The science and the logic behind the suite — how each score is computed, why it is built the way it is, and the one thing it sees that no wearable on the market can: your nutrition and body composition.
Your sleep, training and recovery — understood. On hardware you already own. With every score explained in a single tap.
MIntelligence is Metapace's wellness intelligence suite — three features that turn an Apple Health connection and the nutrition data Metapace already holds into coaching-grade insight. It is built on a single premise:
Nutrition, body composition, sleep, training and recovery — all in one place, on a device you already own.
No proprietary band. No subscription ring. No extra hardware to buy. Because the suite reads from Apple Health, it works with any watch or band that writes to it — and every score it produces is transparent, computed privately on your device, and explained in one tap.
The M-family has two pillars: MIntelligence — the umbrella brand for the feature suite (MInsights + MPace + MRecover), what the app can do for you; and MTribe — Pro members, the membership and identity layer. Going Pro = joining the MTribe.
MInsights — know your sleep
Nightly stages, hypnogram, overnight vitals (HRV, respiratory rate, wrist temperature, SpO₂), trends and a transparent, physiologically-grounded sleep score. The score uses 5 weighted components — duration, stage quality, efficiency, HRV signal, and consistency — and publishes exactly how every point is earned.
MPace — train smarter
Training load (Acute:Chronic ratio), true time-in-zone from per-second heart-rate data, aerobic efficiency trending, personal bests, weekly zone mix with coaching pattern reads — and body-composition-aware training that frames your effort against your weight and energy balance. No wearable knows your body composition.
MRecover — your daily readiness
Six signals — sleep quality, HRV status, training load, nutrition & energy, resting heart rate, and consistency — fused into one 0–100 readiness score that answers: how ready are you today? Two of those six signals are unique to Metapace.
Every recovery and sleep product on the market today shares the same limitations. MIntelligence is built specifically to close them.
The dedicated recovery players — the subscription wearables at $30/month, the premium smart rings at $300+ upfront — lock their best science behind a proprietary band or ring. A recurring membership, a device to charge, one more thing to buy and wear. MIntelligence runs on a watch or band you already own.
Why this matters: the hardware lock-in is a business model, not a scientific requirement. The sensors on an Apple Watch are equivalent or superior.
A single number appears each morning and you are asked to trust it. Apple's own watchOS 26 Sleep Score is a behavioural/timing measure — roughly half its weight is how long you slept, the rest bedtime consistency and interruptions. It does not use HRV, resting heart rate, respiratory rate, or sleep-stage quality. Garmin's Training Readiness is notoriously opaque — users can't tell why the score is what it is. MIntelligence shows its work.
Why this matters: a number you can't interrogate is a verdict. A number you can explain is a conversation.
Physiology without nutrition is half the picture. A deep caloric deficit measurably impairs recovery and suppresses deep sleep — yet no wearable has any idea you are in one. The nutrition apps know your food but have no recovery science. Metapace holds both sides because nutrition and body composition are its home turf.
Why this matters: a sustained −500 kcal/day deficit is the single most actionable lever on recovery that pure sensor data cannot detect.
These come from what Metapace already knows and how it is built. A competitor cannot match them without rebuilding into a full nutrition and body-composition platform first.
Metapace already knows your energy expenditure (Mifflin-St Jeor or Katch-McArdle when body fat is provided), your deficit or surplus trend, your body-fat percentage, your fasting windows and your weight trajectory. Recovery can be read against whether you are actually fuelled — a signal pure wearables simply do not have.
Why we built it this way: a recovery score without nutrition context is like a weather forecast without wind — technically correct but practically incomplete.
Someone at 35% body fat in a 500-calorie deficit recovers very differently from someone at 15% at maintenance. Metapace can reason about that difference because it holds both sides of it. MPace frames training energy as kcal per kg of body weight — relative output, not an absolute number — so progress reads correctly regardless of your size.
Why we built it this way: absolute calories burned means nothing without context. A 60kg runner and a 100kg lifter burning "400 kcal" are doing very different things.
The competing scores are black boxes. Metapace shows exactly what went into every score, with a plain-English, one-tap explanation for each contributor. MInsights publishes the exact 5-component formula (25 pts duration, 25 pts stage quality, 20 pts efficiency, 20 pts HRV, 10 pts consistency). MRecover shows all 6 contributors with their weights. This turns the score from a verdict into a conversation.
Why we built it this way: the single biggest UX difference is trust. An explained score you believe is more useful than a precise score you don't.
Everything runs on a device you already own. Because the suite reads from Apple Health, it works with a wide range of watches and bands — not a single proprietary one — and the honest empty states make clear when richer sensors would add more. The known bug in the closest App Store competitor (failing to correctly merge split sleep sessions) is exactly solved here.
Why we built it this way: adding hardware is a tax on the user. The sensors already on their wrist are more than sufficient when the math is right.
Every score in the suite rests on a deliberate engineering choice. Here is the reasoning behind each one, in the open.
When you wake at 3 a.m. for an hour and fall back asleep, the watch writes two or more separate sleep records. Naïve code keeps only the largest block and silently throws away the rest — corrupting total sleep time and every score built on it. This is the exact failure that breaks the closest app-only alternatives.
MInsights runs a 90-minute gap-merge: sessions separated by less than 90 minutes are treated as one continuous night; a true daytime nap, separated by more, is kept distinct and never pollutes the nightly score. The threshold is chosen to span an ordinary sleep-cycle gap or a bathroom break without ever fusing a nap into the night.
Total sleep time sums only the actual asleep stages — core, deep and REM — never time merely in bed. If a watch only reports undifferentiated sleep, stage metrics are marked unavailable rather than invented. Sleep efficiency is calculated as time asleep ÷ time in bed, and WASO (wake after sleep onset) sums only awake samples within the sleep window.
The merge, step by step
Why 90 minutes? It spans one full sleep cycle — the longest gap a normal bathroom break or settling period could produce without being a deliberate wake-up.
How HRV is shown
Last night vs your 14-day baseline
+8%above average
Shown as deviation from your own history — never a raw millisecond figure presented as a universal "good" or "bad".
The 14-night baseline uses the same recent-14-night enriched series that feeds the Last Night card and Vitals Trend — not a separate HealthKit query — so the explainer's "% vs your 14-night average" always agrees with the charts.
The watch exposes SDNN — the standard deviation of beat-to-beat intervals — not the RMSSD that some dedicated devices use internally. These are numerically different values. Converting between them without the raw beat-to-beat interval series (which Apple does not expose) would be a fabrication.
So MIntelligence never invents a conversion and never shows a bare SDNN value dressed up as "your HRV." It builds a baseline from your own history and reports deviation — "8% above your 14-day average" — with the methodology explained. A signal you can trust beats a number that looks precise but isn't.
The same deviation-from-your-own-baseline approach protects against age, fitness, and device differences that make absolute HRV numbers meaningless to compare across people.
Every "vs baseline" comparison weighs a 7-day exponentially-weighted average against a 28-day mean. Only a drop beyond roughly one standard deviation reads as a genuine negative signal. This validated, two-window method (inspired by the approach used by the leading subscription wearable) is far more robust than naïve day-over-day comparison — your readiness reflects a real trend, not yesterday's noise.
Why two windows? A single rolling average over-smooths real changes. Two windows — one fast (7d), one slow (28d) — capture acute shifts while anchoring against your long-term norm. This is standard sports-science methodology adapted for consumer wellness.
MPace measures a 7-day acute load against a 28-day chronic baseline, weighting each session by heart-rate zone (a Banister-TRIMP-inspired approach). Where per-second heart-rate data is available, zones are computed from true time-in-zone — real per-second distribution, not a single average. When it's estimated, the card says so.
The ratio is withheld until 28 days of history exist — never fabricated early. This is one of the most evidence-based training-stress metrics in sports science and virtually no consumer app visualises it clearly.
The wellness domains are independent — you might have workouts but no tracked sleep, or sleep but no logged food. When an input is missing, MRecover does not punish you for it. It drops that contributor and re-normalises the rest to still sum to 100%.
Scoring a missing signal as 0 would falsely crater readiness; scoring it as 100 would falsely inflate it. Both are dishonest. Instead, the score reflects only what is actually known — and a banner names what is missing and how to restore it.
There is a hard floor: if too few contributors remain to produce a meaningful score (for example, only the 5% consistency signal is available), the score is withheld entirely and the building state explains why — rather than presenting a hollow, confident-looking number.
The re-weighting rule
score = Σ(presentᵢ · weightᵢ)
Σ(present weightᵢ)
If too few contributors remain to be meaningful, the score is withheld entirely and the building state explains why — rather than presenting a hollow, confident-looking number.
One exception: Training Load defaults to neutral 100/100 when absent (no workouts = no fatigue, not a penalty). Every other absent contributor is dropped.
A readiness score appears only after at least 7 nights of sleep and 14 nights of HRV exist; a sleep score waits for at least 3 nights; the Acute:Chronic training ratio requires 28 days; wrist-temperature deviation needs 5+ baseline nights. Below those thresholds, the app says "building your baseline" with honest progress — it does not show an impressive number it cannot stand behind. Trust is the product.
Why gate at all? Early scores with thin baselines look precise but swing wildly night to night. The first impression of a health score sets the user's trust for months — a shaky first number is worse than no number.
Metapace derives your energy expenditure from established equations (Mifflin-St Jeor, or Katch-McArdle when body fat is known) and tracks your real intake and weight trend. So MRecover can say what no wearable can: that a sustained deficit is likely suppressing your recovery, and your body needs fuel to repair.
An extended aggressive deficit (> −700 kcal/day) hard-caps the nutrition contributor at 40/100 regardless of other signals — a health-relevant guardrail that suggests a diet break. An active fasting window applies a small additional penalty. This is guidance a sensor alone cannot give.
Why cap at 40? Because published research shows deficits above ~700 kcal/day measurably impair muscle protein synthesis and recovery. Capping the score prevents the readiness number from reading "fine" when the user's nutrition strategy is actively undermining their recovery.
Before noon, the freshest completed sleep night is still yesterday's. Open the app at 1:45am and it should not flag "No sleep tracked last night" — the night is still happening. The code respects when a signal was actually filed: sleep filed under the morning it ended, workouts bucketed by their start local time, and a noon cutoff that prevents premature missing-night banners.
Why this matters: a false "no sleep tracked" alarm at 2am destroys trust. The time boundary logic is invisible when it works correctly — and catastrophic when it doesn't.
If a night wasn't tracked (watch uncharged, not worn to bed), that day appears as a blank gap on the chart — the line connects across it. It is never rendered as a zero bar, never interpolated, and never silently borrowed from the prior week. "X of Y nights tracked" on the weekly summary makes the coverage honest.
Why not zero-fill? A zero bar reads as "something terrible happened." A gap reads as "we don't know." The difference is the difference between alarming the user and being honest with them.
Honest time boundaries, gaps left blank instead of zero-filled, trustworthy units filtered first, self-titled with the real date when the shown data isn't the expected latest, and any value withheld until its baseline pool is deep enough. The biggest risk in health features is an impressive number that is quietly incorrect. Every metric here degrades to a clear, explained empty state instead.
The built-in score that ships on your watch is a timing-and-duration measure — roughly half its weight is how long you slept, the rest is bedtime consistency and interruptions. It does not factor in your heart-rate variability, resting heart rate, respiratory rate, or how much deep and REM sleep you actually got, and it is not exposed for apps to read. MInsights computes its own, physiologically-grounded score from raw stage data — and publishes exactly how.
100
total points
Sleep duration
Actual sleep vs an age-based target (7–9h for adults, per NSF guidelines)
Stage quality
Deep + REM as a share of total sleep (target: ~20–25% each)
Sleep efficiency
Time asleep vs time in bed — a tight ratio means less restless wakefulness
HRV signal
Overnight SDNN vs your personal 14-night baseline (deviation, not raw)
Consistency
Regularity of your sleep timing (std dev of onset over the 14-night window)
And it degrades honestly: if an older watch can't report stages, those 25 points redistribute to duration and efficiency; if HRV is missing, its 20 move too; below three nights, you see "building your baseline" instead of a shaky number. Tap any score in the app to see this exact breakdown for your night.
MRecover fuses six signals into one 0–100 readiness score. Four of them the dedicated devices also use — but the nutrition and consistency reads are Metapace's alone. MRecover is a read-only consumer — it owns no HealthKit data of its own. Its biometric inputs arrive through the already-authorised sleep and workout scopes; its nutrition inputs come from the user's own logged data.
Sleep quality
30%Last night's MInsights score — the entire physiological sleep assessment as one input. Because sleep is the foundation of recovery, it carries the most weight.
Missing → dropped & re-weighted. "Wear your Apple Watch to bed for a full score."
HRV status
25%SDNN deviation from your dual-window baseline (7d EWMA vs 28d mean). The strongest single biomarker of autonomic recovery — a meaningful drop means your nervous system hasn't fully restored.
Requires ≥14 nights of HRV to baseline. Missing → dropped & re-weighted.
Training load
20%Acute:Chronic ratio proximity to optimal (0.8–1.3 = 100, degrades symmetrically). Are you training harder than your body has adapted to?
Missing → defaults to neutral 100 (no load = no fatigue, not a penalty).
Nutrition & energy
15%Caloric-balance trend vs your expenditure. Deep deficits (> −700 kcal/day) hard-cap this at 40/100 and suggest a diet break. An active fasting window applies a small additional penalty.
Resting heart rate
5%Sleeping RHR vs your baseline. A low weight because RHR is a confirming signal — it rarely tells you something HRV hasn't already said, but it validates the story.
Missing → dropped & re-weighted.
Consistency
5%Sleep timing regularity + training frequency. Social jet lag (wildly varying bedtimes) is one of the most underrated recovery suppressors — MInsights measures it and MRecover penalises it.
Each contributor is rated 0–100, weighted, and — crucially — re-normalised whenever an input is missing, so the score always reflects only what is genuinely known. Tap any contributor in the app for a plain-English what, why and what-to-do.
MPace doesn't just show workouts — it produces coaching-grade intelligence from them. Here are the three capabilities no wearable can match, and why each one matters.
Metres per heartbeat — distance / (avgHR × minutes) — trended for your dominant cardio kind. Rising means "same pace, lower heart rate" — the most motivating proof of fitness improvement, computed from data already collected.
Withheld until ≥3 eligible same-kind sessions. Manual entries are excluded (PB anti-cheat doctrine). The delta compares the recent third vs the earlier third so one outlier can't swing it.
Why we built it: "fitter, not just busier" — the one metric that proves your training is working, not just happening.
Weight, BMI, weekly active energy, and relative output (kcal/kg) — plus a cross-domain insight tying your weight trend to training: "down ~0.4 kg/week — the same training is relatively harder per kilo; keep protein up."
No logged weight → card absent. No height → BMI omitted. No active energy → kcal/kg omitted. Single weight entry → neutral "weight steady" copy, never a fabricated trend.
Why we built it: no wearable knows your body composition. A 400 kcal burn means very different things at 60 kg vs 100 kg. Relative output normalises the comparison.
The 12 most recent workouts are enriched with the real per-second heart-rate series — so HR zones and the TRIMP load come from actual time distribution, not a mean-HR estimate. The weekly zone mix aggregates into easy / moderate / hard with a coaching pattern read (Polarized / Pyramidal / Gray-zone / Balanced).
When per-second data isn't available, the estimate is used — and the card discloses "approximate." Older workouts degrade gracefully.
Why we built it: "12 min in Z4, 3 min in Z5" is actionable. "Average HR 148" is not. The difference is the difference between coaching and counting.
When recovery dips, the suite can point to a multi-day deficit as the likely cause — a read no pure wearable can make. Extended aggressive deficits hard-cap the nutrition contributor and suggest a diet break.
Every score is explainable in one tap. The number is never a verdict — it's the start of a conversation. The full component weights are published, not hidden.
The 90-minute gap-merge correctly stitches a mid-night wake-up back into one night — the edge case the closest app-only rivals get wrong. Naps are separated, never polluting the nightly score.
An automatic Monday review across sleep, training and nutrition — best and worst recovery day, average vs prior week, and a correlation insight — a health journal nobody had to write.
When data is thin, you get an honest empty state — never an impressive number that's quietly wrong. Baseline gates, missing-contributor re-weighting, and worn-night filtering all enforce this.
Every score is computed on your phone and cached locally — file-protected, never on a server. The dedicated devices are cloud services that stream your physiology to their servers by design; here, the raw data stays on your device, and only pre-computed values ever leave it, for an optional narrative layer alone. Privacy isn't a setting — it's the architecture.
Respiratory rate, sleeping wrist-temperature deviation (with a 5-night baseline gate), blood O₂ average and lowest dip, plus a full-night hypnogram — each framed against your own baseline, never absolute "good" or "bad." Missing signals degrade to "not measured", never fabricated.
Every card carries a contextual disclaimer. When respiratory rate or wrist temperature runs above your baseline, the language is calm and personal — "often an early sign your body is fighting something" — never "you may be sick." No alarming colours, no push notifications for vitals. Health guidance without health anxiety, because the goal is awareness, not fear.
Compared by product category, not by name. ✓ present & strong · ~ partial · ✗ absent.
| Capability | MIntelligence | Built-in watch score |
Subscription wearable |
Premium smart ring |
Watch readiness |
Other watch apps |
|---|
Capabilities reflect each category's typical offering as of June 2026. "No extra hardware" credits products that run on a device you already own.
The dedicated devices have physiology but no idea what you eat. The watch-native scores have training science but won't explain themselves. The built-in score is free but timing-only and locked away from apps. The nutrition apps have your food but no recovery science. MIntelligence is the one product that brings them together — on hardware you already own, with every score explained in a single tap.
Sleep, training, recovery, nutrition, body composition, and family — understood. Computed on your device. Explained in the open. Always.