software engineer living with bipolar disorder. I track my mental health daily and I’m building the tool that makes it possible. More at steadyline.app.