Journal · Inside Wave

Correlation is a question, not an answer

Track enough numbers and you will always find patterns, and many of them mean nothing. The fix is not more data. It is a small experiment on yourself.

· 5 min read · by Wave Wellness

Track twenty things about yourself and you have 190 pairs of them that could move together. Use the usual statistical bar, a 1-in-20 chance of a fluke, and you should expect nine or ten of those pairs to look meaningful by luck alone. Not because anything is wrong with your data. Because that is what happens when you look in enough places.

This is the quiet problem at the heart of every app that promises to find patterns in your life. More metrics do not just produce more insight. They produce more coincidence, and coincidence looks exactly like insight from the inside.

Patterns are cheap

There is a well-known collection online of real, tightly matched curves that have nothing to do with each other: cheese consumption tracking deaths from tangled bedsheets, that sort of thing. They are funny because the absurdity is obvious. The ones in your own record are not funny, because they are plausible.

Say your record shows that on days you drink coffee after 2 pm, your sleep is worse. Plausible. Caffeine lingers. But look at what else is true on those days. You drank the late coffee because you were tired, maybe because the night before was already bad. Or because it was a deadline day, and the deadline is what kept you up. Or because it was a Monday, and Sunday nights are rough for a lot of people regardless of what they drink.

Those are the three ways a correlation misleads. Chance: with enough pairs, some line up for no reason. A third thing: stress, a schedule or a season drives both numbers. The arrow points backward: being tired causes the coffee, not the other way round.

A personal record is especially prone to the last two, because your behavior is not random. You choose what to do based on how you feel, and how you feel is exactly what you are measuring. People train less when they are run down, so “workout days have better HRV” can mean the workout helped, or that you only worked out on days you were already recovered. The data alone cannot tell those apart.

None of this means patterns are worthless. It means a pattern is where a question begins.

The experiment of one

Medicine has a tool for exactly this, and it scales down to a single person. It is called an n-of-1 trial: one person, one change, switched on and off on a schedule, with the outcome measured the same way throughout. Clinicians have used the design for decades to settle questions like which of two treatments suits a particular patient, because the average result from a big trial does not always describe the person in the chair.

You can run the everyday version without a clinic. The rules are simple, and they are what turn a correlation into evidence:

  1. Change one thing. Coffee cut off at noon, not coffee plus an earlier bedtime plus a new pillow.
  2. Decide the measure in advance. Pick the number you will judge it by before you start, so you cannot shop for whichever one moved.
  3. Switch it on and off. A week with, a week without, a week with again. If the effect comes and goes with the change, it is far harder to blame Mondays or a busy month.
  4. Give it long enough. One good night proves nothing. Several weeks of each condition starts to.

What makes this work is that you control the timing. In the raw record, you chose when to drink the late coffee, and your reasons got tangled into the result. In the experiment, the calendar chooses, and the tangle comes undone.

How Wave uses it

Wave Wellness does not hand you correlations dressed up as conclusions, and that is deliberate. When something looks connected, say shorter sleep after late workouts, the honest move is to treat it as a question, not a law about your body. It is a candidate for a test, and the test is yours to run: pick one small change, hold it for a week, and look.

You can put the question to Ask, which answers from your own record and nothing else: how did I sleep on the nights after a late session, compared with the others? It describes what the numbers show. It does not tell you the cause, because one person’s record cannot.

The rest of the record is what makes the test readable. Today’s Wave reads your HRV, resting heart rate and sleep against your own 30-day baseline every morning, the same way each time. The WellnessScan asks the same question the same way: forty seconds, the same paced breathing, the same signals. The ten-second check-in records how you felt on a 1-to-5 scale where 5 is always better. A measure that is taken identically every day is what lets a week with the change be compared fairly with a week without it.

Then you look. Did the number you chose in advance move when the change was on, and move back when it was off? If so, you have something much better than a correlation: a result about you, produced by you.

The limits, plainly

Insights finds patterns; it cannot certify causes, and it does not pretend to. Some correlations it surfaces will be coincidence, simply because of the arithmetic at the top of this piece. Wave does not randomize your weeks or run the experiment for you. That part is yours, and it takes some patience. And a self-experiment answers a narrow question, about one change, in one person, over one stretch of time. It will not tell you what is true for everyone, and it was never meant to.

That is the right size for the question, though. You do not need to know what works on average. You need to know what works on you, and the honest way to find out is to treat every pattern as a question you can test.

Wave Wellness is free on the App Store and Google Play. Download it and see all of it working together today.

Wave is for general wellness and is not a medical device. Its insights are prompts from your own data, not medical advice; talk to a doctor before changing anything about a condition or a medication.

See it in the app: Ask

See this number in your own record.

Wave reads it every morning and tells you, in plain English, what changed.