“How accurate is it?” is the first question anyone asks about a fitness tracker, and it is the wrong question — because a smartwatch measures several different things with wildly different levels of reliability, and lumping them together produces a useless answer.
The short version: heart rate is good, calories are not, and everything else falls somewhere in between. Understanding why tells you which numbers to build decisions on and which to treat as decoration.
What the Research Actually Found
The most-cited work here comes from Stanford, which tested seven wrist-worn devices — Apple Watch, Basis Peak, Fitbit Surge, Microsoft Band, Mio Alpha 2, PulseOn and Samsung Gear S2 — against continuous 12-lead ECG monitoring and clinical-grade indirect calorimetry, across 60 volunteers of varying age, skin tone and body mass.
Two findings, and they point in opposite directions.
Heart rate: six of the seven devices measured within 5% error. For a consumer device strapped to a wrist, against a hospital 12-lead reference, that is a genuinely strong result.
Energy expenditure: none of the seven were accurate. The best device averaged 27% error. The worst averaged 93%. Not 93% accurate — 93% off.
That gap is the single most important thing to understand about wearables, and it has held up as devices have improved. More recent work, including 2026 research on how central adiposity and body composition affect readings, continues to find the same pattern: the physiological signal is measured reasonably well, and the derived estimate built on top of it is not.
Why Heart Rate Works
Your watch uses photoplethysmography, or PPG. Green LEDs shine light into the skin; blood absorbs green light; the volume of blood in your capillaries rises and falls with each heartbeat; a photodiode measures the changing reflection and derives a pulse from the rhythm.
It works well because it is measuring something physical and periodic with a strong signal. Rest your arm on a table and a modern watch will track your heart rate about as well as a clinical monitor.
Note what it is not doing: it is not measuring the heart’s electrical activity. That is what an ECG does, and why a chest strap — which reads electrical signal — behaves differently. It also explains why the ECG feature on some watches is a separate sensor with separate electrodes, not an upgrade to the optical one.
Why Calories Do Not
Nothing in your watch measures energy expenditure. It cannot — measuring that properly requires analyzing the gases you exhale.
What the watch does is estimate. It takes heart rate and motion data, combines it with the profile you entered (age, weight, height, sex), and applies a population-level model to guess how many calories that combination probably burned.
Every step introduces error:
- The model is population-based. Your actual metabolic rate can differ substantially from the average for your demographic.
- Your profile data is stale. Most people enter a weight once and never update it.
- Body composition is invisible to the watch. Muscle and fat have very different metabolic rates, and the watch has no way of knowing your ratio. Research in 2026 specifically identified central adiposity as a factor degrading both heart rate and calorie accuracy.
- Exercise type is inferred, not known. The same heart rate from cycling and from anxiety produce different real energy costs.
The practical consequence: use calorie numbers as a relative index, never as an absolute. “I burned 15% more this week than last week” is a defensible reading of the data. “I burned 620 calories, so I can eat 620 calories” is not, and it is how people end up frustrated that tracking is not working.
Where Heart Rate Accuracy Breaks Down
Even the reliable metric has failure modes, and they cluster around predictable situations.
High-intensity intervals. The optical signal lags. Sprint from a walk and there is a delay of several seconds before the reading catches up — sometimes much longer. For interval training, wrist readings routinely miss the peaks entirely.
Weight training and anything gripping. Wrist flexion and forearm muscle contraction physically deform the tissue under the sensor and press the watch against the skin unevenly. Readings during heavy lifting are frequently nonsense.
Cold weather. Peripheral vasoconstriction reduces blood flow to the extremities, which reduces the signal the sensor has to work with. Cold hands produce dropouts.
Cadence lock. A specific and sneaky failure: during running or cycling, the sensor sometimes locks onto the rhythm of your arm swing or pedal stroke rather than your pulse, producing a reading that is stable, plausible, and completely wrong. If your heart rate looks suspiciously flat and matches your cadence, this is why.
Loose fit. The largest single controllable factor. A watch that slides during movement lets ambient light into the sensor gap.
Tattoos. Dense ink directly under the sensor absorbs green light before it reaches blood vessels, and can prevent readings outright.
Skin tone. Melanin absorbs green light, reducing signal on darker skin. Manufacturers compensate with brighter LEDs and multi-wavelength sensors, and current devices are much better than early ones, but published research continues to find measurable differences. It affects passive background monitoring more than deliberate readings.
Seven Things That Actually Improve Your Readings
- Tighten the band. Snug enough that it does not slide, loose enough to fit a fingertip under. This alone fixes most complaints.
- Move it up your arm. About one finger-width above the wrist bone — not down against the hand where wrist flexion moves it.
- Tighten further for workouts. Many people keep a second, tighter hole for exercise and loosen it afterward.
- Clean the sensor. Sweat residue and sunscreen build a film over the optical window.
- Update your profile. Weight in particular, since it feeds directly into every derived estimate.
- Wear it on the non-dominant wrist. Less movement, fewer artifacts.
- Use a chest strap for interval and strength training. Most watches — Apple, Garmin, Samsung, Polar, Coros — will pair with one over Bluetooth and use its data instead of the wrist sensor.
What About Sleep, SpO2 and Steps?
Sleep staging is the most over-trusted metric on any wearable. Determining sleep stages properly requires measuring brain activity, and your watch measures movement and heart rate variability instead. Total sleep duration is reasonably good; the light/deep/REM breakdown is an inference and should be read as a trend rather than a measurement. Wearables are better at detecting changes in your sleep than at describing any single night. Devices designed specifically for this do better — see best smartwatches for sleep tracking.
SpO2 (blood oxygen) is a wellness feature on nearly every consumer wearable, not a medical one, and it is explicitly labeled as such. It is prone to error from fit and motion, and readings are typically only taken while still. Useful for spotting a trend, not for assessing a respiratory problem. Our oxygen sensor roundup covers which devices have it.
Step counting is generally very good on a wrist for walking and running, and generally poor for pushing a stroller, a shopping cart, or anything else that immobilizes your arms. It also picks up phantom steps from hand gestures. Clip-on and ankle trackers do better in those specific cases.
HRV deserves its own explanation, because it is the metric most likely to be misread — see what HRV means on a smartwatch.
The Honest Summary
Ranked from most to least trustworthy:
- Resting heart rate — very good, and the single most useful health number your watch produces.
- Steady-state exercise heart rate — very good with a proper fit.
- Step count (walking/running) — good.
- Total sleep duration — reasonably good.
- Heart rate variability trend — useful as a trend, meaningless as a one-day number.
- Interval and strength-training heart rate — unreliable; use a chest strap.
- Sleep stage breakdown — directional only.
- SpO2 — directional only.
- Calorie burn — the least reliable number on the device by a wide margin.
None of this means the devices are not worth wearing. Consistency is what makes a wearable valuable: even an imperfect measurement, taken the same way every day, reveals changes you would not otherwise notice. The mistake is treating an estimate as a measurement — and calories are the estimate everyone treats as a measurement.
If you are shopping and want to know which devices do the sensing best, start with our best smartwatches of 2026 and best fitness trackers 2026 guides.
This article summarizes published research for general information. It is not medical advice.