How a Wearable Actually Measures You

A wrist-worn heart-rate sensor typically shines green light into your skin dozens of times per second and measures how much bounces back — blood absorbs more light than surrounding tissue, so the reflected amount rises and falls with every heartbeat. One measurement is not one heartbeat. It is dozens of individual light readings per second, stitched together afterward into something that looks like a single number.

The specific technique has a name: photoplethysmography, or PPG. An LED fires green light — green because tissue absorbs and scatters it strongly, which makes the small changes caused by blood flow easier to pick out — and a photodiode sitting right next to it measures how much light comes back. During systole, when the heart pushes a fresh pulse of blood into your wrist, more light gets absorbed and less reflects back. During the trough between beats, more light reflects. That rise-and-fall pattern, read many times a second, is the entire raw material a smartwatch has to work with — there is no direct electrical reading of your heart happening anywhere on your wrist.

Toggle to Raw signal below and look at how jagged it is sample to sample. Every small wrist movement, every shift in how tightly the band sits against your skin, and the ambient light in the room all show up as noise layered directly on top of the real pulse — none of it removable by simply asking the sensor to try harder.

Toggle to Smoothed signal and the same data resolves into a clean, repeating pulse shape. That smoothing — averaging each point with its close neighbours — is not optional polish. Without it, the number you would see on screen would jump around too violently to mean anything, swinging on tiny movements rather than tracking your actual heart rate.

Every consumer wearable does some version of this processing before a single number ever reaches its display. The reading you trust every day is already several steps removed from the raw physical signal — a fact to remember in the next chapter, which is specifically about how much you should trust it.

Motion is the obvious problem — running or lifting weights shakes the sensor against the skin and produces movement artefacts far larger than the pulse signal itself, which is exactly why heart-rate readings during a workout are the least reliable ones a wearable produces. It is not the only one.

Darker, more melanin-rich skin absorbs more green light than lighter skin does, which means less light reaches the blood vessels and less of it returns to the photodiode — a weaker signal sitting under the same amount of motion and ambient noise. A tattoo directly over the sensor compounds the same problem: ink absorbs light too, regardless of skin tone underneath it, and can degrade the reading further still. Neither effect is a hypothetical edge case. The same light-absorption problem in pulse oximeters, which use red and infrared light rather than green, led to a 2021 FDA safety communication warning that their readings can be less accurate on patients with darker skin pigmentation.

A hospital sleep study — polysomnography — measures sleep stages directly, from electrodes reading brain waves, eye movement, and muscle activity. A wearable has none of those sensors. It infers a sleep stage from proxies: how still your wrist is, how your heart rate and its beat-to-beat variability shift, and on some devices your breathing rate estimated from the same optical sensor.

Those proxies correlate with sleep stage without being the same thing as it. Published validation studies comparing consumer wearables against clinical polysomnography typically find that total sleep time and overall time awake line up reasonably well, while agreement on which specific stage — light, deep, or REM — a given stretch of sleep belongs to is considerably weaker. A wearable telling you that you got ninety minutes of deep sleep is reporting the output of a model's best guess from movement and heart-rate patterns, not a measurement of your brain activity.

Not every wearable measurement is held to the same standard before it ships, and the gap is bigger than most people assume. A step count and a pulse oximeter reading both sound like straightforward numbers. The testing behind each one is not remotely comparable.

None of that makes the step count useless — it was never built for the job a cleared pulse oximeter is built for. The mistake is treating every number that comes off a wearable as though it cleared the same bar, when the testing behind them can differ by an entire category.