How Accurate Are Fitness Trackers?

A fitness tracker shows you a dozen numbers with the same confident typography, and they do not deserve the same confidence. Some are close to measurements. Some are estimates built on assumptions about you. Knowing which is which is the difference between a device that improves your training and one that quietly annoys you.

Here is the honest ranking, from most trustworthy to least.

1. Steps: reliable enough to act on

Step counting is a mature technology. An accelerometer detects the rhythmic acceleration of walking, and modern algorithms are good at it. Expect small errors rather than large ones, with the usual quirks: pushing a shopping trolley or a pushchair under-counts because your wrist is still, and hand-heavy activities like clapping or chopping can add phantom steps.

More importantly, the errors are consistent. If your tracker says 8,000 steps today and 11,000 tomorrow, you genuinely moved more — and that comparison, not the absolute figure, is what a step count is for.

2. Heart rate: excellent at rest, variable under load

Resting and steady-state heart rate from a wrist sensor is good. Published work on optical wrist sensing describes strong agreement with reference measurement during steady activity and wider disagreement as intensity varies — which matches what people experience: fine on a walk, unreliable in intervals.

Four things break wrist readings: a loose band, tendon movement under the sensor during gripping (rowing, pull-ups, kettlebell work), cold weather reducing peripheral blood flow, and rapid changes in effort that smoothing algorithms lag behind. The occasional worst case is cadence lock, where the sensor tracks your arm swing rather than your pulse and produces a stable, convincing, wrong number.

The fix is not a better watch. It is a chest strap, which measures the heart’s electrical signal instead of inferring it from light.

3. Sleep duration: good. Sleep stages: directional

Wearables are reasonably good at knowing whether you are asleep, and much weaker at knowing which stage of sleep you are in. A clinical sleep study measures brain activity; a wearable infers stages from movement and heart rate. Cleveland Clinic makes exactly this point about wearable sleep tracking, and a review of consumer sleep apps in a clinical journal reaches the same conclusion.

Use the numbers that hold up: total sleep time, bedtime consistency, and how your resting heart rate and heart rate variability move across weeks. Ignore the nightly deep-sleep percentage. There is also a documented irony here — anxiously optimising a sleep score can itself cost you sleep. Our sleep tracker roundup is written on that basis.

4. Calories: the weakest number on the screen

Energy expenditure is not measured at all. It is modelled from heart rate and movement combined with the age, weight, height and sex you typed in during setup — and individual metabolism varies more than any of those inputs capture. Two people of the same size doing the same session genuinely burn different amounts, and no wrist device knows which one you are.

Treat calorie figures as a relative scale for your own sessions: this workout was harder than that one. Do not treat them as a budget to eat back against. If weight management is the goal, our weight loss equipment guide explains why the equipment matters less than consistency.

5. Blood oxygen, temperature and stress scores: context only

SpO2 readings from a wrist are sensitive to fit, movement and skin contact, and consumer devices generally publish them as wellness features rather than medical measurements. Skin temperature is reported as a deviation from your own baseline rather than a thermometer reading, which is a sensible design and also a reminder that the absolute number means little. “Stress” and “readiness” scores are proprietary composites, mostly of heart rate variability — useful as a trend line for your own body, meaningless as a comparison to anyone else’s.

The exception: regulated features

A few wearable health features have been through actual regulatory review. The FDA granted a De Novo classification for the first over-the-counter irregular rhythm notification feature, and cleared wrist ECG apps followed. These are meaningfully different from a stress score — they have been evaluated against a standard — and they have prompted real diagnoses.

They are still screening tools. A notification is a reason to see a doctor, not a diagnosis, and the absence of a notification is not a clean bill of health. Blood pressure remains the metric people most often assume is on their wrist and mostly is not — we cover that separately in do fitness trackers measure blood pressure?

How to get better data from the device you own

• Wear it properly.A finger’s width above the wrist bone, snug enough not to slide. This fixes more complaints than any firmware update.
• Enter your details honestly. Height, weight and age feed every estimate the device makes.
• Pair a strap for hard sessions. Intervals, lifting and cold-weather training are exactly where a wrist sensor is weakest.
• Compare yourself to yourself. Every estimated metric here is far more useful as a trend across weeks than as a number on one day.

The bottom line

A fitness tracker is a consistency tool, not a laboratory. Its real value is that it notices patterns you would not: that your resting heart rate has crept up during a stressful month, that your step count collapses at weekends, that you sleep an hour less when you train late. Those observations change behaviour, and behaviour is what changes fitness. Ask it for trends and it is excellent; ask it for precision and it will make something up with a straight face.

None of this is medical advice, and no wearable on this site is a medical device unless a specific feature has been cleared as one. If you have symptoms, or a reading that worries you, that is a conversation with a clinician rather than an app.