58 articles · every figure traceable

The numbers, explained.

Health apps hand you numbers with a lot of confidence and very little explanation. These pages say what each one actually measures, how wrong it can be — and what it can never tell you.

The scale & body

0.06 kgof an overnight kilo is actually fat You didn’t gain a kilo overnight The scale can jump a kilo between Tuesday and Wednesday. Almost none of that is fat — it's water held by sodium, freshly stored glycogen, and food still in transit. Here's the arithmetic, and how to read the number that actually matters. +0.7 kgtypical Monday-to-Friday swing Your weight has a weekly rhythm Daily weighers are typically heaviest on Monday morning and lightest by Friday. The wave is mostly water, glycogen and gut contents, not fat. Here is what drives it, and how to read a scale that oscillates on a seven-day cycle. +18%more space per kilo of fat than muscle What 1 kg of fat looks like next to 1 kg of muscle Fat and muscle weigh the same per kilo — that is what a kilo means. They differ in density: fat fills about 1.1 litres per kilogram, muscle about 0.94. That gap is why body weight and body shape can move independently. 0.45 kgfat a 500 kcal daily deficit removes in a week The first week of any diet is water. The second is the truth A 500 kcal daily deficit removes about 0.45 kg of fat in a week. Most first weeks show three to five times that. The difference is glycogen and water, and it tells you nothing about whether the diet works. 0.5keep your waist under half your height Waist beats weight: the one tape measurement worth tracking The scale reports total mass — water, glycogen, gut contents and fat, all in one number. A tape around your waist tracks the fat most tied to health risk. Here is what each can tell you, and how to measure so it means something. ±1–2 ptseven DEXA misses by this much How accurate is a photo body scan, honestly A photo scan estimates body fat to within roughly 2.5–5 percentage points. That is worse than DEXA, similar to a smart scale, and enough for exactly one job: tracking change under conditions you keep fixed. Here is where the error comes from. ~50%of high-body-fat adults missed by BMI “Skinny fat” is two numbers, not one Two people can weigh the same and carry very different bodies. “Skinny fat” is what happens when weight stays normal while body fat climbs. One number cannot see it. Two can. 3 gof water stored with every gram of glycogen Why you look worse mid-cut before you look better Mid-cut, glycogen empties, water shifts under the skin, and muscles flatten while fat loss carries on underneath. The mirror reports all of it at once. Here is the timeline, the mechanism, and the numbers that see through it.

Food & logging

±20%the legal margin on stated calories A calorie label is allowed to be 20% wrong Nutrition labels are estimates with a legal margin — roughly 20% in the US and EU — and almost nobody checks. Here is where the error comes from, how big it gets in a real day, and how to log food anyway. 440 kcalone large blended coffee order The 400-calorie coffee you didn’t order Coffee itself is roughly two calories. The milk, syrup, cream and blended base stacked on top can take one order past 400 — more than many meals — and drinks are the entries people forget to log first. 1 in 5tested dishes ran 100+ kcal over the menu Restaurant calories vs what’s on the plate Lab measurements of restaurant food find the average dish close to its stated calories — and one in five at least 100 kcal over. Here is what a menu number can tell you, what it cannot, and how to log around the gap. 25%higher 24-hour synthesis from an even split Protein at breakfast is the meal most people lose A typical food log holds around 90 g of protein, skewed hard toward dinner — often four times what breakfast carries. Research on distribution suggests the skew costs you, and breakfast is the cheapest place to fix it. 3.3×more lentils than chicken for the same 30 g What 30 g of protein looks like on a real plate Thirty grams of protein is a common per-meal target. Here is what it actually weighs in chicken, salmon, tofu, eggs, yogurt and lentils — and the three things the gram count doesn't tell you: calories, cooked weight and label maths. <3%of one food app's downloaders became active users Logging two days a week still works Daily food logging fails for most people within weeks. Sampling two days a week — one weekday, one weekend day — captures most of what logging is for, at a fraction of the effort. Here is what it can and cannot tell you. 66%of the planned weekly deficit, gone by Sunday The weekend calorie gap: why Mon–Thu perfection stalls You hit the target Monday to Thursday, then stop counting. In a worked week, three unplanned weekend days erase 66% of the deficit — the maths, the reasons logging goes blank on Saturdays, and a budget shaped like your actual life. ~500 kcalwhat a day of unmeasured pours can add Oil, the invisible 500 — the cooking fat nobody logs One tablespoon of oil is about 119 kcal. A free pour is often two or three. Across a day of cooking, the fat that never reaches the diary can erase a 500-kcal deficit while the log looks perfect. −47%of intake unreported in the classic study “I barely eat” — under-reporting is nutrition’s best-replicated finding Measured against doubly labelled water, self-reported food intake falls short by about 20% on average — and by nearly half in some dieters. This is measurement error, not character. Here is what the gap looks like, and how to log around it.

Training

~75%of a five-set stimulus, from two hard sets Two hard sets do most of what five do Muscle growth rises with every set you add, but each one buys less than the last. Two honest sets near failure capture most of what five deliver. Here is the dose-response data — and what it cannot tell you. ~5 repsthe window that does most of the work The last rep before failure is worth more than the first Ten reps are not ten equal reps. Muscle fibres are recruited in order, and the reps that recruit the biggest ones are the slow, ugly ones near the end. Here is why effort — not rep count — decides what a set is worth. 24–48 hhow long repair runs after a session How long a muscle actually needs between sessions Muscle protein synthesis runs for roughly 24–48 hours after a hard session, longer after heavy eccentric work. That gives each muscle group a recovery window you can plan around — and a few things that quietly stretch it. ~0 kgof new muscle behind week-2 PRs Strength before size: what the first 8 weeks really build New lifters get stronger long before they get bigger. The first eight weeks are mostly your nervous system learning to use the muscle you already have. Here is the timeline, why it looks like nothing is happening, and what to track instead. −5.4%strength after pre-lift static stretching The warm-up that takes 4 minutes and isn’t stretching Long static holds before lifting reduce strength by around 5% and rehearse nothing. Ramp sets — the empty bar, then three rising sets with falling reps — raise temperature, groove the movement, and take about four minutes. Here is the exact protocol. ≈0muscle-growth gap in head-to-head trials Machines vs free weights: what actually differs in the data Head-to-head trials keep finding the same thing: machines and free weights grow muscle about equally. The real differences sit elsewhere — in how strength transfers to the test, how fast you learn the movement, and almost nowhere in the injury data. 14 kgbest-to-worst spread on a 140 kg squat, one week Why your lifts wobble 10% day to day The same bar feels light on Tuesday and welded to the floor on Friday. That is not lost muscle. Daily strength output swings around 10% for reasons that have little to do with training — and the trend underneath still decides. 28 weeksto double your squat at +2.5% weekly Progressive overload has 5 dials, not 1 Adding weight is one way to make training harder — not the only one. Load, reps, sets, tempo and range each move the stimulus, in very different step sizes. What each dial does, when it sticks, and the order worth turning them in. 4–7%VO2max lost in a fortnight of full rest What detraining actually costs: two weeks off, by the numbers Two weeks of complete rest costs roughly 4–7% of VO2max, almost no strength, and less muscle than the mirror suggests. The losses are real but shallow, and the regain runs faster than the original gain. Here are the numbers.

Recovery & sleep

14signals a fully-measured morning score draws on Your recovery score, decoded signal by signal Heala's recovery score compresses sleep, HRV, resting heart rate, yesterday's strain and a handful of lifestyle signals into one morning number. Here is how each nudges the score, why your baseline beats anyone else's, and what the number cannot see. 55 msone reading, two opposite meanings HRV only means something as YOUR trend Two people wake up to the same HRV: 55 ms. For one it is a red flag, for the other a personal high. HRV carries almost no meaning as a raw number — only as a deviation from your own baseline. 1–2all-nighters hidden in a fortnight of 6-hour nights One bad night vs a bad week of sleep A single short night is recoverable and you can feel it. A week of short nights is different: the deficit compounds, your sense of sleepiness plateaus, and by day fourteen you can perform like someone who has not slept at all. +8 bpmtypical overnight heart-rate gap after late drinks Alcohol’s exact fingerprint on your sleep graph Two drinks with dinner don’t just make you sleepy. They raise your overnight heart rate, hold your HRV down, and push REM into the back half of the night. The pattern is consistent enough to spot on a graph — and to test on yourself. 2–4 daysmetric drift before the first symptom The cold that showed up in your metrics before your throat By the time your throat hurts, your immune system has been at work for days — and the work shows up in resting heart rate, HRV and overnight temperature. Here is what the early drift looks like, and what it cannot prove. 3–5 bpmtypical drop after three months of cardio Resting heart rate is the cheapest fitness test you own Your resting heart rate drops as aerobic fitness improves — steeply in the first three months, then slowly for years. It costs nothing to measure. Here is how to read your own trend, and what a single morning's number cannot tell you. 66 mgstill circulating at a 23:00 bedtime Caffeine’s half-life, drawn on your actual evening A caffeine half-life of about five hours means a 15:00 coffee is still a third present at a 23:00 bedtime. Here is the decay curve drawn hour by hour, what it changes about your sleep, and where the average stops applying to you. +54%alertness after a 26-minute nap Naps: what 20 minutes buys and 90 minutes costs Nap length decides which sleep stage you wake from. Twenty minutes buys measurable alertness with a clean wake. Forty-five wakes you into grogginess. Ninety wakes clean but spends the pressure you need to fall asleep tonight. Monthsof decline before overtraining syndrome applies Overtraining is rare. Under-recovering is everywhere Genuine overtraining syndrome takes months of declining performance to earn the name. What most people call overtraining is a normal training load landing on too little sleep, too much stress, or too little food. The fix usually sits on the recovery side.

Cardio, steps & strain

~80%of elite endurance training is easy Zone 2 feels too easy — that’s the point Zone 2 is the pace that feels like you are not trying. The evidence says that unglamorous pace builds most of your aerobic engine — if you can resist the urge to speed up. Here is how to find it and hold it. 7,000steps a day, where the curve flattens 10,000 steps was a marketing number. The curve bends at 7,000 The 10,000-step target was coined to sell a Japanese pedometer in 1965. Modern cohort studies put the bend in the benefit curve closer to 7,000 steps a day — and show the first few thousand do most of the work. 3×load spread across equal-calorie workouts What “strain” measures that calorie burn doesn’t Two workouts can burn the same 400 kcal and cost your body very different amounts. Calorie burn estimates fuel; cardio load estimates the dose. Here is what each number can and cannot tell you. 80%of elite endurance sessions are easy Running slower to race faster: the 80/20 split Across running, cycling, rowing and skiing, elite athletes spend roughly four in five sessions at easy intensity. The split is well replicated, slightly counterintuitive, and mostly ignored by recreational runners — who tend to run everything at a pace that is neither. 13%lower mortality per extra MET of fitness VO₂max: the number your watch guesses, and what it means Your watch has never measured your VO₂max. It infers it from how fast you move at a given heart rate — typically off by 5–15% versus a lab test. What the number means, where you sit by age, and what moves it. ≈10 METswhat a sub-45-second climb suggested The stairs test: 60 seconds that predicts a lot A stopwatch and four flights of stairs give a rough read on aerobic capacity. Under 60 seconds is a good sign; over 90 is a prompt to investigate. What the test measures, what it misses, and how to retest it properly. −22%evening post-meal glucose rise, type 2 diabetes trial Walking after meals moves glucose more than walking before Ten minutes of easy walking lowers the post-meal glucose rise — but only if the walk lands on the peak. Trials that timed the same dose of walking differently found the after-meal version does more. Here is the size, the shape, and the schedule.

Consistency

2consecutive misses is where habits die The 2-day rule: streaks break in pairs In the best habit-formation data we have, a single missed day barely registered. The second consecutive miss is where streaks tend to die. The 2-day rule defends that line: never miss twice, and keep the day after a miss deliberately small. 66 daysaverage time for a new action to feel automatic Motivation is a lagging indicator — action comes first The common model says motivation produces action. Trials of behavioural activation and habit-formation studies point the other way: action comes first, and the feeling follows it. Here is the evidence, and a way to start without waiting for anything. 2.6%of 189,770 app downloaders became active users Why “all or nothing” always lands on nothing All-or-nothing plans fail on schedule: at the first missed day. The adherence data favours the unglamorous 80% middle — and the fix is structural, not motivational. Build a plan that survives a bad Tuesday. 14 daysof flat trend before a stall counts The plateau is data, not punishment A flat scale is information: your intake and expenditure currently match. Before you cut harder, decide whether the stall is real, audit the log, and change one variable at a time. A four-step ladder for reading a plateau. +13%what 1% a week compounds to What 12 weeks of “average” actually compounds to Twelve unremarkable weeks — most sessions done, most meals logged — add up to more than one heroic fortnight. Here is the honest arithmetic of 1% a week: what compounds, what stays linear, and why the curve barely bends.

Cycles & seasons

Trivialaverage performance effect of cycle phase Training with your cycle: what the evidence supports Phase-based training plans promise a lot. The largest meta-analysis found trivial average performance differences across the cycle, on low-quality evidence. Here is what holds up, what is oversold, and how to use your own data instead of a textbook calendar. +0.5–2 kgof water, days before day one Your period moves the scale before it arrives In the week before a period, hormones shift how much water your body holds. The scale can rise 0.5–2 kg with no fat gained. Here is what moves, when it peaks, and how to read your weight trend through it. 0.37 kgthe average holiday gain, once measured The seasonal weight wave almost everyone rides Average body weight follows an annual curve: a winter peak around the holidays, a slow spring fall, a late-summer low. The wave is smaller than people believe, stickier than they expect, and easy to misread on a single weigh-in.

Heala vs the alternatives

2022the year barcode scanning went Premium Heala vs MyFitnessPal: logging a meal in 2026 MyFitnessPal has the deepest food database in the business. Heala starts from a photo of the plate instead. Both end at the same calorie number; the difference is every tap in between. 0extra hardware for a Heala recovery score Heala vs WHOOP: recovery without the strap WHOOP pairs excellent dedicated hardware with a membership and measures you around the clock. Heala reads the same overnight signals from the watch or phone you already own, and puts the score next to food, training and body. 1 photowhere both apps start — one of them stops there Heala vs Cal AI: what happens after the photo Cal AI turns a photo into a calorie estimate fast, and it does that job well. Heala takes the same photo and connects it to targets, training, recovery and a coach. This is about what the number is for. +3domains Heala runs beyond the diet: training, recovery, cardio Heala vs MacroFactor: two ways to run a diet MacroFactor runs the best pure macro-adjustment algorithm on the market and never polices your food. Heala runs the same loop with training, recovery and a 3D body scan in the same app. Which you need depends on what else you track. 0wearables required for a recovery score Heala vs Oura: the ring and the whole picture Oura is one of the best sleep trackers you can wear, and it measures the night with real rigor. It doesn't know your training block, your lifts or your calorie target. Heala reads the ring's numbers through Apple Health and completes the picture. 1person who sees your Heala run Heala vs Strava: the feed and the plan Strava turned cardio into a social network: segments, kudos, a feed worth checking. Heala points the same GPS data inward — strain, recovery, an adaptive plan for tomorrow. They do different jobs, and plenty of people will run both. 0daily lessons — Heala assigns instruments, not homework Heala vs Noom: lessons or instruments Noom's bet is that you need teaching: daily lessons, food colours, a coach in your pocket. Heala's bet is that you need instruments: honest numbers in seconds and a plan that adjusts. Both bets are reasonable. They suit different people. 41actions Heala's coach can take Heala vs ChatGPT as your fitness coach ChatGPT is the better generalist: broader knowledge, better conversation, and it remembers what you tell it. Heala's coach is narrower but wired in — it reads your meals, sets, sleep and HRV, and it can act on them.

Heala opens 1 October 2026

Free forever — Pro is optional.

Photo food logging, adaptive training, a morning recovery score and a 3D body twin, in one app. Join the list and your spot — and your founding discount — are saved.

  • 15% off for life · first 1,000 members
  • Reserve your @username
  • One email at launch · no newsletter
Save my spot → Free · takes ten seconds