Self-tracking

Does that supplement actually work? How to tell with your own data.

You started magnesium three weeks ago. You feel like you're sleeping better. But "I feel like" is exactly the sentence that fools people — including careful, skeptical people. Here's how to actually find out.

Published July 6, 2026 · Updated July 20, 2026 · 7 min read

Every supplement aisle promises something. Every friend has a story about the one that changed everything. And every one of us has, at some point, taken something for a few weeks, felt a little different, and quietly decided it worked — without ever asking what "worked" would even look like as evidence.

This isn't a piece about which supplements are legitimate and which are snake oil. It's about a much narrower, much more useful question: once you've decided to try something on yourself, how do you tell — honestly, without kidding yourself — whether it did anything?

Why "I feel like it's working" isn't evidence

Three things conspire against you the moment you start paying attention to how you feel after starting something new:

• Regression to the mean. People tend to start a new supplement when they're feeling worse than usual — that's often the reason they started. Bad stretches end on their own. If you begin tracking during a rough patch, you'll almost always see things improve next, whether or not the supplement did anything.
• Confirmation bias. Once you've spent $30 and three weeks on something, you're motivated to notice the good days and explain away the bad ones. This isn't a character flaw — it's just how attention works.
• Confounds. Real life rarely holds still while you run an experiment on yourself. New supplement, but also: the weather changed, work got less stressful, you started sleeping with the window cracked. Any of those could be the actual cause.

None of this means the supplement isn't working. It means a good feeling, on its own, can't tell you either way.

The fix: treat yourself like an experiment of one

You don't need a lab, a control group, or a statistics degree to do this reasonably well. You need four things: a baseline, a clear start date, an outcome you're actually measuring, and enough days on both sides to tell signal from noise.

1. Pick one outcome, not a vague feeling

"Better" isn't measurable. Pick something you can rate the same way every day: sleep quality on a 1–5 scale, hours of sleep, a mood or energy rating, resting heart rate, HRV if you have a wearable. The specific number matters less than picking one and rating it consistently — the same scale, the same rough time of day, every day, whether or not you remember to think about the supplement that day.

2. Log the two weeks before you start

This is the step almost everyone skips, and it's the one that matters most. Without a real "before," there's nothing to compare the after to — you're just left with a vague sense of "before" reconstructed from memory, which is exactly what regression to the mean and confirmation bias feed on. Two weeks is a practical example, not a scientific threshold; the important part is collecting enough repeated days to see your ordinary variation.

3. Mark the actual start date — and don't change anything else

Write down the day you actually started, not "sometime in March." And to the extent you can control it: don't start a new sleep routine, a new workout split, and a new supplement all in the same week. If you do, and something changes, you'll never know which of the three actually caused it.

4. Give it real time on the other side — then look at the average, not the anecdotes

One great night three days after you start proves nothing; neither does one bad night. What you want is roughly the same window as your "before" — another two weeks or so — averaged, then compared. Was the average meaningfully different? Not "did I have one incredible Tuesday," but did the whole two-week stretch move.

What this looks like in practice

Concretely: two columns in a notes app or a spreadsheet, one row per day, one number per day for whatever you're tracking. Two weeks of "before" numbers, a marked start date, two-plus weeks of "after" numbers. Average each side. Compare.

It's not complicated. It's just tedious enough that almost nobody actually does it by hand for more than a few days — which is the entire reason most people end up relying on "I feel like it's working" instead of an actual before-and-after.

---

Why we built Traceline

This is the work Traceline makes easier. For one planned supplement change, it records a baseline and the actual start date, then compares your logged Sleep Quality across usable days before and after the change. With permission, sleep duration, HRV, and resting heart rate can sync from Apple Health, while the other signals you log help preserve the context around the result.

It does not run a clinical trial or prove that a supplement caused a change. It keeps the fixed windows, coverage, and data strength visible, separates planned Trial Results from less-controlled Observed Trends, and leaves room for the honest answer when the evidence is thin.

Sources and further reading

• Barnett et al.: regression to the mean in health research
• BMJ: CONSORT extension for formal n-of-1 trials