How Daytlas reads your data.

Sleep score

A 0–100 score for the night from your ring’s app, built from contributors such as total sleep, efficiency, timing and sleep stages. Daytlas shows it as it arrives and never recalculates it.

How Daytlas shows it

• Day shows the score for the selected day and its difference from the average of the 30 calendar days before it, once at least 5 of those days have a score.
• Day also lists the contributors your ring’s app reports, each on its own 0–100 scale.
• Values are coloured with the bands your ring’s app uses for its scores: below 60 “Pay attention”, 60–69 “Fair”, 70–84 “Good” and 85 or more “Optimal”. A colour describes the number, not your health.
• Overview compares the latest score with the average of the up to 7 days before it.

Limits

The formula belongs to your ring’s app and can change with its updates, so older and newer scores aren’t always strictly comparable. A consumer sleep score can’t diagnose a sleep disorder.

Research

• Khosla et al., 2018: Consumer sleep technology can support a conversation with a clinician but is not validated or cleared for diagnosing or treating sleep disorders.
• de Zambotti et al., 2024: A Sleep Research Society review found that consumer wearables now outperform traditional actigraphy for sleep, with clear limits such as misclassified wake, naps and artefacts.
• de Zambotti et al., 2019: Reviews how consumer sleep trackers are validated and cautions that proprietary algorithms and firmware updates can change their numbers.

Also cited: Buysse, 2014

Total sleep

The time you were asleep during the main sleep period of the day, not counting time awake in bed.

How Daytlas shows it

• When a day has more than one sleep period, the longest one (the main sleep) carries that day’s sleep metrics. Naps and rest periods don’t add to it; on Day you can still open an additional sleep period.
• Durations arrive in seconds and are shown in hours.
• Monthly sleep summaries only count nights longer than 0 and up to 24 hours.

Limits

Wearables are good at telling sleep from wake but less precise than a sleep lab, especially on restless nights. Sleep needs differ between people; Daytlas shows your numbers without judging them against a target.

Research

• Watson et al., 2015: A joint panel of two sleep societies recommends that adults regularly sleep 7 hours or more a night.
• Hirshkowitz et al., 2015: An expert panel judged 7–9 hours appropriate for adults and 7–8 hours for older adults, and noted that needs differ from person to person.
• Chinoy et al., 2021: Compared with lab sleep recordings, most consumer trackers detected sleep well, were weaker at detecting wake, gave inconsistent sleep stages and did worse on disrupted nights.

Also cited: Chaput et al., 2018

Time in bed and sleep efficiency

In Daytlas: Time in Bed · Sleep Efficiency

Time in bed is the length of the main sleep period, from going to bed to getting up. Sleep efficiency is the share of that time you spent asleep, as a percentage.

How Daytlas shows it

• Both come from the main sleep period as your ring’s app reports them; Daytlas doesn’t recalculate efficiency.
• Day compares the night with the average of the previous 30 days; Trends can average it by week, month, quarter or year.

Limits

Efficiency depends on what counts as time in bed. Reading in bed or lying awake in the morning lowers it even when the sleep itself didn’t change.

Research

• Ohayon et al., 2017: Experts agreed that time to fall asleep, awakenings, time awake after falling asleep and sleep efficiency are reasonable markers of sleep quality, but found little agreement on sleep stages.
• Lee et al., 2023: Eleven consumer sleep trackers varied widely in sleep stage accuracy against lab recordings, and the wearables showed systematic bias in sleep efficiency.
• Ancoli-Israel et al., 2015: A society guide to movement-based sleep monitoring: what it estimates well, such as timing and duration, and where it struggles, such as lying awake but still.

Also cited: Reed & Sacco, 2016

Time to fall asleep

In Daytlas: Sleep Latency

How long it took to fall asleep after the sleep period began.

How Daytlas shows it

• Reported in seconds and shown in whole minutes.
• Day compares it with the average of the previous 30 days and shows the direction of a change without calling it good or bad.

Limits

It’s an estimate: the exact moment of falling asleep is hard to pin down without a sleep lab. Typical values also change with age.

Research

• Ohayon et al., 2017: Experts agreed that time to fall asleep, awakenings, time awake after falling asleep and sleep efficiency are reasonable markers of sleep quality, but found little agreement on sleep stages.
• Boulos et al., 2019: Across 169 studies of healthy adults, each decade of age brought about 10 minutes less sleep and lower efficiency, while the shares of deep and REM sleep barely changed.
• Ohayon et al., 2004: Across 65 studies of healthy people, total sleep, sleep efficiency and the share of deep and REM sleep fell with age in adults, while time to fall asleep and time awake rose.

Deep, light, REM and awake time

In Daytlas: Deep Sleep Time · Light Sleep Time · REM Sleep Time · Awake Time

Your ring estimates which stage of sleep you were in through the night. Daytlas shows the time in each stage and, on Day, the stage timeline.

How Daytlas shows it

• The timeline uses 30-second estimates when your data includes them, otherwise 5-minute estimates, drawn from bedtime to wake-up.
• Stage totals come from the main sleep period and are shown in hours.
• Trends lets you chart each stage and overlay up to three other metrics.

Limits

Stage estimates from any wearable agree with a sleep lab much less than sleep-versus-wake does. Read stage time as a rough trend over weeks rather than an exact number for one night.

Research

• Boulos et al., 2019: Across 169 studies of healthy adults, each decade of age brought about 10 minutes less sleep and lower efficiency, while the shares of deep and REM sleep barely changed.
• Lee et al., 2023: Eleven consumer sleep trackers varied widely in sleep stage accuracy against lab recordings, and the wearables showed systematic bias in sleep efficiency.
• Miller et al., 2022: Six wearables, a ring among them, matched lab recordings well for sleep versus wake but agreed on the exact sleep stage for only about 50–65% of the night.

Also cited: Chinoy et al., 2021

Bedtime, wake-up time and midpoint

In Daytlas: Bedtime · Wake-up Time · Midpoint

When your main sleep period started and ended, and the clock time halfway between them.

How Daytlas shows it

• Times use the local clock recorded with your data, so a night in another time zone keeps its own clock.
• Bedtimes after midnight continue past 24:00 (01:30 becomes 25:30), so late and early bedtimes sit next to each other on charts.
• Weekly and monthly averages of clock times use a circular mean, so 23:00 and 01:00 average to midnight rather than noon.
• Your Year draws every night as a bar from bedtime to wake-up; its average bedtime and wake-up time use the same circular mean.
• Correlations leave clock times out because the wrap at midnight breaks a straight-line comparison.

Limits

Bedtime is the start of the sleep period your ring detected, which may not match when you meant to go to sleep.

Research

• Sletten et al., 2023: An expert panel agreed that consistent sleep and wake times matter for health, safety and performance.
• Fischer et al., 2017: In about 54,000 time-use diaries, mid-sleep on weekends was latest around age 19 and moved earlier with age, with sex differences that reversed around 40.
• Roenneberg et al., 2003: Timed sleep separately on work days and free days and found large differences between people and between the two kinds of day, with light exposure shifting sleep timing.

Also cited: Berens, 2009

Sleep regularity

How consistent your sleep timing and duration are from night to night. Daytlas doesn’t reduce this to a single score; it gives you views to see it.

How Daytlas shows it

• Your Year’s sleep rhythm view draws every night as a bar from bedtime to wake-up, with gaps where nights are missing.
• The weekday view on Overview shows the median bedtime, or another metric, for each day of the week, with the middle half of values around it.
• Your months, on the Year page, show the average and the standard deviation of main sleep for each month, and name the steadiest month (the lowest standard deviation) once at least two months have 7 or more nights.

Limits

Regularity measures need more than a week of data to settle. These views describe your schedule; they don’t rate it.

Research

• Chaput et al., 2020: Later sleep timing and more night-to-night variability were generally linked to worse health outcomes, though the quality of evidence ranged from very low to moderate.
• Windred et al., 2024: In about 61,000 UK Biobank participants, more regular sleep timing was linked to lower mortality risk and predicted it more strongly than sleep duration did.
• Phillips et al., 2017: In 61 students followed for 30 days, irregular sleepers had later body-clock timing and lower grades; the study introduced the Sleep Regularity Index.

Also cited: Fischer et al., 2021

Readiness score

A 0–100 score from your ring’s app that combines signals such as resting heart rate, HRV, body temperature, recent sleep and activity. Daytlas shows it as it arrives and never recalculates it.

How Daytlas shows it

• Day shows the score, its contributors (each 0–100) and the difference from the average of the previous 30 days, once at least 5 of those days have a score.
• Values are coloured with the bands your ring’s app uses for its scores: below 60 “Pay attention”, 60–69 “Fair”, 70–84 “Good” and 85 or more “Optimal”. A colour describes the number, not your health.
• Overview averages the score over the last 30 days and compares it with the 30 days before and with the same 30 days a year earlier.

Limits

Because the score is built from other measurements, it correlates with them by design: a strong link between readiness and HRV in Trends partly reflects how the score is made. Heart-rate-based recovery measures can’t tell every kind of fatigue apart.

Research

• Kellmann et al., 2018: Experts agreed that balancing stress and recovery needs regular monitoring, because responses to training and recovery vary within and between people.
• Düking et al., 2021: Across 8 studies, training guided by wearable HRV had a medium effect on submaximal physiology but only a small, non-significant effect on performance.
• Bellenger et al., 2016: In endurance athletes, resting HRV rose slightly when training went well and barely changed with overreaching, so heart-rate measures alone can’t tell every kind of fatigue apart.

Also cited: Buchheit, 2014

Average HRV

Heart rate variability is the variation in time between heartbeats. Your ring reports the night’s average in milliseconds, based on RMSSD, a standard measure of beat-to-beat change.

How Daytlas shows it

• Daytlas uses the average HRV of the main sleep period for each day.
• Day shows the overnight HRV curve with a reference line at the average of the previous 30 days, and compares the night’s average with it.
• Overview shows the average of the last 30 days, its change from the 30 days before and from the same period a year earlier, and the highest recorded day.
• Trends shows daily values with a dashed line for the prior 60-day mean and a shaded band of ±1 standard deviation.

Limits

HRV depends strongly on measurement conditions, and one night says little. The band is your own spread, not a clinical range.

Research

• Task Force, 1996: Set the standard definitions of heart rate variability measures, RMSSD among them, and described how recording conditions affect them.
• Dobbs et al., 2019: Across 23 studies, HRV from portable devices differed from ECG by a small but variable amount, which the authors judged acceptable for use outside the lab.
• Cao et al., 2022: In 35 healthy adults sleeping at home, a ring’s nightly heart rate and RMSSD agreed closely with a chest ECG, while some other HRV measures were less accurate.

Also cited: Laborde et al., 2017

Reading HRV over time

HRV differs a lot between people and changes with age, so the useful comparison is with your own history.

How Daytlas shows it

• Daytlas never compares your HRV with population norms; every comparison in the app is with your own earlier days.
• Weekly averages in Trends smooth out night-to-night noise before you read much into a change.
• Shifts in your baseline, on Overview, point out dates where your 30-day average before and after differs by more than your usual spread.

Limits

Many things move HRV at once, such as training, alcohol, illness and stress, so a change is a reason to look at the context, not an explanation.

Research

• Altini & Plews, 2021: Across 28,175 people, HRV was lower with age, and day-to-day HRV responded to stressors such as hard training, alcohol and illness: sensitive, but not specific.
• Natarajan et al., 2020: In 8 million wearable users, HRV fell with age and varied strongly over the day, while RMSSD did not differ notably between men and women.
• Umetani et al., 1998: In 260 healthy people aged 10 to 99, HRV fell with age; RMSSD fell fastest and levelled off from about the sixth decade.

Also cited: Shaffer & Ginsberg, 2017

Resting heart rate

In Daytlas: Average Resting HR · Lowest Resting HR

Your heart rate during the main sleep period. Average resting heart rate is the mean across the period; lowest resting heart rate is the lowest value your ring recorded in it.

How Daytlas shows it

• Day shows the overnight heart-rate curve with a reference line at the average of the previous 30 days, and compares both values with their 30-day averages.
• Overview’s “Your body over time” and its relationship chart use the lowest resting heart rate.
• Trends and Tag Lab can show both.

Limits

Each person has their own normal resting heart rate, and it changes with season, sleep and much else. Population studies link resting heart rate with long-term health on average; Daytlas doesn’t apply those findings to you.

Research

• Zhang et al., 2016: Pooling 46 cohort studies, a higher resting heart rate went with higher mortality on average; this describes populations, not any one person.
• Stone et al., 2021: Against a multi-lead ECG, seven consumer technologies measured resting heart rate accurately, while their RMSSD accuracy varied widely.
• Quer et al., 2020: In 92,457 wearable users, each person’s resting heart rate was fairly stable over time, but what was normal differed between people by up to 70 bpm, with a small seasonal pattern.

Also cited: Speed et al., 2023

Heart rate during the day

The heart-rate samples your ring records through the day, outside your sleep.

How Daytlas shows it

• Day plots every sample of the selected day in this device’s time zone. Gaps longer than 15 minutes stay visible as breaks in the line; samples at the same moment are averaged.
• Daytime resting heart rate is the average of that day’s samples marked as awake or resting. Today’s value is partial.
• It’s compared with the average of the previous 30 days that have samples, once there are at least 5.
• Imported export files don’t include these samples, so this view needs a live connection.

Limits

Optical heart-rate sensors are reasonably accurate at rest and less accurate during movement.

Research

• Fuller et al., 2020: Across 158 studies, wearables counted steps and measured heart rate well in the lab, but no brand estimated energy expenditure accurately.
• Zhang et al., 2020: Across 44 studies, wrist-worn optical heart rate was close to the reference during sleep, rest, walking and running, and less accurate during resistance training and cycling.
• Bent et al., 2020: Optical heart-rate sensors were reasonably accurate at rest, errors were about 30% larger during activity, and accuracy did not differ significantly across skin tones.

Also cited: Shcherbina et al., 2017

Cardiovascular age

An estimate from your ring’s app of the age of your cardiovascular system, based on the shape of your pulse signal and estimated pulse wave velocity. Daytlas shows it but doesn’t calculate it.

How Daytlas shows it

• Overview shows the latest reading from the past year, the average of the readings in the 30 days before it (once there are at least 5) and the last 90 days as a chart with an average line.
• Readings outside 18 to 100 years are ignored as invalid.
• Imported export files don’t include it.

Limits

It’s an estimate of vascular ageing, not a clinical measurement of arterial stiffness and not a diagnosis. Look at the trend over weeks rather than single readings.

Research

• Laurent et al., 2006: A European expert network summarised how arterial stiffness is measured, pulse wave velocity among the methods, and the issues that affect the measurement.
• Reference Values for Arterial Stiffness’ Collaboration, 2010: In 16,867 people from 13 European centres, pulse wave velocity, a measure of arterial stiffness, rose with age and blood pressure.
• Charlton et al., 2022: Reviews how the shape and timing of the optical pulse signal relate to vascular ageing, and how far such estimates have been validated.

Also cited: Allen, 2007

Activity score

A 0–100 score from your ring’s app that reflects how active you were and how your activity was spread through the day and week. Daytlas shows it as it arrives and never recalculates it.

How Daytlas shows it

• Day shows the score, its contributors (each 0–100) and the difference from the average of the previous 30 days, once at least 5 of those days have a score.
• Values are coloured with the bands your ring’s app uses for its scores: below 60 “Pay attention”, 60–69 “Fair”, 70–84 “Good” and 85 or more “Optimal”. A colour describes the number, not your health.
• Today’s score is compared with previous full days, so it shows progress rather than a final result.

Limits

The score’s targets and formula belong to your ring’s app. Today’s activity is incomplete until the day ends.

Research

• Bull et al., 2020: Recommends that adults get 150–300 minutes of moderate or 75–150 minutes of vigorous activity a week, and spend less time sedentary.
• Ekelund et al., 2019: Pooling accelerometer data from about 36,000 adults, any physical activity, light activity included, went with lower mortality, with the steepest difference at low levels.
• de Zambotti et al., 2019: Reviews how consumer sleep trackers are validated and cautions that proprietary algorithms and firmware updates can change their numbers.

Steps and walking equivalency

In Daytlas: Steps · Walking Equivalency

Steps your ring counted during the day, and walking equivalency: the day’s activity expressed as a walking distance.

How Daytlas shows it

• Distances arrive in metres and are shown in kilometres.
• Day compares a day with the average of the previous 30 days; today so far is compared with previous full days.
• Your Year can colour every day of the year by steps.

Limits

Step counts from wearables hold up well in lab tests but differ between devices and activities, so compare your own days rather than devices.

Research

• Banach et al., 2023: Across cohort studies, more daily steps went with lower all-cause and cardiovascular mortality, starting from roughly 2,300 to 3,900 steps a day.
• Paluch et al., 2022: Across 15 cohorts, more daily steps went with lower mortality, levelling off at about 6,000–8,000 steps for adults aged 60 and over and 8,000–10,000 for younger adults.
• Fuller et al., 2020: Across 158 studies, wearables counted steps and measured heart rate well in the lab, but no brand estimated energy expenditure accurately.

Also cited: Tudor-Locke et al., 2011

Active and total energy

In Daytlas: Activity Burn · Total Burn

Estimated energy burned through activity (active energy) and in total, including the energy your body uses at rest (total energy), in kilocalories.

How Daytlas shows it

• Shown as reported and compared with the average of the previous 30 days on Day.
• Trends can chart both, daily or as weekly and monthly averages.

Limits

No consumer wearable estimates energy expenditure precisely. Read these as rough estimates, most useful for comparing your own days with each other.

Research

• Fuller et al., 2020: Across 158 studies, wearables counted steps and measured heart rate well in the lab, but no brand estimated energy expenditure accurately.
• O’Driscoll et al., 2020: Across 60 studies, the accuracy of energy estimates from wrist and arm devices varied with the activity, and adding heart rate to movement data helped.
• Shcherbina et al., 2017: Seven wrist devices measured heart rate reasonably well in the lab, but none estimated energy expenditure with an error below 20%.

Activity intensity and MET

In Daytlas: Average MET · High Activity · Medium Activity · Low Activity

MET, the metabolic equivalent, expresses effort as a multiple of the energy you use at rest. Your ring estimates it through the day and sorts your time into low, medium and high activity.

How Daytlas shows it

• Day’s activity intensity chart shows the MET estimates of the selected day as bars, up to the moment the data was loaded.
• Time in each intensity arrives in seconds and is shown in minutes on Day and in hours in Trends.
• Average MET is the day’s average as reported.

Limits

The intensity thresholds are set by your ring’s app and may not match the minutes counted in physical activity guidelines.

Research

• Bull et al., 2020: Recommends that adults get 150–300 minutes of moderate or 75–150 minutes of vigorous activity a week, and spend less time sedentary.
• Herrmann et al., 2024: Lists the measured energy cost, in METs, of 1,114 activities for adults, from resting to vigorous sport.
• Migueles et al., 2017: Shows how choices such as non-wear rules, valid-day definitions and intensity cut-points change the activity estimates that come out of an accelerometer.

Also cited: Byrne et al., 2005

Inactive, resting and non-wear time

In Daytlas: Inactive Time · Resting Time · Non-wear Time

Inactive time is time awake with little movement. Resting time covers rest and sleep. Non-wear time is when the ring wasn’t worn.

How Daytlas shows it

• Reported in seconds and shown in minutes on Day and in hours in Trends.
• Daytlas doesn’t fill non-wear time: when the ring was off, the readings for that time are simply missing.

Limits

Long non-wear periods leave the rest of the day’s activity numbers incomplete. Sitting still for a while isn’t the same as being inactive overall.

Research

• Tremblay et al., 2017: Agrees common definitions that separate sedentary behaviour, low-energy sitting or lying while awake, from physical inactivity.
• Patterson et al., 2018: Independent of physical activity, more total sitting went with higher risk, rising above about 6–8 hours a day for mortality.
• Choi et al., 2011: Validated a rule for telling non-wear from still wear in accelerometer data, which changes how much time counts as sedentary.

Temperature deviation and trend

In Daytlas: Temperature Deviation · Temperature Trend Deviation

How your skin temperature during sleep differed from your own baseline, in degrees Celsius. The trend deviation is a smoothed version that changes more slowly. Both come from your ring’s app.

How Daytlas shows it

• Day shows the deviation for the selected day and compares it with the average deviation of the previous 30 days.
• Overview’s “Your body over time” charts the daily deviation.
• Trends can chart both, daily or as weekly and monthly averages.

Limits

Skin temperature changes with room temperature, bedding, timing and much else. A deviation describes your recorded data; it isn’t a fever reading or a diagnosis.

Research

• Smith et al., 2010: Small wireless skin sensors gave valid skin temperature readings against a reference, though devices differed by up to about 1 °C with ambient temperature and exercise.
• Harding et al., 2019: Reviews evidence that falling asleep is closely tied to the body and brain cooling down.
• Okamoto-Mizuno & Mizuno, 2012: Room temperature, humidity and bedding shape sleep; heat in particular increases wakefulness and reduces deep and REM sleep.

Also cited: Kräuchi, 2007

Breathing rate

In Daytlas: Respiratory Rate

Your average number of breaths per minute during the main sleep period, estimated from your ring’s signals.

How Daytlas shows it

• Day compares it with the average of the previous 30 days; Trends and Tag Lab can show it over time.

Limits

Breathing rate is estimated indirectly rather than measured with an airflow sensor, so small changes can be within measurement error.

Research

• Douglas et al., 1982: In 19 healthy adults, breathing became faster and shallower during sleep, and ventilation fell most during REM sleep.
• Nicolò et al., 2020: Argues that breathing rate responds to stress, heat, effort and fatigue more readily than most other vital signs.
• Charlton et al., 2018: Reviews algorithms that estimate breathing rate from heart and pulse signals, and how their accuracy should be tested.

Blood oxygen (SpO2)

In Daytlas: Average Oxygen Saturation

The average estimated oxygen saturation of your blood during sleep, as a percentage. It’s available only when your ring and account provide it.

How Daytlas shows it

• Trends shows the nightly average; Daytlas doesn’t derive anything else from it.
• If this data isn’t available, the rest of Daytlas works without it.

Limits

Pulse oximetry is less accurate with movement or poor circulation, and studies have found that oximeters can overestimate oxygen levels in people with darker skin. A ring isn’t a medical oximeter.

Research

• Jiang et al., 2023: Consumer smartwatches differed in blood oxygen accuracy against a clinical oximeter, and some failed to give a reading in about 30% of attempts.
• Sjoding et al., 2020: In hospital patients, pulse oximeters missed low blood oxygen about three times as often in Black patients as in White patients.
• Gries & Brooks, 1996: In healthy sleepers, median oxygen saturation was about 96.5%, the lowest values averaged about 90%, and people over 60 ran slightly lower.

Also cited: Jubran, 2015

Days, nights and missing data

Every chart starts from one row per calendar day, built from the daily summaries and sleep periods your ring’s app provides.

How Daytlas shows it

• Sleep periods belong to the day your ring’s app assigns them to; the longest one carries that day’s sleep metrics.
• Missing readings stay missing. Daytlas never fills a gap with zero or an invented value, and averages use only the days that have a reading.
• Lines break where a day is missing; a faint dashed line bridges gaps inside the chart so you can follow the trend without mistaking it for data.
• Every comparison needs a minimum number of readings, listed in the sections below. Below that minimum, Daytlas leaves the comparison out rather than guess.

Limits

If readings tend to be missing on particular days, for example nights you took the ring off, averages describe only the days that were recorded.

Research

• Tackney et al., 2021: Proposes using wear time to define missing wearable data, since a device that records nothing can mean either no wear or no movement.
• Cho et al., 2021: Researchers and data-quality experts named completeness, accuracy and plausibility as the main quality problems of wearable data.
• Sterne et al., 2009: Explains that analysing only complete records can bias results when the reason data are missing is related to the values themselves.

Averages and periods

How Daytlas groups days into weeks, months, quarters and years.

How Daytlas shows it

• Trends shows daily values or averages them by week (Monday to Sunday), month, quarter or year. Each period is the plain mean of the days with a reading; clock times use a circular mean.
• Each Trends chart draws a dashed average line for the values on screen.
• The Trends section header splits the readings in your selection into an older and a newer half and shows the difference between their averages. It doesn’t compare with a previous period.
• Your Year groups days into calendar weeks (Monday to Sunday, cut at the start and end of the year) and months.

Limits

Averaging smooths noise but can also hide short changes, and it makes two metrics look more strongly related than their daily values are.

Research

• Cremers & Klugkist, 2018: A tutorial showing that data on a circle, such as angles or clock times, need their own methods because 0 and 360 degrees are the same point.
• Berens, 2009: Explains why angles and times of day need circular statistics, such as the circular mean, instead of ordinary averages.
• Robinson, 1950: Showed that correlations between group averages can differ sharply from correlations between the underlying individual values.

Baselines and bands

A baseline is your own recent average, used as the reference for a comparison.

How Daytlas shows it

• Day: each reading is compared with the average of the 30 calendar days before the selected day, not counting the day itself. The comparison appears once at least 5 of those days have a reading; otherwise Day says “Not enough history”.
• Trends, daily view: the dashed line is the mean of the previous 60 calendar days within your selected range, not counting the day itself. The shaded band is ±1 standard deviation around it and needs at least 2 earlier readings.
• Overview’s score cards compare the latest score with the average of the up to 7 days before it.

Limits

A baseline describes your own recent past, not a healthy or target range. A difference smaller than your usual day-to-day spread is likely ordinary variation.

Research

• Buchheit, 2014: Recommends reading a change in resting heart rate or HRV against its usual day-to-day error and the smallest change that matters.
• Plews et al., 2013: Argues that HRV is better followed with appropriate averaging over several days than as single daily values.
• Hopkins, 2000: Defines typical error, the normal spread of a person’s repeated measurements, as the yardstick for judging whether a change is real.

Also cited: Plews et al., 2013

Comparison windows

Side-by-side comparisons of two stretches of your own history.

How Daytlas shows it

• Last 30 days (Overview): the average of the past 30 calendar days, including today, compared with the 30 days before and with the same 30 days a year earlier. Each comparison needs at least 15 readings in both windows. The highest recorded day is the highest single value in the current window.
• This month vs last: the last 30 days against the 30 days before, as a percentage change, for metrics with at least 15 readings in each window, sorted by the size of the change.
• Your week: the last 7 calendar days against the 60 days before them. A metric qualifies with at least 4 readings this week and 30 in the baseline; the three metrics furthest from their baseline, measured in standard deviations, are shown.
• Colours show direction only, up or down, never better or worse.

Limits

Two windows can differ simply because one contained unusually high or low days; the next window often drifts back towards your average.

Research

• Hopkins, 2000: Defines typical error, the normal spread of a person’s repeated measurements, as the yardstick for judging whether a change is real.
• Swinton et al., 2018: Sets out how measurement error, typical error and the smallest worthwhile change help decide whether one person’s change is real.
• Hopkins et al., 2009: Recommends judging effects by their size and precision, and showing standard deviations, rather than relying on significance tests alone.

Also cited: Barnett et al., 2005

Medians, percentiles and distributions

Ways to see your typical range rather than one average.

How Daytlas shows it

• Your usual, on your profile: the median and the middle half (25th to 75th percentile) of the last 90 calendar days, from at least 14 recorded days. The track behind it spans the 5th to 95th percentile so one rare day can’t flatten it.
• Weekday patterns, on Overview: the median and the middle half for each day of the week across your history, with the number of days behind each.
• “Where do your last 30 days sit?”, on Overview: every recorded day sorted into 12 equal-width ranges, showing the share of earlier days and of the last 30 days in each. It needs at least 10 recorded days.
• Percentiles use linear interpolation between neighbouring values, a common definition in statistics software.

Limits

With only a few days, medians and quartiles move noticeably as new days arrive. Weekday differences can reflect your schedule rather than the day itself.

Research

• Weissgerber et al., 2015: Many different distributions can produce the same bar or line chart, so showing the full distribution helps readers judge the data.
• Wittmann et al., 2006: Describes social jetlag: many people sleep at different times on work days and free days, late chronotypes most of all.
• Krzywinski & Altman, 2014: Explains how box plots show the median, the middle half of the values and their spread, which makes distributions easy to compare.

Also cited: Hyndman & Fan, 1996

Correlations

A correlation describes how closely two metrics rise and fall together in your own data.

How Daytlas shows it

• Daytlas uses Pearson’s correlation coefficient, r, from −1 to +1, on pairs of values from the same day, or the same week, month, quarter or year when you average.
• Only periods where both metrics have a value count; gaps are never filled. At least 5 pairs are needed, and no r is shown when one of the metrics doesn’t vary.
• Clock times (bedtime, wake-up time and midpoint) are left out because they wrap at midnight.
• Wording follows the size of r, ignoring its sign: below 0.2 “No clear linear pattern”, below 0.4 weak, below 0.7 moderate, and 0.7 or more strong. The number of pairs is shown with the result.

Limits

A correlation is not cause and effect, and Daytlas doesn’t run significance tests. With a few dozen pairs, r can swing widely, and a matrix of many pairs will show some strong-looking values by chance. Day-to-day values depend on the days before them, which can make two metrics look more related than they are, and so can averages and scores built from the same inputs.

Research

• Schober et al., 2018: A tutorial on when to use Pearson’s r, what its scale from −1 to +1 means and how to interpret its strength.
• Dean & Dunsmuir, 2016: Shows that correlating two time series whose values depend on their own past can suggest relationships that are not there.
• Schönbrodt & Perugini, 2013: Simulations showed that correlations from small samples swing widely; in typical cases, stable estimates need around 250 observations.

Also cited: Benjamini & Hochberg, 1995

Shifts in your baseline

Dates around which a metric’s level seems to have changed and stayed changed for a while.

How Daytlas shows it

• For average resting heart rate, HRV, sleep score and total sleep, Daytlas steps through your history at least 7 days at a time and compares the average of the 30 days before a date with the 30 days from it.
• A shift is listed when the two averages differ by more than one standard deviation of the metric’s whole history, both windows have at least 24 readings, and the metric has at least 60 readings overall.
• Shifts in one metric are at least 30 days apart; the card shows the four most recent.

Limits

These are descriptive candidates, not statistically tested change points, and they say nothing about why a change happened.

Research

• Truong et al., 2020: Reviews methods for finding the points where a time series shifts, each built from a cost function, a search method and a limit on the number of changes.
• Aminikhanghahi & Cook, 2017: Surveys and compares methods for detecting abrupt changes in time series, from supervised to unsupervised algorithms.
• Barnett et al., 2005: Explains why unusually high or low values tend to be followed by values closer to the average, which can look like real change.

Tag Lab comparisons

How your recorded metrics differ on the days after a habit you tagged in your ring’s app.

How Daytlas shows it

• A tag covers every day from its start to its end date, and tags with the same name are grouped.
• For each metric, Daytlas compares the average on days that follow a tagged day with the average on all other recorded days.
• Each group needs at least 5 days with a reading, and the number of days in each group is shown.
• The metrics compared are average HRV, average resting heart rate, sleep score, readiness score, total sleep, deep sleep and breathing rate, sorted by the size of the difference relative to the comparison average.

Limits

These are unadjusted averages. An untagged day doesn’t prove the habit was absent, other habits and schedules can explain a difference, and a few unusual days can move a small group. With seven metrics compared at once, one of them can differ by chance. Read a difference as a question, not a result.

Research

• Shaffer et al., 2018: Reviewed 54 published single-person trials, found that many lacked the detail needed to judge their results, and called for more rigour.
• Daza, 2018: Sets out when an association in one person’s self-tracked data could be read as an effect, allowing for trends, carryover and autocorrelation.
• Barnett et al., 2005: Explains why unusually high or low values tend to be followed by values closer to the average, which can look like real change.

Also cited: Bender & Lange, 2001

Milestones, streaks and records

Runs of high scores and your highest and lowest recorded values.

How Daytlas shows it

• A streak counts consecutive calendar days with a sleep or readiness score of 85 or more. A missing day ends a streak, and the current streak only counts if the latest score is from today or yesterday.
• The longest streak is the longest such run in your history; the highest recorded HRV is your single highest nightly average.
• Your Year lists the highest and lowest day and, from months with at least 7 recorded days, the highest and lowest month.

Limits

Records are extremes by definition, and the values after a record usually move back towards your average. Higher and lower describe values, not health.

Research

• Gilovich et al., 1985: Shooting records didn’t support players’ and fans’ belief in hot streaks, suggesting people see streaks in largely random sequences.
• Barnett et al., 2005: Explains why unusually high or low values tend to be followed by values closer to the average, which can look like real change.
• Morton & Torgerson, 2003: Shows how regression to the mean can make a change look like the effect of an action when it is partly chance.

Your Year and your months

Whole-year views that show seasons and variability rather than single days.

How Daytlas shows it

• The three rings show weekly averages (Monday to Sunday, cut at the year’s edges) of the sleep, readiness and activity scores. Darker means higher within each metric for that year; grey means no reading.
• The calendar colours each day on a 7-step scale from the year’s lowest to its highest value of the chosen metric.
• Your months show the 12 months ending with your latest record: the average main sleep per month, its standard deviation and how many days are missing. The longest and steadiest months come from months with at least 7 nights, once at least two months qualify.

Limits

Colour scales are relative to your own year, so the same colour can mean different values in different years or metrics.

Research

• Bei et al., 2016: Argues that night-to-night variability is a dimension of sleep in its own right, beyond the average, and reviews 53 studies of what it relates to.
• Lemola et al., 2013: In 441 adults tracked for a week, night-to-night variability in sleep duration, more than its average, went with poorer sleep quality and well-being.
• Crameri et al., 2020: Shows how uneven or red-green colour maps distort data, and recommends perceptually even scales that also work with colour-vision deficiency.