Why "it worked for me" fools you

Anecdote sits at the bottom of the pyramid not because personal experience doesn't matter, but because a sample of one has no defences against the things that fool everyone. When you try a supplement, at least four forces conspire to manufacture a "success":

  • No control. You have no idea what would have happened if you'd done nothing. There's no placebo-you running in parallel.
  • The placebo effect. Believing you're being helped produces real, measurable improvement — especially in symptoms you rate yourself: pain, energy, mood, sleep.
  • Confirmation bias & memory. You notice and remember the days it "worked" and quietly forget the rest.
  • Regression to the mean. The big one, and the least intuitive — so it gets a picture.

Regression to the mean

We start supplements when we feel worst — the peak of a cold, the flare of a joint, the low point of a slump. But extremes are, by definition, unusual; things tend to drift back toward your normal on their own. Take something at the trough and the natural rebound gets credited to the pill.

Symptom severity fluctuating over time, with a supplement taken at the worst point symptom severity time → your normal you start the supplement you feel better — but you'd get there anyway
You act at the worst moment, so improvement is nearly guaranteed. Symptoms swing around your normal. You reach for a remedy at the peak — the only rational time to — and the inevitable drift back toward baseline feels like proof it worked. It would have happened with a sugar pill, or nothing at all.

What a personal experiment can tell you

Here's the joyful part: a sample of one isn't worthless — it's just usually run badly. For the right kind of question, a careful "n=1" trial is genuinely useful. It works best for effects that are subjective, fast-acting and reversible: does caffeine after 3 pm wreck your sleep? Does this pre-workout give you the jitters? Does whey upset your stomach? You can actually test these on yourself if you borrow the RCT's tricks:

  • Blind yourself if you can. Identical capsules, one real, one placebo, shuffled by a friend so you don't know which week is which.
  • Pick your outcome in advance and write it down — "hours of sleep", "1–10 gut comfort" — so you can't move the goalposts later.
  • Use a washout. Give the last thing time to leave before testing the next.
  • Repeat. One good week proves little; alternate a few times and look for a consistent pattern.
⚠
What n=1 can never tell you

Anything slow, invisible or statistical: long-term disease risk, whether it's quietly protecting your heart, rare side effects. You cannot feel your cholesterol or your cancer risk change. For those, you're back to needing trials — no amount of self-tracking substitutes.

Worked example The vitamin C “it cured my cold” story

You feel a cold coming on, take a big dose of vitamin C at the worst moment, and three days later you're fine. Obvious conclusion: it worked.

But colds resolve on their own in a few days — you started the vitamin exactly when you felt worst, so recovery was coming regardless (regression to the mean). Add the placebo effect and the fact that you'd never notice the times it didn't help, and you have a rock-solid personal belief built entirely on the four traps above. The large RCTs agree: routine vitamin C doesn't meaningfully prevent colds, and barely shortens them. Your experience was real; your conclusion was the mirage.

So take your own experience seriously as a signal — it's the reason to test something — but not as proof. And when the effect is fast and you can feel it, don't just wonder: run the little blinded experiment. It's a genuinely fun way to out-think your own biases.

What to remember

Take your experience as a signal, never as proof.

  • Four traps rig “it worked for me”: no control, the placebo effect, confirmation bias, and regression to the mean — you act at your worst moment, so rebound is nearly guaranteed.
  • n=1 is useful for fast, subjective, reversible effects — sleep, jitters, gut comfort — if you blind yourself, pre-pick the outcome, wash out, and repeat.
  • n=1 can't touch slow or invisible outcomes — disease risk, rare harms. You can't feel those change; for them you still need trials.
Try it · ~1 min

Design a tiny honest self-experiment.

  1. Pick one fast, feelable question ("does X improve my sleep?").
  2. Write the outcome you'll measure before you start, and decide how you'll blind it.
  3. Plan to alternate and repeat. Notice how much harder this is to fool than "I took it and felt great".

Next in this path

Lesson 07Putting it together — how much evidence to demand Lesson 02RCTs — the blinded, controlled version of your self-experiment

References

1
Barnett AG, van der Pols JC, Dobson AJ. Regression to the mean: what it is and how to deal with it. Int J Epidemiol. 2005;34(1):215–220. doi:10.1093/ije/dyh299
2
Hemilä H, Chalker E. Vitamin C for preventing and treating the common cold. Cochrane Database Syst Rev. 2013;(1):CD000980. doi:10.1002/14651858.CD000980.pub4

Progress saves automatically in this browser · skip ahead anytime · no streak.