Is My Wearable Reading Wrong? How to Tell an Artifact From a Signal

Some mornings the number on your wrist is not a weak signal. It is an artifact. The sensor recorded something real, and that something was not you.

This is an uncomfortable page for a company built on personal health data to publish, and it is the honest one. Every measurement carries error. Knowing which readings to set aside is not a way of avoiding your data. It is a precondition for using it.

An artifact is not the same as a bad night

The distinction matters, because the two feel identical at 7am.

A bad night is signal. You slept four hours, your resting heart rate is up, your heart-rate variability is down, and the reading is an accurate description of a real state. Unpleasant, but true.

An artifact is noise wearing the clothes of signal. The band sat loose against your wrist. Your hands were cold and peripheral circulation dropped. You shifted for an hour and the accelerometer read restlessness that was really a bad mattress. The device dutifully recorded what reached the sensor, and what reached the sensor was not a faithful account of your physiology.

Acting on the first is sensible. Acting on the second is a real decision made on a number that was never about you.

Four cases where a reading deserves to be set aside

None of these is a rule. Each is a reason to hold the number loosely until something else confirms it.

The value sits far outside your usual range and nothing upstream explains it. Bodies move, but they rarely leap. If a metric lands well outside where it has lived for weeks, and the preceding two or three days contain no late night, no unusual load, no illness, no alcohol, no travel and no time-zone change, the most economical explanation is often the measurement rather than the body.

One variable moved and the others held steady. Physiology is correlated. A genuine strain response tends to show up in more than one place at once. When heart-rate variability collapses while resting heart rate, respiratory rate, sleep duration and temperature all sit exactly where they were, you are usually looking at one noisy channel, not a body in trouble.

Sensor contact was poor. This is the most common and the most under-reported cause. A loose band, a strap worn higher or lower than usual, a cold room, a ring turned on the finger, dry skin, tattoos under an optical sensor. Optical heart-rate sensing depends on light reaching tissue and returning. Anything that interrupts that path degrades the reading without announcing that it has.

The device is inside its first few weeks. Before a system has learned your normal it has nothing of yours to compare you against, so it compares you to other people. That is a different question from the one you are asking. Early readings are frequently correct as measurements and misleading as judgements.

The test that settles it is corroboration

There is one question worth asking before you act on any surprising reading.

Did anything else move with it?

One signal moving on its own is a question. Several signals moving together, in a direction that makes physiological sense, against your own history rather than a population band, is an answer.

This is why single-metric alarm is so often wasted. The metric that moved may simply be the one with the widest error bars on that particular night. Corroboration is what separates a body telling you something from a sensor having a bad morning.

One night is weather, six weeks is climate

The other correction is temporal.

A single reading is a point estimate with error attached. A trend is many measurements, and the errors in a trend tend to cancel while the underlying direction persists. This is ordinary statistics rather than anything exotic, and it has a practical consequence: the direction of your data over weeks is far more trustworthy than any individual morning, including the mornings that look alarming and the mornings that look excellent.

If you take one habit from this page, make it this. Read the direction. Do not rearrange your week around a single number.

What to do when a reading is genuinely wrong

Very little, which is the point.

Note it, so that if the same pattern recurs you can see whether it clusters around particular conditions, a particular strap position or particular nights. Check the obvious physical causes before the physiological ones. Then carry on with the plan you had, and let the next few days adjudicate.

The failure mode worth avoiding is the one where a bad reading produces a cascade: you train lighter, sleep worse because you are now worried, and the following morning confirms a decline that your own reaction created.

Why your own history is the only useful reference

Almost every consumer score compares you to a population band before it has enough of your own data to do better. That is a reasonable default and a poor long-run answer, because the range of healthy human physiology is wide and your position inside it is not a verdict.

A reading can only be called unusual against something. If that something is other people, the word unusual is doing very little work. If it is your own trailing behaviour, it starts to mean something.

Where NuVARD fits

NuVARD reads from more than 300 devices and apps and orders what it finds across 53 physiological variables and 15 body systems, against your baseline rather than a population average. The reason for the breadth is exactly the corroboration test above: a system that can see several signals at once can tell you whether a surprising value is supported by anything else, which a single app looking at a single silo structurally cannot.

Every forecast it makes ends in one of three states, Held, Still open or Missed, scored against what actually happened. That includes the misses.

Early access opens in cohorts through 2026. You can join the waitlist at nuvard.ai.

NuVARD provides wellness and health intelligence. It does not diagnose, treat, cure or prevent disease, and nothing on this page is medical advice.