How it works
How Levend works out your scores
What goes into each number, what moves it, and what it cannot tell you.
Your band records your heart rate, the time between heartbeats and the movement of your wrist. Levend collects that from the band over Bluetooth and works everything out on your iPhone. With an Apple Watch instead, Levend reads what the Watch saved in Apple Health (below). This page explains how, for anyone curious. The numbers are estimates for your own insight: Levend is not a medical device, and none of its scores is a diagnosis.
Where the formulas come from
Levend's scores are its own. Some building blocks come from OpenStrap, an open-source project: how a night's sleep and its stages are found, and how the nightly heart rate, heart rate variability and breathing rate are measured. Levend checks its versions of those against OpenStrap's own tests. The scores built on top were designed for Levend, from published research where it exists and from stated product choices where it does not. Where a number is a choice rather than a research result, this page says so.
A few rules apply to every score:
- You against yourself. Most scores compare you with your own recent nights and days, not with other people.
- A gap, never a guess. When the band did not record enough, there is no score, never a made-up number or a zero. Missing time stays visible in charts.
- Versioned. Every score records the version of the formula that made it, and formula changes normally get a new version, so days you already finished keep the score they had. The exceptions so far are all Load's, explained under Load.
- Not validated. No score has been checked against a laboratory or medical reference. The constants are starting values that may change.
Sleep
A score from 0 to 100 for your main sleep of last night. Most of it is how long you slept; up to a fifth is how regular your timing was.
- How long you slept
- Levend counts the time you were actually asleep, leaving out time awake and time the band did not measure, and compares it with 8 hours. The share is squared, so short nights lose points quickly: 4 hours gives 25, 6 hours 56, and 8 hours or more 100. Eight hours is a provisional reference for adults, not an estimate of what you personally need.
- How regular your timing was
- Levend compares when you fell asleep and when you woke up with your average times over the seven nights before. An average difference of up to 30 minutes counts in full, 1 hour 45 minutes half, and 3 hours or more not at all. This part grows with your history: with one, two or three earlier nights it weighs a fifteenth, two fifteenths and then a fifth of the score. Without earlier nights the score is duration alone.
- When the night starts
- The band first finds a long stretch in which your wrist rests. Lying still awake in bed also rests the wrist, so since 22 October 2026 the night starts later: at the first 20 minutes that are scored as sleep without any waking and in which your wrist is as still as it normally is in sleep. Time before that counts as awake before bed, not as sleep. Nights before that date keep their earlier start.
- What does not count
- Sleep stages, heart rate, heart rate variability, naps and your activity the day before. Levend does not calculate a personal sleep need or a sleep debt.
- Sleep stages
- The night is divided into awake, light, deep and REM sleep, estimated from your movement and heart rhythm. Stages describe the night; they are not part of the score. A wrist band estimates them far less exactly than a sleep laboratory.
- When there is no score
- The score waits until the night has finished arriving from the band, and needs at least 80% of the night measured. A night ends once the band's data runs a few minutes past your wake-up, so if you open the app in bed, the night still shows as arriving until you are up. Without that you see no score rather than a low one. The band cannot confirm a night without sleep, so a missing night is shown as missing, not as zero.
- Once it appears
- A night's score stays put once shown. It changes if you edit the night yourself, or once automatically if a later reading finds the night at least 15 minutes longer, for example because you went back to sleep.
Why: getting enough sleep and keeping regular times both have scientific support. The exact mappings, such as the 8 hours, the squaring and the one-fifth weight, are transparent product choices, not a clinically validated scale.
Readiness
An index of how last night's heart signals compare with your own usual nights. A typical night for you scores 65.
- Inputs
- Your heart rate variability (HRV) while asleep counts for 70%, your resting heart rate, the average of your heart rate while asleep, for 30%. Both are measured only over the time scored as sleep, so a stretch lying awake does not count. Both come from the same heartbeats, so Levend does not treat them as two independent votes. The 70/30 split is a transparent starting assumption, not a value from the research.
- Your baseline
- Your usual is the median of your last 28 nights within the previous six weeks, with a measure of how much you normally vary. Medians are used because one odd night barely moves them. After a gap of more than 14 days without a night, the baseline starts again.
- What moves it
- HRV above your usual and resting heart rate below it raise the score; the opposite lowers it. A change within a quarter of your normal variation counts as no change. Good news is capped sooner than bad news, because a high HRV is not always good: it can also rise during overreaching. The best possible night reads 99.
- Breathing rate
- Breathing rate can only lower Readiness, when it is clearly above your usual and once there are seven nights to compare with. A lower rate never raises it: illness tends to raise breathing rate, and nothing supports rewarding a lower one. When resting heart rate or breathing rate is clearly up, good news from HRV counts for less.
- Reading it
- Below 50: below usual. 50 to 79: around usual. 80 and above: above usual. It is an index, not a percentage of recovery or a probability, and it cannot tell you the reason or whether training is safe.
- The first nights
- On nights 1 and 2 there is no baseline yet. You see a grey, preliminary number that compares the end of the night with its earlier parts; this method is unvalidated. From night 3 the full formula applies. The label says "Calibrating" up to night 7 and "Learning your usual" up to night 28, but the formula does not change at those points. The research review recommended no number before seven nights; we chose to show one from the first night, clearly marked, so the app is useful from the start.
- When there is no score
- Without a usable HRV or resting heart rate for the night, there is no Readiness. The resting heart rate needs at least three hours scored as sleep. The measurements that were taken are still shown.
- Skin temperature
- On a 5.0 band, the Readiness detail also shows a provisional skin temperature row, which does not count toward Readiness and has not been checked against a thermometer.
Why: the design follows a literature review of heart rate variability and resting heart rate. Research supports the direction of these signals, but no study gives an optimal weighting, baseline length or scale, and the benefit of training by heart rate variability is small and mixed. Every constant is a starting value. Since October 2026 both are measured over sleep only, as most overnight research does; the earlier version used the whole night, awake time included, and the lowest 30 minutes of heart rate. Days scored by earlier Readiness versions keep their score, but your baseline starts again from nights measured the new way, so Readiness calibrates again for a few nights.
Load
A score from 0 to 100 for how hard your heart and circulation worked over your day.
- The day it covers
- A Load day runs from the moment you fall asleep to the next time you fall asleep, so the night belongs to the day that follows it. Today's Load appears once last night has been confirmed, or once the band's history reaches noon without a night.
- Inputs
- Your heart rate for every measured minute, and three personal references:
- Sleeping heart rate: the lowest 30-minute average of each night (Readiness's resting heart rate averages the whole time asleep instead), as the median of your last 28 nights. Until there are earlier nights: your first settled night, otherwise the resting heart rate you entered, otherwise an average for your age and sex.
- Quiet level: your heart rate while awake and still. Once a day has 30 still, awake minutes, their median counts, and Levend uses the median of these daily values over the last 28 days. Before that, your sleeping heart rate plus 15.
- Maximum: the maximum heart rate you entered, otherwise 208 − 0.7 × your age. When the band measures a higher one-minute heart rate, that becomes your maximum.
- What moves it
- Each minute counts by how far your heart rate sits between your sleeping heart rate and your maximum, weighted so that hard minutes count far more than easy ones (the Banister weighting from training research). Only the part above your quiet level counts: a heart rate at or below it adds nothing. Light effort counts only partly: below 25% of the way from your sleeping heart rate to your maximum a minute adds nothing, and between 25% and 45% its share rises smoothly to all of it, without a step. Movement is not required, so effort with a still wrist counts, and so does a day that runs well above your normal. Every measured hour, sleep included, also adds a small amount: a night of eight hours gives about 12. A logarithmic scale then turns the total into 0 to 100, so the last points are the hardest to reach. Above 85 it bends further and never reaches 100: a marathon gives about 93, a whole day at a high heart rate 97 or 98, and a rounded 100 would take about five marathons in one day.
- Why the quiet level
- The earlier version counted every minute from your sleeping heart rate. Someone whose heart rate while sitting is well above their sleeping rate, because of medication, caffeine, stress, illness or simply their physiology, then built up Load all day without any exertion. The quiet level fixes that starting point. Because it is a median over many days, one unusual day does not move it, and such a day reads higher. A heart rate that is always raised, for example by daily medication, becomes part of your normal. Load cannot tell medication from stress.
- Why light effort counts less
- Load should tell a day of calm work at a low heart rate, such as standing and cooking, apart from training. Under the previous version a long day on your feet could score as high as an hour's run. Now such a day scores clearly lower, while training, runs and long hard days score as before. The 25% and 45% are Levend's choice, not a physiological boundary, and they follow your own sleeping heart rate and maximum. A side effect: an easy walk also counts a little less.
- When the formula changed
- Load's formula has changed three times: when the quiet level came in, when light effort started to count less, and when the top of the scale was bent so that 100 is not reached. Each time the last 30 days were worked out again once with the new formula, but only days whose measurements were still on the phone. Older days keep their score and say it comes from an earlier formula.
- Calibrating
- Under seven measured nights, or under three days behind your quiet level, Load says "Calibrating". It still shows a score.
- When there is no score
- Time the band did not measure adds nothing, and the measured share of the day is shown. Without any heart rate, or without the age and sex in your profile, there is no Load rather than a zero.
- What it does not see
- Load only sees heart effort. Strength training and other muscular work read lower than they feel.
Why: the Banister weighting comes from training research; the scale, the per-hour amount and the references are Levend's own product choices, checked with worked examples in tests. When this version arrived, the earlier version read so high on ordinary days that the two could not be shown side by side, so it replaced the recent days once, where nearly all their measurements were still on the phone. Days where that was not possible keep their old score and say it comes from an earlier formula. The light-effort change (October 2026) replaced the recent days in the same way, so a week does not mix the two.
Body Activation Experimental
A score from 0 to 100 for each minute of the last 24 hours, shown as Low (0–32), Medium (33–66) or High (67–100). It is an activation estimate, not a stress measurement.
- Two readings, the higher one counts
- Exertion: how much of your heart-rate reserve, the room between your resting and maximum heart rate, you are using. Medium starts at 20% of it, High at 40%, where moderate exercise begins in the standard intensity classes, and 100 is 90%.
Beyond movement: how far your heart rate sits above what is usual for you at that amount of wrist movement, learned from your own last 14 days. Half of your usual variation above it begins Medium, one and a half begins High, and three reaches 100. Each minute shows which of the two readings set it, so a raised heart rate without movement is shown as such. - While you sleep
- Your heart rate is compared with your usual sleeping heart rate, and the bar is higher: Medium from one and a half of your usual variation, High from three. During sleep, heart rate variability also counts for 40% of the second reading, and each minute is a three-minute average. The first and last 15 minutes of a night count as awake, because you often are.
- Why this design
- It follows the published idea of additional heart rate: heart rate above what your movement explains. Several versions were scored on two public research datasets of wrist measurements with labelled stress and rest. Heart rate compared with the person's own usual did best. Wrist heart rate variability did not help while awake, so it is not used then; during sleep it is used, untested, since the datasets contain no sleep. Single waking minutes separated stress from rest as well as or better than three-minute averages. Scored against each person's calm periods, the version these figures come from caught roughly a quarter to half of the stress minutes, while about one calm minute in ten also read High. The datasets used another wristband, young healthy adults and laboratory tasks, and contained no sleep, so the sleep thresholds are judgements.
- When there is no score
- Minutes without measurements stay gaps. The beyond-movement reading needs about three days of your own data first, and is left out during logged activities and while your heart rate settles after hard effort. Until then the label says "Calibrating".
Limits: it cannot tell excitement, tension, caffeine or heat apart, and a calm reading does not prove you were calm. It has not been checked against how people report their stress.
Calories burned Experimental
An estimate of the energy you used today, in Nutrition: resting, plus activity, plus an assumed amount for digestion.
- Resting
- From your age, sex, height and weight, with the Mifflin–St Jeor equation. Levend needs your profile for this.
- Activity
- For each awake minute the band measured, Levend takes how far your heart rate sits between your awake, still heart rate and your maximum, and turns that into energy with an estimate of your aerobic capacity from your maximum and sleeping heart rate. Activity counts only while your wrist is moving, or inside an activity you logged, because cycling and weights keep the wrist still. A raised heart rate while sitting still adds nothing.
- Why the still heart rate
- The still heart rate comes from the same day once it has 30 still, awake minutes, and from your recent days before that. An awake heart rate sits above the sleeping one without any extra work, and on some days far above it all day, with illness, caffeine, heat or stress. Taking the starting point from the same day cancels that shift.
- Digestion, assumed
- Digestion is shown as its own line, taken as a tenth of the total, as if you ate what you burned. Logging a meal does not change it. An earlier version based it on the meals you logged; then the burn fell when you logged lunch, which read as a mistake, and the extra detail was small next to the uncertainty of the estimate itself.
- Rest of the day
- For today, a separate line projects the whole day: what you burned so far, resting for the hours left, and the activity you usually burn from this time to midnight on your earlier, well-measured days. With too few such days it adds little or no activity, and it does not know what you plan to do.
- Eat per day
- The Nutrition goal's daily target is your usual day's burn plus or minus the weekly pace you chose, at 7,700 kcal per kilogram. Your usual day is the average of the days among the last 28 that the band measured for at least 85%. Once there are seven such days, the target is the same on every day, so one busy day does not raise it. It is a number you chose, not advice, and only as good as the burn estimate.
- Your own adjustment
- If the estimate runs consistently high or low for you, you can adjust it by up to 20% either way. Levend never sets this for you.
- When the band did not measure
- Asleep minutes and minutes without data count as resting only. The unmeasured time is shown. A past day is saved once and not recalculated when your profile changes later.
- Sources
- Resting energy: the Mifflin–St Jeor equation. Maximum heart rate from age: Tanaka and colleagues. Aerobic capacity from the ratio of maximum to resting heart rate: Uth and colleagues. Energy from where your heart rate sits between still and maximum (heart-rate reserve): Swain and Leutholtz. Combining heart rate with movement: the branched model of Brage and colleagues.
Limits: not checked against a laboratory measurement or any reference device. Strength training reads high, a wrong maximum heart rate shifts every active minute, and the movement limits are engineering choices, not calibrated values.
Heart rate
- Live and through the day
- While the app is open, your heart rate updates every second; the live value disappears when readings stop. The day's chart uses what the band stored, with gaps where it measured nothing, and shows how much of the time was measured.
- The nightly signals
- Readiness uses three measurements from your sleep: resting heart rate (the average over the time scored as sleep), heart rate variability (from the time between cleaned heartbeats, in the five-minute windows that are entirely sleep) and breathing rate (worked out from the rhythm of your heartbeat, not measured directly). Because the heartbeat samples breathing once per beat, a low heart rate while asleep limits how fast a breathing rate can be read: below about 48 beats a minute Levend reads only slower breathing, and shows no breathing rate rather than a guess when it may be faster. Load and heart-rate zones use the lowest 30-minute average of the night instead. These methods are adapted from OpenStrap; measuring over sleep only is Levend's own choice.
- Heart rate zones
- Activities show five effort zones. They follow your maximum heart rate: the one you entered, otherwise one estimated from your age; when the band measures a higher one-minute heart rate, that replaces either. Once Levend has 14 nights of your lowest 30-minute heart rate, the zones use your heart-rate reserve instead. You can also set the zone limits yourself.
A loose band or a lot of wrist movement can give a wrong or missing reading.
Steps
- Where they come from
- Your band's motion chip keeps its own running step count, and the band stores it with every second of its history. Levend adds up how much that count rose, minute by minute. Nothing is guessed for the time the band recorded nothing: those hours show faded, and a day without band records has no count.
- Which minutes count
- A wrist also moves without walking, in bed, in the shower or while cooking, so a minute counts only when it looks like steps: 30 or more on a 5.0 band, 70 or more on a 4.0 band, whose chip also counts more arm movement, and either your arm hung down for a good part of it (40% of the stepping seconds), as it does while walking, or it was part of at least three minutes of steady stepping, as when you walk with your phone or a bag in your hand, or it runs straight on from such a minute, as the start and end of a walk do. Pedalling on a bike can look like steady stepping, so minutes where the count jumps quickly while your wrist stays still are not taken as steady stepping and do not join a walk. A moment of stepping in bed stands on its own, so it does not count. A short walk on its own with something in your hand and very slow walks can therefore count fewer steps. On an MG band, steps are experimental: its counter has not been checked.
- Your phone's steps
- Levend shows the band's steps only and writes nothing to Apple Health, so your phone's own count, which misses walks without it, can differ.
Steps are not a score and do not change Sleep, Readiness or Load.
With an Apple Watch
- Where the data comes from
- Instead of a band, you can let Levend read Apple Health. It reads what your Apple Watch saved: heart rate, the heartbeat timing it records with its heart rate variability readings, sleep and its stages, steps, energy and workouts. Levend only reads, and what it reads and works out stays on your iPhone.
- Sleep
- Levend uses Apple's own sleep stages and counts them the way it counts a band's night, with the same Sleep score. Apple decides when you fell asleep and woke up, so the band's rules for the start and end of a night do not apply. Naps are Apple's other sleeps of the day.
- Readiness
- The same formula as with a band, from your heart rate variability (from the Watch's heartbeat recordings while you were asleep), sleeping heart rate and breathing rate, against your own Watch nights. The Watch records far fewer heartbeats than a band, so on your first nights Readiness is a preliminary score from heart rate.
- Load
- The same formula as with a band. Outside a workout started on the Watch, it saves a heart rate about every five minutes; each reading counts until the next one, for at most 15 minutes. That makes Load an estimate: start workouts on the Watch to give them their full heart rate.
- Calories burned
- The same model as with a band. Apple Health holds no wrist movement, so whether you are moving comes from your steps and from Apple's own active energy, whichever shows more. Apple's own estimate is shown next to Levend's; the two are worked out differently and can differ.
- Activities and steps
- Workouts you start on the Watch arrive named by Apple. Other stretches of raised heart rate are suggested for you to name. Steps are Apple's own count.
- Not available
- Body Activation needs heart rate every second, and the band's alarm needs a band.
Scores from an Apple Watch are worked out from fewer measurements than a band's, are kept apart from them and are not validated.
Habit patterns
In Behavior you can answer Yes or No to your habits each day. Levend then looks at whether your Sleep score, your Readiness or how rested you said you felt the next morning tends to differ after Yes days compared with No days, over up to 180 days, allowing for the previous day's score, weekends and slow trends. It needs at least 15 usable days, with at least four Yes and four No, and statistical checks before it shows a result. Until then it says it needs more days, and the habit's page shows which of the three is still short. With enough days the result can still be unclear, and it says so; a difference that shows but is not yet certain is called an early sign. It shows what goes together, not what causes what, and it is never a reason to change a medicine or treatment. Expecting a habit to matter can change how rested you say you feel, so answering before you see your scores helps.
Check-ins
Check-ins are your own answers, not measurements, and Levend does not score them. In the morning, one question asks how rested you feel, from 1 to 10 (five steps before November 2026; those answers stay as given), shown in Readiness's colours from red to green; it comes before your scores, because seeing a low score first tends to make people feel worse, and you can turn that order off in Settings. In the evening, four questions ask about your focus, stress, energy and satisfaction, each in five steps, and any of them can be left open; a missed evening can still be answered the next morning. Neither is asked on the day you first wear a band, only from the morning after its first night. Your answers can differ from your scores, and that difference is worth seeing.
Learning your normal
Readiness, the timing part of Sleep, Load and the heart rate zones compare you with your own usual, so they first need to learn it. Labels such as "Calibrating" show how far that is. Wearing the band day and night makes it go faster; a day with large gaps does not count.
When you pair a different band, Readiness, the timing part of Sleep and the heart rate zones carry on from the nights you wore your earlier band, so they work from the first night on the new one. Each new night replaces the oldest one, until the baseline is entirely the new band's.
On a new iPhone, or after reinstalling the app, Readiness, the timing part of Sleep, Load and the heart rate zones carry on from the nights restored from your Levend account in the same way.
When a formula changes
Every score keeps the version of the formula that made it, and formula changes normally get a new version number. Finished days keep their score, so a jump caused by a new formula is not presented as a change in you. The exceptions so far are Load's, described under Load: each time, recent days were recalculated once, only where their measurements were still on the phone, and the other days say they come from an earlier formula. A few figures are worked out each time they are shown rather than saved, such as the assumed digestion in calories burned and today's projection, so a change to them also shows on past days.
Questions about how a score works? Email support@shuen.app.