What is an n-of-1 experiment? (And how to run one on yourself)

An n-of-1 experiment (or n-of-1 trial) is a controlled study with exactly one participant — you. The 'n' is the sample size in statistics, so n-of-1 literally means 'sample size of one.' Instead of asking 'does this work on average across thousands of people?', an n-of-1 trial asks the only question that affects your life: 'does this work for me?'

Why population averages fail individuals

A randomized controlled trial might show that a supplement improves sleep by an average of 8 minutes. That average hides enormous variation: it could mean everyone gained 8 minutes, or that half the people gained 30 minutes and half gained nothing. If you're in the 'nothing' half, the headline is true and still useless to you. n-of-1 trials exist precisely because individual response varies — genetics, baseline levels, lifestyle, and dozens of other factors mean the average rarely describes any single real person.

What makes it a real experiment, not just trying something

'I took it for a while and felt better' is not an n-of-1 trial — it's a guess wearing a lab coat. A real one has structure:

• A baseline period: you measure the outcome before changing anything, to know your true starting point.
• A defined intervention: one change, one dose, one schedule — not five new habits at once.
• An objective outcome: a number you can measure (sleep duration, HRV, resting heart rate) rather than a mood you can rationalize.
• Enough duration: long enough for the effect to appear and for daily noise to average out.
• A noise-aware comparison: a result only counts if the change exceeds your normal day-to-day variation.

A simple n-of-1 protocol you can copy

1. Pick one claim and one measurable outcome (e.g. 'magnesium improves my deep sleep').
2. Measure your baseline for about a week without changing anything.
3. Run the intervention for two to three weeks, keeping everything else as constant as you can.
4. Compare your intervention data against your baseline.
5. Decide: keep it, drop it, or run it longer if the result is borderline.

The hard parts are the statistics — establishing a fair baseline, deciding whether a change is real or just noise, and reporting an honest confidence level. That's the math Protocol handles for you, using the health data your phone already collects.