Wearable data quality and how to interpret it

In 2017 a Stanford team put seven wrist devices on 60 volunteers and compared them with laboratory references. Six of the seven kept median heart rate error under 5 percent during cycling. None of them estimated energy expenditure with an error below 20 percent (Shcherbina and colleagues, 2017, Journal of Personalized Medicine). The same sensors on the same wrists produced one number you could train by and one you could not. Most training mistakes made with wearable data start in the gap between what a device measures well and what it only infers.

This guide sorts wearable numbers by how far you can trust them. It separates measurements from estimates, gives the published accuracy for each common metric, summarizes the devices serious athletes wear as of September 2026, and covers the habits, noise limits and failure modes that decide whether a reading means anything. Subjects with their own article get a short summary and a link. For the physiology behind the signals, how heart rate, HRV and sleep respond to training, start with the performance physiology guide.

Consider a morning that plays out constantly. You check your wrist or ring and see a recovery score of 42 out of 100 on a day the plan calls for a threshold session. Whether to run it depends on questions most athletes never ask. Which inputs fed that 42? What is the measurement error on each one? Is 42 different from yesterday's 48, or is the gap inside the device's noise? Did you sleep on the wrist that wears the sensor? The wearable metrics glossary entry defines the terms used below.

01Measurements and estimates

The most useful distinction in wearable data is between a measurement and an estimate. A measurement is a value a physical sensor captures directly. An estimate is a value an algorithm computes from one or more measurements. The two carry different levels of certainty, and treating them alike leads to bad decisions.

A wearable holds a small number of sensors. The common ones are an optical heart rate sensor (photoplethysmography, or PPG), a three-axis accelerometer, a gyroscope, a skin temperature sensor and, in many devices, red and infrared LEDs for blood oxygen. Some add a barometric altimeter or electrodes for a single-lead ECG. The PPG sensor shines green light into the skin and measures how much comes back. Blood absorbs green light, and capillary blood volume pulses with each heartbeat, so the reflected signal oscillates at your heart rate. Motion, fit and ambient conditions add noise even at this level.

Everything else is derived. No sensor on your wrist detects deep sleep or REM. The device combines movement, heart rate, heart rate variability and sometimes breathing and temperature, then feeds them to a classifier trained against polysomnography, the clinical sleep study that records brain waves with EEG. VO2 max comes from the relationship between heart rate and pace or power. Calorie burn combines heart rate, movement and your profile with a metabolic model. Recovery, readiness and stress scores weight several estimated inputs into one number.

Each layer adds uncertainty, and stacked estimates inherit every layer's error. If the PPG signal was poor overnight because of wrist position, the HRV value suffers, the sleep staging suffers, and a composite score built on both inherits both errors at once.

A practical ranking follows. Direct measurements from a well-seated sensor earn moderate to high confidence. These include resting heart rate taken during stillness, a full day's step count, and skin temperature change across nights. Single-layer estimates built on good measurements earn moderate confidence. These include HRV computed from beat intervals, respiratory rate inferred from the PPG signal, and exercise heart rate during steady work. Multi-layer composites earn the least. Sleep stage breakdowns, recovery scores, VO2 max, calorie burn and stress scores all sit here. A rise of 8 beats per minute in resting heart rate across three mornings deserves attention. A 15-point drop in a composite recovery score on one morning deserves investigation before action.

Precision and accuracy are separate properties. Precision means the device gives consistent readings under the same conditions. Accuracy means those readings sit close to the true value. A device that always reads resting heart rate 3 bpm above an ECG is precise and biased. For tracking your own trend, precision matters more, because a constant offset cancels out when you compare today with your own history. Consistent wearing conditions protect precision, and that is the part you control.

02Accuracy metric by metric

Each wearable metric has its own error profile. This section gives the published numbers where good studies exist, says so where they do not, and links to the article that covers each metric in full.

Resting and exercise heart rate

At rest with good contact, optical sensors read heart rate well. A living meta-analysis of Apple Watch accuracy by Lambe and colleagues (2026, npj Digital Medicine) pooled 82 studies and found a mean heart rate bias of -0.27 bpm, with 95 percent limits of agreement from about -7 to +7 bpm. Any single reading can still be several beats off. Dial and colleagues (2025, Physiological Reports) found the Oura Ring Gen 3 and Gen 4 within about 2 percent of ECG for resting heart rate across 536 nights, and WHOOP 4.0 within 3 percent.

Movement is the main source of error. Bent and colleagues (2020, npj Digital Medicine) found that absolute error during activity averaged 30 percent higher than at rest. Gillinov and colleagues (2017, Medicine and Science in Sports and Exercise) compared wrist devices with ECG on the treadmill, bike and elliptical. A Polar chest strap agreed almost perfectly, with a concordance of 0.996. The Apple Watch was the best wrist device at 0.92, the others ranged from 0.67 to 0.83, and on the elliptical with arm levers every wrist device fell below 0.80. A wrist sensor can also lag the true rise at the start of a hard interval, or lose the pulse during burpees and kettlebell work and fill the gap with a smooth guess.

Resting heart rate from a wrist or ring is one of the most dependable daily numbers you have. Exercise heart rate from the wrist is good enough for steady aerobic work. For threshold intervals and zone work, a chest strap remains the better tool.

Heart rate variability

Heart rate variability describes the variation in time between beats. Most consumer devices report RMSSD, which reflects vagal activity. Apple Health stores SDNN instead, as Apple's HealthKit documentation states, and the two formulas give different numbers from the same heartbeats. HRV needs each beat timed to within a few milliseconds, which PPG does best overnight while you lie still.

Dial and colleagues (2025) compared five devices with ECG. For overnight HRV, the Oura Ring Gen 4 agreed most closely, with a concordance of 0.99 and a mean absolute percentage error of 5.96 percent. The Oura Gen 3 followed at 0.97 and 7.15 percent, WHOOP 4.0 at 0.94 and 8.17 percent, the Garmin Fenix 6 at 0.87 and 10.52 percent, and the Polar Grit X Pro at 0.82 and 16.32 percent. Stone and colleagues (2021, Frontiers in Sports and Active Living) found Oura's RMSSD within 6.84 percent of ECG. In Miller and colleagues' one-night test of six devices in 53 adults (2022, Sensors), every device read RMSSD lower than ECG, from 4.5 ms lower for WHOOP to 22.4 ms lower for Garmin. The authors' lab receives funding and equipment from WHOOP.

Even good devices carry errors of 6 to 16 percent on a given night. For an athlete whose true RMSSD is 55 ms, 8 percent is about 4 ms, so a day-to-day change of that size can come from the device. That is why the useful unit is the trend across days, and why Titan measures today's change against your own typical spread instead of a fixed number. The Apple Watch HRV explainer covers how the watch samples HRV, and the HRV and recovery readiness guide covers HRV trends.

Blood oxygen

Pulse oximetry estimates the share of hemoglobin carrying oxygen from how tissue absorbs red and infrared light. For Apple Watch, the Lambe meta-analysis found a mean blood oxygen bias of -0.04 percentage points, with 95 percent limits of agreement of about ±4 points. Pipek and colleagues (2021, Scientific Reports) found the Series 6 read 0.8 points high on average in 100 patients with lung disease, with limits from -2.7 to 4.1. A single reading of 95 could reflect a true value anywhere from about 91 to 99.

Skin pigmentation matters more here than for heart rate. Sjoding and colleagues (2020, New England Journal of Medicine) found that hospital pulse oximeters missed arterial oxygen below 88 percent, while reading 92 to 96, in 11.7 percent of measurements in Black patients and 3.6 percent in White patients. Wrist and ring devices rely on the same optics. For athletes, blood oxygen is most useful at altitude, where real changes exceed the error. A reading that stays low, or any breathlessness, is a reason to see a clinician with a medical device.

Sleep staging

Wearables infer sleep stages from movement, heart rate, HRV and sometimes temperature, which makes sleep tracking one of the most marketed and least exact wearable features.

Miller and colleagues (2022) compared six devices with polysomnography. On the simple question of asleep or awake, Apple Watch, Garmin, Polar, Oura and WHOOP agreed 86 to 89 percent of the time. On sleep stages, agreement fell to 50 percent for Garmin, 51 for Polar, 53 for Apple Watch, 60 for WHOOP and 61 for Oura. Chinoy and colleagues (2021, Sleep) tested seven other devices. Sensitivity for detecting sleep was 0.93 or higher on every device, while specificity, the ability to catch wake, ran from 0.18 to 0.54. Devices miss short awakenings and score them as light sleep, so they tend to overstate total sleep.

Duration and timing are dependable enough to act on, and stage breakdowns are rough context. If your device reports 30 more minutes of deep sleep than last night, doubt it. If it reports 90 minutes less total sleep, believe it. Titan labels Apple Watch stages Core, Deep and REM. Sleep from the direct Oura connection arrives without stages, so those nights show no stage breakdown. The Apple Watch sleep accuracy article covers the watch's staging study by study.

Calories burned

Calorie burn is the weakest common wearable metric. In Shcherbina's study no device got energy expenditure error below 20 percent. Fuller and colleagues (2020, JMIR mHealth and uHealth) reviewed 158 publications across nine brands and concluded that no brand measured energy expenditure accurately. O'Driscoll and colleagues (2020, British Journal of Sports Medicine) pooled 60 studies and found that adding heart rate to movement helped.

Heart rate and wrist motion cannot tell a 70 kg runner on flat ground from the same runner carrying a 15 kg pack uphill, and much of the cost of resistance training is anaerobic or arrives after the session. Use calorie numbers to compare your own days, and set food intake from the weight trend over weeks.

Step count

Fuller's review found that Fitbit, Apple and Samsung devices counted steps accurately in the lab. Accuracy drops when the arm does not swing, as when pushing a stroller or holding a rail, and repetitive arm motion can register as steps. A gap of 300 steps between two days means nothing, and 9,000 against 5,000 is real. When a phone and a watch both count steps, Titan's Allowed Sources setting keeps it to one of them.

Respiratory rate

Devices estimate respiratory rate from the way breathing modulates the PPG signal, which works as a trend during sleep and fails during exercise. Published validation for consumer respiratory rate is thin, so trust the trend over the absolute value. A rise in overnight respiratory rate over several nights can come with illness or heavy fatigue, which makes it a useful supporting signal that rarely justifies a decision on its own. Titan plots it on the Vitals screen against your own range.

Skin temperature

The useful skin temperature number is the change from your own baseline across nights, and most devices report it as a deviation. Smarr and colleagues (2020, Scientific Reports) used Oura data from the TemPredict study to show that temperature rises were visible in the first 50 COVID-19 cases and matched self-reported fever. The luteal phase of the menstrual cycle also raises it, and so do room temperature, bedding and fit, so one warm night means little. Several warm nights in a row, especially with resting heart rate up, mean more.

VO2 max estimates

Wearable VO2 max is inferred from submaximal signals, using the relationship between heart rate and pace, power or walking speed. Molina-Garcia and colleagues (2022, Sports Medicine) pooled 14 studies for the INTERLIVE network. Algorithms that used exercise data averaged almost no bias, -0.09 ml/kg/min, with individual limits of agreement near ±10 ml/kg/min. Apple Watch reads low. Lambe and colleagues (2025, PLOS ONE) found it underestimated lab VO2 max by 6.07 ml/kg/min on average in 30 adults, and a 2026 follow-up on the Series 10 in Mayo Clinic Proceedings: Digital Health found a similar gap of 6.25. The trend within one device is usable, and the absolute number can be off by a full fitness category. Norms and testing protocols live in VO2 max and aerobic capacity.

Summary table

03Device profiles as of September 2026

Five platforms dominate among athletes who take their data seriously. Prices and models are US figures as of September 2026. No independent validation of the Series 12, Ultra 4, WHOOP 5.0, WHOOP MG or Oura Ring 5 had been published at the time of writing, so the evidence column names the older model each study tested.

The evidence column mixes studies and hardware generations, so compare each device with its own history. Titan reads Apple Watch data natively, and the watch writes every data type Titan scores with, so Recovery, Stress, Battery, Sleep, Exertion and Titan Age all work from it alone. Oura has two routes into Titan. Neither route writes resting heart rate. Stress compares your heart rate with your usual resting heart rate, so it still needs resting heart rate from an Apple Watch or WHOOP. For Garmin and Polar, whatever Garmin Connect or Polar Flow writes to Apple Health is what Titan can read, so check the Health app for which types appear before relying on either for a Titan score.

The WHOOP and Apple Watch comparison covers sensors, three-year costs, lab results and wearing both. The Oura and Apple Watch comparison covers ring versus wrist HRV and the two Oura routes into Titan.

04Data hygiene

The best sensor in the world returns poor data when it sits loose, sits in the wrong place, or runs under changing conditions. The habits below are dull, and skipping them is one of the most common sources of bad data in everyday use.

Placement and fit. A wrist sensor should sit flat against the skin with steady pressure, about a finger's width above the wrist bone. Closer to the hand, it sits over tendon and bone with less capillary blood under it, and the PPG signal weakens. The band should be snug enough that the watch does not slide and loose enough that it leaves no deep mark. For hard sessions, one notch tighter helps. A ring needs a correct size, and the index or middle finger usually fits best. Fingers shrink in the cold and swell in the heat, so a ring that fits in summer can lose contact in winter. A chest strap needs damp electrodes and should sit just under the chest muscles. Dry electrodes produce dropouts and spikes in the first minutes of a session.

Firmware and algorithm updates. An update can change how a device computes HRV, stages sleep or estimates calories. Note major updates in your training log, so an abrupt shift on the day of an update does not read as a change in you.

Consistent timing. A resting heart rate taken lying down on waking differs from one taken at a desk mid-morning, and HRV from a whole night differs from HRV from a one-minute standing test. Mixing methods wrecks the trend, so pick one protocol and keep it. Titan's Recovery score keeps this rule for you. It reads HRV and heart rate from your main sleep, or from 10 PM to 10 AM on a night with no recorded sleep, so the same kind of reading feeds the score each night.

Band wear. Silicone and fabric bands stretch. A band that held the sensor firmly in month one may let it shift by month six. Replace it when it loosens.

Skin, hair and tattoos. Tattoo ink under the sensor absorbs light and can block the PPG signal, and Apple's support documentation says so directly for Apple Watch. The evidence on skin tone is more mixed, and the FAQ below covers it. If readings are erratic and you have a wrist tattoo, move the device to untattooed skin or use a chest strap for training.

Why poor hygiene is worse than random noise. Random noise averages out over a week. Poor fit tends to push readings in one direction night after night. A loose ring misses beats and distorts HRV. A loose watch reads band movement as wakefulness. These biases build a trend that looks physiological, which makes them the most misleading kind of error a wearable can produce.

05Building a trustworthy baseline

Every useful reading compares you with yourself. A resting heart rate of 58 bpm means one thing for an athlete whose normal is 52 and another for one whose normal is 62. Titan builds the Recovery baseline as the median of your measured nights across the Recovery window, 60 nights by default. Titan will not compute a Recovery baseline until the window holds at least 7 measured nights covering at least a quarter of it, so the 60-night window needs 15. A new device, a significant illness, travel across several time zones or a new heart rate medication each call for a fresh start. The baseline glossary entry covers medians and spread, window length, device switches and when to restart, and the personal baselines article shows the window inside Titan.

06Signal and noise

Every measurement system has a noise floor, the smallest change it can reliably detect. A change below it may be real or may be random, and you cannot tell which. Acting on changes below that limit is acting on chance.

Sport science formalizes this as the smallest worthwhile change, the smallest shift in a metric that reflects a real change beyond measurement noise and day-to-day biology. Hopkins (2000, Sports Medicine) set out the statistics of typical error and reliability that most of this work rests on, and recommended expressing typical error as a coefficient of variation. Buchheit (2014, Frontiers in Physiology) applied the same logic to heart rate measures in athletes and argued that a change in HRV means something only when judged against the error of measurement and the smallest important change.

Two sources of variation make up the noise floor. One is biology, the ordinary day-to-day swing in your physiology. The other is the device's own imprecision. You cannot separate them from a wrist, and you do not need to. What you need is your total typical spread, measured over enough days under consistent conditions. Titan's Recovery model uses median absolute deviation for that spread because it resists outliers. One night of bad contact or a late dinner barely moves it, while a standard deviation gets pulled wide by a single strange reading and makes real changes look small.

For decisions outside the app, this blog uses one rule. A morning HRV more than one median absolute deviation below your baseline median for two or more consecutive days, with resting heart rate above baseline, is a signal to reduce training intensity. Each of its three conditions removes a class of false alarm. One bad night fails the repetition test. A drop within your normal swing fails the size test. An HRV artifact with a flat resting heart rate fails the agreement test.

Suppose your last 60 mornings give an HRV median of 55 ms and a median absolute deviation of 6 ms, so the line sits at 49 ms, and your resting heart rate median is 54 bpm. A reading of 51 ms is noise. A single reading of 46 ms with resting heart rate at 56 is worth noting and not worth acting on. Two consecutive mornings at 47 and 45 ms with resting heart rate at 56 and 57 meet the rule, and the next hard session should become an easy one.

Inside Titan, the Recovery score does this weighing for you. It compares last night's HRV and sleeping heart rate with their baseline medians, in units of your median absolute deviation, and places the result on a fixed curve. A night with both values at baseline scores 56, and sleeping heart rate counts about half as much as HRV. Low, 0 to 39, is a day to favor easy work, sleep and nutrition. Moderate, 40 to 74, means train and keep an eye on intensity. High, 75 to 100, means ready for quality training. The Recovery score article gives the weights.

07Using more than one device

Many athletes wear more than one device, and the numbers rarely match. An Apple Watch can report HRV of 45 ms on the same night an Oura ring reports 62 ms, and neither is broken. The sensor site, the part of the night measured, the artifact filtering and the formula all differ, and Apple Health stores SDNN while WHOOP, Oura, Garmin and Polar report RMSSD. Name one device as the source for each metric, use a second device only to confirm an odd reading, and never average across devices. The Oura and Apple Watch comparison explains each cause of disagreement, and the WHOOP and Apple Watch comparison covers wearing a band beside a watch. The Titan settings are in the last section of this guide.

08Turning data into a training decision

Data is worth collecting only if it changes what you do. Sort each morning into one of three states. Proceed when HRV and resting heart rate sit within your normal spread and you feel normal. Train and watch intensity when one signal has moved and the other has not, keeping the session's purpose with fewer repetitions or longer recoveries. Reduce training intensity when the rule from the signal and noise section is met, or when Titan shows Recovery in the Low band, 0 to 39. Rate your own readiness before you look at the device so the rating stays an independent check. The most frequent error is reacting to one morning. If you change sessions more than twice a week because of wearable data, you are probably reacting to noise.

The guide to adjusting training to recovery turns these states into a decision tree, a worked week and deload rules. Recovery scores explained covers what each band means and why two devices give different scores on the same morning.

09Failure modes

Every metric has conditions under which it misleads. Knowing them in advance lets you spot compromised data before it drives a decision.

Cardiac drift

In long sessions, especially in heat, heart rate rises while pace or power stays flat. Coyle and González-Alonso (2001, Exercise and Sport Sciences Reviews) describe stroke volume falling progressively after 10 to 20 minutes of exercise, and argue that the rising heart rate itself, more than blood diverted to the skin, drives that fall. The performance physiology guide covers the cardiovascular side. The drift inflates any load model built on heart rate zones, so a 90-minute easy run averaging 142 bpm in the first half and 156 bpm in the second at the same pace scores the second half as harder work than it was. If your load comes from zone minutes, as in TRIMP-style models and Titan's training load, expect long hot sessions to be overcounted, and read pace and perceived effort alongside heart rate.

Alcohol

Alcohol lowers HRV, raises resting heart rate, fragments sleep and raises skin temperature, which looks the same on a dashboard as overreaching or an illness starting. Pietilä and colleagues (2018, JMIR Mental Health) measured the effect in 4,098 Finnish employees. An HRV-based recovery measure during the first hours of sleep fell by 9.3 percentage points after low intake, 24.0 after moderate intake and 39.2 after high intake. Low meant up to 0.25 g of alcohol per kg of body weight, a little over one standard US drink for a 70 kg person, so even one drink showed up.

Tag Alcohol in Titan's Journal on the days you drink, and blame the drink first when recovery is down the next morning. If the numbers stay down past a day and a half, look for another cause.

Illness before symptoms

Here the wearable sees something before you feel it. Mishra and colleagues (2020, Nature Biomedical Engineering) studied smartwatch data from people who caught COVID-19. Of 25 infected people with enough data, 22 showed changes at or before symptom onset, four of them at least nine days earlier, and a real-time alert based on resting heart rate could have flagged 63 percent of cases before symptoms. If HRV drops while resting heart rate, temperature and respiratory rate climb together with no obvious cause, treat it as a possible illness. Cut training stress, protect sleep and watch for symptoms. Tag Sickness in the Journal once it declares itself, so later analysis does not count the illness against your training.

Altitude

Above roughly 1,500 meters, blood oxygen falls, heart rate rises, HRV often drops and sleep is usually worse for the first days. These are normal acclimatization responses that a wearable reports as deviations from a sea-level baseline. Do not make intensity decisions against that baseline during the first week at altitude. Once your numbers settle, treat them as a temporary baseline for the rest of the camp, and expect another adjustment on the way down.

Medication

Beta-blockers lower heart rate and blunt its response to exercise, which breaks heart rate zones and heart rate based recovery scores. Stimulants, including many ADHD medications and some decongestants, raise heart rate. Hormonal contraceptives can shift resting heart rate and temperature patterns. When you start, stop or change a dose of anything that affects heart rate or the autonomic nervous system, restart your baseline.

Dehydration, sleep position, heat and cold

Low fluid intake reduces plasma volume, which raises resting heart rate and can lower HRV. If resting heart rate is up after a long session or a hot day with too little to drink, rehydrate and see whether it settles within a day before calling it fatigue. Sleeping on the wrist that wears the watch can press the sensor for long stretches and produce gaps, spikes or odd HRV. If a night looks bad and you woke with a numb hand, discount it. Heat raises resting heart rate, lowers HRV and disturbs sleep, so during a heat wave expect the data to show thermal stress. Cold reduces blood flow to the skin, which weakens the wrist PPG signal, and a chest strap is the safer source for structured sessions in the cold.

Travel and jet lag

Waterhouse and colleagues (2007, The Lancet) describe jet lag as the result of a body clock still timed to the old zone, with lost sleep and lower performance as its main costs. Heart rate, HRV, temperature and sleep timing all shift while the clock adjusts. Treat the first days after crossing several time zones as transition data. Tag the trip with Travel and return to data-led decisions once you fall asleep and wake at local times without an alarm doing the work.

10Reading your data over weeks

Learning what moves your numbers only works across weeks, and Titan has three places built for it.

Trends

The Trends screen charts your history on one page. A period menu at the top opens on Year and also offers 30 Days, 60 Days and 6 Months. The Training load chart plots long-term and short-term training load, which Titan builds from zone-weighted workout minutes with a 42-day and a 7-day exponentially weighted average, and prints the latest load ratio and its zone under the chart. Further down, Recovery vs exertion draws both scores on one scale, Training balance shows whether effort has run ahead of recovery over the last 7 days, and Training adaptation shows how much your daily HRV varied across a rolling 30 days.

Read those charts together. A resting heart rate that drifts down over two months on the Vitals screen while load rises is a sign of rising fitness. The same drift upward during a stable load block is a problem worth finding. A load ratio climbing into Titan's High zone while Recovery slides on the Recovery vs exertion chart is the combination to catch early. The training load article explains the ratio zones Titan uses.

Journal and Journal Insights

A wearable cannot see the drink, the late meal, the flight or the argument at work. The Journal records them. It opens on yesterday, offers 15 built-in tags such as Alcohol, Caffeine, Late Meal, Travel, Sickness and Job Stress, and lets you add your own.

Journal Insights compares the days you tagged with the days you did not across the last 60 days. For each tag it takes the average Recovery or Sleep score on tagged days, subtracts the average on untagged days, and divides by the untagged average. If Recovery averages 72 on untagged days and 63 on days tagged Alcohol, the impact reads -13%. Tags used on fewer than 3 days in the window sit under Low Confidence. Tag the day the event happened, because Titan lines up each tag with the day and night it describes. Impact is a correlation, and a tag that tends to land on the same days as hard training or illness picks up their effect too.

Personal baselines

The HRV Baseline and RHR Baseline tiles on the You screen show the values and window in use. The Vitals screen shows HRV, resting heart rate and respiratory rate against a median and a shaded band covering the middle 60 percent of your days for a period you pick, from 7 days to 6 months. A reading outside the band means unusual for you in that period. It is not a clinical range. Blood oxygen is drawn against a fixed Normal band of 95 to 100 percent instead, with days below 95 percent in red.

Context changes what a single number means. A resting heart rate of 62 bpm is an ordinary day-after rise for an athlete with a 90-day median of 58 who ran intervals yesterday. For one whose median is 54 and whose resting heart rate has climbed four mornings straight, it is a warning.

11Guidance by population

Different athletes need different things from the same device.

Endurance athletes

Heart rate and HRV carry the most information here, because cardiovascular load is the main training stimulus. Resting heart rate is the most dependable daily number. It moves down over months with fitness and up over days with fatigue. HRV trend pairs with it and forms the core of the readiness approach in HRV and recovery readiness. Session heart rate feeds load models like TRIMP, so clean session data pays off. For an athlete training 10 to 15 hours a week, short sleep is the first thing to fix before reading much into HRV.

The load ratio in Titan's Trends, short-term load over long-term load, is most useful read alongside the HRV trend. A ratio rising into Titan's High zone, 1.4 to 1.6, while the HRV baseline slides is a stronger reason to back off than either signal alone.

Strength athletes

Heart rate says little about heavy resistance work. A set of three heavy deadlifts might push heart rate to 140 bpm for a moment, and the training stimulus is mechanical tension and neural drive, which no wrist sensor measures. Titan's training load counts a strength workout without heart rate data at 0.9 per minute, against 1.0 for endurance work.

Overnight HRV, resting heart rate and sleep still reflect systemic recovery, and one practical use is timing deloads. If the HRV trend has slid across a three-week loading block and resting heart rate has crept up 3 to 4 bpm from the block's start, a deload at that point often protects more of the block than waiting for bar speed to drop. Heavy lower-body sessions often lower HRV the next morning, and a one-day dip is expected. Concern starts when the rule is met, or when the dip persists three days past the last heavy session. Use bar speed, RPE and completed volume at target loads as the main training feedback.

General wellness users

For people who train for health, the most useful function is tracking behavior. Step count is accurate and tied to health outcomes, sleep duration and regularity over weeks predict how you feel better than any single morning, and resting heart rate over months tracks cardiovascular fitness. Composite scores can be a helpful daily nudge and can also cause anxiety. If checking a score every morning creates stress or guilt, check weekly, and look at average steps, average sleep and the resting heart rate trend.

Older adults

HRV falls with age. Umetani and colleagues (1998, Journal of the American College of Cardiology) measured 24-hour HRV in 260 healthy people aged 10 to 99 and found that RMSSD fell to 47 percent of its youngest-decade value by the sixth decade and then leveled off. Nunan and colleagues (2010, Pacing and Clinical Electrophysiology) pooled 44 studies and found very wide variation between healthy individuals. A lower absolute HRV in an older adult is expected, and the trend says more than the number. A fall from 25 to 18 ms over several days is the same kind of signal as a younger athlete falling from 60 to 43 ms. Because Titan scores against your own median and spread, the Recovery score already accounts for this.

Daily steps make the most useful single target for this group. Paluch and colleagues (2022, The Lancet Public Health) pooled 15 cohorts and 47,471 adults. Mortality risk fell with more steps up to 6,000 to 8,000 a day in adults 60 and older, and up to 8,000 to 10,000 in younger adults, before leveling off. Apple Watch fall detection and irregular rhythm notifications sit outside training, and for some older adults they are the best reason to wear one. For older adults on medications that affect heart rate, such as beta-blockers, calcium channel blockers or some antiarrhythmics, compare only against your own baseline on the current medication, never against population norms.

12Privacy, ownership and export

A wearable builds a detailed record of your body and your days, including heart rate, sleep, movement and often location. It is worth ten minutes of reading before you commit to a platform.

Export varies by platform. Apple Health exports everything as an XML archive from the Health app. Garmin Connect exports activities as FIT, TCX or GPX files and offers a full account export. Oura and WHOOP both provide data exports from their apps or web accounts. Raw data can move with you when you switch. Proprietary scores, baselines and trend models stay behind.

Check retention and sharing. Each company's privacy policy governs how long it keeps your data, whether it shares de-identified data for research, and what happens when you close the account. Before you choose a platform, confirm you can export your full history without an active subscription. That matters most on the day you want to leave.

Titan's own storage. Titan reads from Apple Health on your iPhone. Journal tags stay in Titan's database on the device and are not written to Apple Health or sent to a server. The direct Oura connection stores your access token in the iPhone Keychain and sends it only to Oura's API, as the Oura direct connection article describes.

13FAQ

Should I wear a tracker on my wrist or my finger

It depends on the metric that matters most to you. For overnight HRV, resting heart rate and sleep, a ring gets a strong pulse signal from the finger and moves little in bed. For exercise heart rate, a wrist device does better than a ring and a chest strap does better than both. For daily activity, a wrist device tracks walking and running more directly. Many athletes wear a ring for sleep and a watch or strap for training. If you want one device, choose by your top priority.

How do skin tone and tattoos affect wearable accuracy

For heart rate the evidence is mixed. Bent and colleagues (2020) found no statistically significant difference in wrist heart rate accuracy across skin tones, while Shcherbina and colleagues (2017) found error was higher with darker skin, higher body mass index and male sex. Blood oxygen is clearer. Pulse oximeters, including clinical ones, miss low oxygen more often in people with darker skin (Sjoding 2020). Tattoos are a known problem, and Apple's support documentation says the ink, pattern and saturation of some tattoos can block light from the sensor. Wear the device on untattooed skin, use a chest strap for training, or use a ring for overnight data.

When should I ignore my wearable data

Ignore it when you have a clear reason outside training for a compromised reading. The morning after drinking, the first week at a new altitude, a night of poor sensor contact, an acute illness and a new medication that affects heart rate all qualify. Also discount a single metric that moves while every other metric and your own sense of readiness stay normal. If checking the data every morning creates more anxiety than information, cut back to two or three looks a week and focus on weekly trends.

14Per-metric data sources in Titan

When you wear more than one device, the setting that decides data quality in Titan is the list of metric sources under Settings > Health & Connected Apps. It gives each of ten metrics, including Sleep, HRV, Resting Heart Rate and Workouts, a Primary Source and a list of Allowed Sources, so Titan reads one copy of each metric from the device you trust for it. The data sources article covers every control on that screen.

15References

Studies

• Bent B, et al. (2020). Investigating sources of inaccuracy in wearable optical heart rate sensors. npj Digital Medicine, 3, 18. doi.org/10.1038/s41746-020-0226-6
• Buchheit M. (2014). Monitoring training status with HR measures: do all roads lead to Rome? Frontiers in Physiology, 5, 73. doi.org/10.3389/fphys.2014.00073
• Chinoy ED, et al. (2021). Performance of seven consumer sleep-tracking devices compared with polysomnography. Sleep, 44(5), zsaa291. doi.org/10.1093/sleep/zsaa291
• Coyle EF, González-Alonso J. (2001). Cardiovascular drift during prolonged exercise: new perspectives. Exercise and Sport Sciences Reviews, 29(2), 88-92. doi.org/10.1097/00003677-200104000-00009
• Dial MB, et al. (2025). Validation of nocturnal resting heart rate and heart rate variability in consumer wearables. Physiological Reports, 13(16), e70527. doi.org/10.14814/phy2.70527
• Fuller D, et al. (2020). Reliability and validity of commercially available wearable devices for measuring steps, energy expenditure, and heart rate: systematic review. JMIR mHealth and uHealth, 8(9), e18694. doi.org/10.2196/18694
• Gillinov S, et al. (2017). Variable accuracy of wearable heart rate monitors during aerobic exercise. Medicine and Science in Sports and Exercise, 49(8), 1697-1703. doi.org/10.1249/MSS.0000000000001284
• Hopkins WG. (2000). Measures of reliability in sports medicine and science. Sports Medicine, 30(1), 1-15. doi.org/10.2165/00007256-200030010-00001
• Lambe R, et al. (2025). Investigating the accuracy of Apple Watch VO2 max measurements: a validation study. PLOS ONE, 20(5), e0323741. doi.org/10.1371/journal.pone.0323741
• Lambe R, et al. (2026). The accuracy of Apple Watch measurements: a living systematic review and meta-analysis. npj Digital Medicine, 9, 63. doi.org/10.1038/s41746-025-02238-1
• Lambe R, et al. (2026). Accuracy of VO2 max estimates from Apple Watch Series 10. Mayo Clinic Proceedings: Digital Health, 4(2), 100357. doi.org/10.1016/j.mcpdig.2026.100357
• Miller DJ, et al. (2022). A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults. Sensors, 22(16), 6317. doi.org/10.3390/s22166317
• Mishra T, et al. (2020). Pre-symptomatic detection of COVID-19 from smartwatch data. Nature Biomedical Engineering, 4(12), 1208-1220. doi.org/10.1038/s41551-020-00640-6
• Molina-Garcia P, et al. (2022). Validity of estimating the maximal oxygen consumption by consumer wearables: a systematic review with meta-analysis and expert statement of the INTERLIVE network. Sports Medicine, 52(7), 1577-1597. doi.org/10.1007/s40279-021-01639-y
• Nunan D, et al. (2010). A quantitative systematic review of normal values for short-term heart rate variability in healthy adults. Pacing and Clinical Electrophysiology, 33(11), 1407-1417. doi.org/10.1111/j.1540-8159.2010.02841.x
• O'Driscoll R, et al. (2020). How well do activity monitors estimate energy expenditure? A systematic review and meta-analysis of the validity of current technologies. British Journal of Sports Medicine, 54(6), 332-340. doi.org/10.1136/bjsports-2018-099643
• Paluch AE, et al. (2022). Daily steps and all-cause mortality: a meta-analysis of 15 international cohorts. The Lancet Public Health, 7(3), e219-e228. doi.org/10.1016/S2468-2667(21)00302-9
• Pietilä J, et al. (2018). Acute effect of alcohol intake on cardiovascular autonomic regulation during the first hours of sleep in a large real-world sample of Finnish employees. JMIR Mental Health, 5(1), e23. doi.org/10.2196/mental.9519
• Pipek LZ, et al. (2021). Comparison of SpO2 and heart rate values on Apple Watch and conventional commercial oximeters devices in patients with lung disease. Scientific Reports, 11, 18901. doi.org/10.1038/s41598-021-98453-3
• Shcherbina A, et al. (2017). Accuracy in wrist-worn, sensor-based measurements of heart rate and energy expenditure in a diverse cohort. Journal of Personalized Medicine, 7(2), 3. doi.org/10.3390/jpm7020003
• Sjoding MW, et al. (2020). Racial bias in pulse oximetry measurement. New England Journal of Medicine, 383(25), 2477-2478. doi.org/10.1056/NEJMc2029240
• Smarr BL, et al. (2020). Feasibility of continuous fever monitoring using wearable devices. Scientific Reports, 10, 21640. doi.org/10.1038/s41598-020-78355-6
• Stone JD, et al. (2021). Assessing the accuracy of popular commercial technologies that measure resting heart rate and heart rate variability. Frontiers in Sports and Active Living, 3, 585870. doi.org/10.3389/fspor.2021.585870
• Umetani K, et al. (1998). Twenty-four hour time domain heart rate variability and heart rate: relations to age and gender over nine decades. Journal of the American College of Cardiology, 31(3), 593-601. doi.org/10.1016/S0735-1097(97)00554-8
• Waterhouse J, et al. (2007). Jet lag: trends and coping strategies. The Lancet, 369(9567), 1117-1129. doi.org/10.1016/S0140-6736(07)60529-7

Manufacturer documentation

• Apple Developer. HKQuantityTypeIdentifier.heartRateVariabilitySDNN. developer.apple.com
• Apple Support. Get the most accurate measurements using your Apple Watch. support.apple.com/en-us/105002
• Apple Newsroom. Introducing Apple Watch Series 12, with the all-new Health Sensing System. September 2026. apple.com/newsroom
• Oura. Introducing Oura Ring 5. May 2026. ouraring.com/blog/introducing-oura-ring-5
• Garmin. Garmin expands its flagship performance smartwatch lineup with fēnix 9 and fēnix 9 Pro. August 2026. garmin.com/newsroom