Your wearable gets resting pulse right, while calories and sleep are another story

Key takeaways

• A 2024 review found that only 34 of 310 consumer wearables released since 2003 had been checked against a lab standard for even one measure¹.
• Wrist heart rate is usually within about 3% of the truth, but in a Duke University test, error during activity ran about 30% higher than at rest¹³.
• Calorie estimates for a workout can miss by as much as about 20%, and sleep trackers tend to count time awake in bed as sleep¹²⁵.
• Hospital pulse oximeters missed dangerously low oxygen about three times as often in Black patients, and the FDA’s 2025 testing plan for oximeters, still a draft as of October 2026, doesn’t cover wellness devices⁷⁹.
• You can run basic checks on your own device in about 20 minutes with a manual pulse count, a step count and, for blood pressure, a validated cuff¹³¹⁴.

Since 2003, at least 310 consumer wearables have reached the market. In 2024, researchers at University College Dublin went looking for evidence that they measure what they claim. They found that only 34 had been tested against a lab standard for even one measure¹.

Those devices report your pulse, steps, calories, sleep stages and blood oxygen, often down to the last beat or calorie. A precise-looking number feels like a measured one, but it may be closer to an educated guess.

The research also faces a speed problem. Companies typically release new hardware every year and update software several times a year¹. Studies can take more than 12 months, so a device is often out of date by the time its accuracy results come out².

So “Is my wearable accurate?” isn’t one question. It depends on which number you’re reading, what you were doing, how the device fits, and sometimes your skin. We walk through what the research shows, metric by metric, and how to test your own device at home.

Accuracy depends on which number you’re reading

The best single source is a 2024 “living umbrella review” in the journal Sports Medicine. It pooled 24 systematic reviews covering 249 validation studies and more than 430,000 participants¹. A systematic review gathers every good study on a question and weighs them together. An umbrella review does the same thing for the reviews themselves.

The table below sums up what that review and other key studies found. Ratings use our four-level evidence scale, applied here to one question: is this reading accurate enough to rely on?

Metric	Rating	What studies found
Heart rate at rest	Strong evidence	Across studies, wrist devices average within about 3%, and they do best when you’re still13.
Heart rate during exercise	Promising but early	Error ran about 30% higher during activity and varied by brand3.
AFib (irregular rhythm) detection	Strong evidence	Pooled across 18 studies, sensitivity was 100% and specificity 95%1. Our smartwatch article covers what everyday alerts miss.
Steps	Strong evidence	Errors from about -9% to +12%, usually undercounting1.
Calories burned	Evidence says no, or unsafe	Average error looks small, but individual errors ranged from about -21% to +15%12.
Fitness (VO2 max) estimates	Promising but early	Estimates from resting data varied by about 15% either way1.
Total sleep time, as a rough nightly estimate	Promising but early	Devices tend to overcount sleep, with errors usually above 10%1.
Sleep stages	Evidence says no, or unsafe	Stage estimates were inconsistent across devices5.
Blood oxygen on a watch	Not enough evidence	One low-quality review of one watch model found average differences up to 2 percentage points1.
Cuffless blood pressure numbers	Evidence says no, or unsafe	In 2025 the FDA told consumers not to use unauthorized watch and ring blood pressure features, citing the risk of inaccurate readings12.

Two patterns stand out. Measures that track a steady signal, like your resting pulse, do well. Measures that depend on a company’s estimate of something the sensor can’t see directly, like calories or deep sleep, do poorly.

How a watch reads your pulse, and what throws it off

Most wrist and ring sensors use a method called photoplethysmography, or PPG. Small lights shine into the skin, and a sensor measures how much light bounces back as blood pulses through the vessels underneath³. Software turns that rhythm into a heart rate.

Duke University engineers named three possible sources of error: skin type, motion, and “signal crossover”³. Crossover happens when the sensor picks up a repeating movement, such as your arm swinging as you walk, and mistakes it for a heartbeat.

To test these, the Duke team checked six devices against a medical ECG while 53 people sat, breathed deeply, walked on a treadmill and typed³. An ECG, or electrocardiogram, reads the heart’s electrical signal directly and served as the reference.

What they found³:

• Movement raised the error. Consumer devices were off by an average of about 7 beats per minute at rest and 10 during activity.
• Brands differed. The Apple Watch 4 was most accurate, off by about 4 to 5 beats. The Xiaomi Miband 3 was least accurate among consumer devices, off by about 10 at rest and 14 during activity.
• Even typing raised the error. Small hand movements disrupted several devices almost as much as walking.
• Readings lagged. The optical signal trailed the ECG by up to 43 seconds, so quick changes during interval workouts may show up late.
• Direction varied. Devices tended to read too high while walking and too low during typing and deep breathing.

Fit and blood flow matter, too. A 2025 FDA clearance summary for one optical pulse feature, Fitbit’s Loss of Pulse Detection, lists motion, hand and finger movement, tattoos on the wrist and poor blood flow as things that can hurt performance¹⁵. Cold skin and poor circulation also affect fingertip oxygen readings, according to the FDA⁶.

The study tested older models with a small group, so today’s devices may do better or worse. Its core lesson holds: a wrist number is most trustworthy when you’re still.

Skin tone: unsettled for pulse, documented for oxygen

Many wrist sensors use green light, and earlier studies had reported larger green-light errors on darker skin, especially during exercise⁴.

The Duke study looked for this and found no significant link between skin tone and heart rate error³. That result is reassuring but not final.

In a published response, Peter Colvonen of the VA San Diego pointed out that only nine people in the study had the darkest skin type⁴. He argued that was too few to rule out a real difference. He also noted that the Fitzpatrick scale used to sort skin tone was built in 1975 to rate sunburn risk and tracks skin color only loosely⁴.

For blood oxygen, the problem is better documented, though the key evidence comes from hospital pulse oximeters, not watches.

In December 2020, University of Michigan doctors led by Michael Sjoding reported in the New England Journal of Medicine on more than 10,000 paired oximeter readings and arterial blood tests taken at their hospital earlier that year¹⁰. When the oximeter showed a reassuring number, Black patients’ blood tests revealed hidden low oxygen 11.7% of the time, compared with 3.6% for White patients⁹¹⁰. The Anesthesia Patient Safety Foundation noted that older studies, dating to 1990, had found similar bias⁹.

The FDA responded in stages:

• 2021: A safety communication warned that skin pigmentation, poor circulation, skin thickness, skin temperature, tobacco use and nail polish can affect oximeter accuracy⁶.
• January 2025: Draft guidance proposed that makers test oximeters on 150 or more people, up from 10 under earlier advice⁸. At least a quarter of participants should fall in each of three skin-tone groups on the Monk Skin Tone scale, a 10-shade color chart⁷.
• As of October 2026: The FDA’s guidance page still labeled it a draft, “not for implementation,” so its recommendations are not yet final⁷.

This is the detail most shoppers miss. The draft guidance covers oximeters sold for medical purposes. Products sold for general wellness or sports are outside its scope⁷.

Watch and ring oxygen features sold as wellness tools fall into that lane⁷¹¹. So even when the new testing plan is final, it won’t automatically apply to the number on your wrist.

Sleep, calories and fitness scores drift the most

Sleep trackers are good at one thing: noticing that you’re asleep. In a 2021 lab study of seven consumer sleep trackers, four of them wearables, every device caught sleep at least 93% of the time⁵.

Spotting wakefulness was another story. The devices correctly flagged only 18% to 54% of the time people were awake⁵. Performance also slipped on nights of disrupted sleep, which is when many people care most⁵. Our sleep tracker article covers stage-by-stage accuracy and how to use the data.

Calories have a similar problem, because a device can’t measure the energy you burn directly. It has to estimate it.

On average the error looks small, but for an individual workout it can run from about 21% too low to 15% too high². That gap can erase a planned calorie deficit. Our fitness tracking article goes further.

Fitness scores such as VO2 max, an estimate of how well your body uses oxygen during hard exercise, work better when the device has exercise data. Estimates built from resting data varied by about 15% either way¹.

Your device can change overnight

A wearable is partly hardware and partly software. Companies push software updates several times a year¹. Each update can change the formulas that turn light and motion into your numbers.

That creates two practical problems. First, a published study may describe last year’s software, not yours². Second, a sudden shift in your sleep score or resting heart rate could reflect a new algorithm rather than a change in your body.

The rules leave much of the checking to the companies. Under the FDA’s January 2026 wellness policy, a wellness wearable may show values that look clinical, such as oxygen or heart rate variability, if they are validated¹¹. That validation can come from the company’s own testing or published research, and the FDA says it does not intend to examine low-risk wellness products¹¹. Our guide to FDA device terms explains that policy in more detail.

The Duke team found that newer, pricier and more widely sold devices tended to be more accurate, though they couldn’t tell which factor mattered³. Price is a hint, not a guarantee.

Test your own device in 20 minutes

You don’t need a lab to learn how much to trust your wearable. You need a quiet chair, a clock and, for some checks, a medical device you already own. Write each result down so you can repeat the test after updates.

1. Resting pulse check. Sit for 10 minutes¹⁴. Place your index and middle fingers on the inside of your wrist, just below the thumb, and count beats for 30 seconds, then double it¹⁴. Compare with the watch at the same moment. Do this on three different days. In the Duke study, consumer watches averaged about 7 beats off at rest, so a gap much larger than that is a reason to check the fit³.
2. Movement check. Walk briskly for five minutes, stop, and count your pulse for 15 seconds, then multiply by 4¹⁴. If the watch reads much higher than your count, it may have locked onto your stride, the signal crossover described above³.
3. Step check. Count 500 steps out loud on a walk and compare. Expect the device to come in a little low¹.
4. Blood pressure check. If your device estimates blood pressure, compare it with a home cuff from the validated list at validatebp.org¹³. Prepare the way the National Heart, Lung, and Blood Institute advises: no exercise, coffee or smoking for 30 minutes, an empty bladder, five quiet minutes seated with feet flat, arm supported at heart level, and no talking¹³. Use the cuff, not the watch, for any decision¹².
5. Oxygen check. If you own a fingertip oximeter sold for medical use, compare readings at rest with warm hands and no nail polish⁶. Remember that pulse oximeters can also read high on darker skin⁹.
6. Sleep check. For one week, jot down when you turned out the light and when you got up. If the tracker reports much more sleep than you remember, it may be counting quiet wake time as sleep⁵.
7. Repeat after changes. Run steps 1 to 3 again after a major software update or a new band¹.

Trust your body first, the device second

Many people who value natural health already pay close attention to how they feel. That habit is the best partner a wearable can have.

The FDA’s advice for home oximeter users applies to every wearable number. Don’t rely on the device alone. Track your symptoms, and contact a clinician if a reading worries you or your symptoms get worse⁶. Shortness of breath with a low oxygen reading, or an irregular pulse you can feel, deserves a call⁶¹⁴.

Counting your own pulse takes a minute, costs nothing and never needs a software update. A normal resting rate for adults runs from 60 to 100 beats per minute, and well-trained athletes may sit at 40 to 60¹⁴. Learning your own pulse gives you a baseline no company can change overnight.

The same skepticism you bring to a supplement label belongs here. A wearable company has products to sell. Ask whether a feature has been tested independently, on people like you, with the software version you’re using. When the answer is no, treat the number as a rough trend.

Questions readers ask

The bottom line

Your wearable is a good pulse counter when you’re still and a decent step counter. It’s a rough guesser for calories, deep sleep and, on most devices, blood oxygen. Motion, fit, software updates and possibly skin tone all shape the number you see, and few devices have been tested independently.

You can learn how far to trust your own device with a manual pulse count, a step count and a validated cuff. Use the wearable for trends and motivation. Use your body’s signals and FDA-cleared medical devices for decisions.

Related reading

• “FDA registered” isn’t “FDA approved”: a buyer’s map to wellness devices
• Smartwatch health alerts are better at warning you than reassuring you
• What a medical test can and cannot tell you
• Why being out of range often means nothing at all

Sources

1. Doherty C, Baldwin M, Keogh A, Caulfield B, Argent R. “Keeping Pace with Wearables: A Living Umbrella Review of Systematic Reviews Evaluating the Accuracy of Consumer Wearable Technologies in Health Measurement.” Sports Medicine, 2024. link.springer.com
2. Doherty C. “How Accurate Are Wearable Fitness Trackers? Less Than You Might Think.” The Conversation, republished by Sleep Review, August 2024. sleepreviewmag.com
3. Bent B, Goldstein BA, Kibbe WA, Dunn JP. “Investigating Sources of Inaccuracy in Wearable Optical Heart Rate Sensors.” npj Digital Medicine, 2020. nature.com
4. Colvonen PJ. “Response To: Investigating Sources of Inaccuracy in Wearable Optical Heart Rate Sensors.” npj Digital Medicine, 2021. nature.com
5. Chinoy ED, Cuellar JA, Huwa KE, et al. “Performance of Seven Consumer Sleep-Tracking Devices Compared With Polysomnography.” Sleep, 2021. pubmed.ncbi.nlm.nih.gov
6. U.S. Food and Drug Administration. “Pulse Oximeters.” Content current as of January 6, 2025. fda.gov
7. U.S. Food and Drug Administration. “Pulse Oximeters for Medical Purposes: Non-Clinical and Clinical Performance Testing, Labeling, and Premarket Submission Recommendations.” Draft guidance, January 2025 (status checked October 2026). fda.gov
8. TCTMD. “FDA Issues Draft Guidance for Improving Skin Tone Disparities of Pulse Oximeters.” January 8, 2025. tctmd.com
9. Feldman J, Lane-Fall M. “APSF Statement on Pulse Oximetry and Skin Tone.” Anesthesia Patient Safety Foundation Newsletter, June 2021. apsf.org
10. HealthDay. “Pulse Oximeter Readings May Be Inaccurate in Black Patients.” January 6, 2021. healthday.com
11. U.S. Food and Drug Administration. “General Wellness: Policy for Low Risk Devices.” Guidance, January 2026. fda.gov
12. U.S. Food and Drug Administration. “Do Not Use Unauthorized Devices for Measuring Blood Pressure: FDA Safety Communication.” September 16, 2025. fda.gov
13. National Heart, Lung, and Blood Institute. “High Blood Pressure: Diagnosis.” 2025. nhlbi.nih.gov
14. MedlinePlus, National Library of Medicine. “Pulse.” 2025. medlineplus.gov
15. Fitbit LLC. 510(k) Summary K242967, Loss of Pulse Detection. U.S. Food and Drug Administration database, February 2025. accessdata.fda.gov