Where the numbers come from
The application shows you a daily target, a deficit and a split of macros. Every one of those is an estimate rather than a measurement, and none of it is medical advice. What follows is where each number came from: where a paper stands behind it, where a guideline does, and where the answer is that we decided it ourselves.
Contents
Where does the resting expenditure come from?
Resting expenditure is calculated with the Mifflin-St Jeor equation:
Resting expenditure, kcal =
10 × weight, kg
+ 6.25 × height, cm
− 5 × age, years
+ 5 for men or − 161 for women
You may leave sex unstated — the midpoint between those two constants is used instead, and the application says that it errs by more when you do. Height, age and weight are never invented: without any one of the three there is no number at all, because an average stranger's figure would look as though it had been worked out for you.
Mifflin MD, St Jeor ST, Hill LA, Scott BJ, Daugherty SA, Koh YO, 1990. A new predictive equation for resting energy expenditure in healthy individuals. The American Journal of Clinical Nutrition 51(2):241–247.
How wrong can that formula be?
The error of the formula itself is around 10% of resting expenditure, and that is the standard error of the estimate rather than the worst case. The number matters: the only statement the application makes that resembles an accusation — that some of the food went unlogged — rests on it. Without a margin for that error the remark would land on honest people, and not at random: sedentary users have the least room above the floor, so the misfire would concentrate on them.
The error is named by the paper itself: Mifflin MD et al., 1990. A new predictive equation for resting energy expenditure in healthy individuals. And a check against independent data: Frankenfield D, Roth-Yousey L, Compher C, 2005. Comparison of predictive equations for resting metabolic rate in healthy nonobese and obese adults. Journal of the American Dietetic Association 105(5):775–789.
Where do the activity multipliers come from?
Resting expenditure is multiplied by the level you pick:
| Level | Multiplier | Example |
|---|---|---|
| Sedentary | 1.2 | Desk job, little walking, no workouts |
| Light | 1.375 | 1–2 light workouts a week |
| Moderate | 1.55 | 3–5 workouts a week |
| High | 1.725 | 6–7 workouts or physical work |
| Athlete | 1.9 | 8+ workouts a week, sport |
Everyone quotes this ladder and there is no primary source behind it — the nearest real thing is the physical activity levels in the joint FAO, WHO and UNU report, which are arranged differently. This is also the crudest part of the whole calculation: the gap between the outer rungs is larger than the error of everything else put together. Given a choice between two neighbouring rungs, take the lower one — that much is our own choice rather than a finding of the report.
FAO / WHO / UNU, 2004. Human energy requirements. Report of a Joint FAO/WHO/UNU Expert Consultation. FAO Food and Nutrition Technical Report Series 1.
Why is a kilogram of body weight 7,700 kcal?
A rate in kilograms per week becomes a daily deficit through that constant. It is a simplification, and a well documented one: over long stretches, multiplying a deficit by the number of days overstates the loss, because expenditure falls along with the weight. This is why the rate is computed back out of the target that was actually set rather than handed back as asked for, and why the forecast has a limit at all.
Where the number comes from: Wishnofsky M, 1958. Caloric equivalents of gained or lost weight. The American Journal of Clinical Nutrition 6(5):542–546. Where it stops holding: Hall KD, Sacks G, Chandramohan D, Chow CC, Wang YC, Gortmaker SL, Swinburn BA, 2011. Quantification of the effect of energy imbalance on bodyweight. The Lancet 378(9,793):826–837.
Why not offer a faster rate?
Faster than 1% of body mass per week is not something we propose. The limit is on our suggestion rather than on you: any rate can be entered, but the target will be worked out from the clamped one, and the clamping is stated in words. There is no primary source here either; the nearest is a comparison of two rates in athletes, where the slower loss preserved strength better.
Garthe I, Raastad T, Refsnes PE, Koivisto A, Sundgot-Borgen J, 2011. Effect of two different weight-loss rates on body composition and strength and power-related performance in elite athletes. International Journal of Sport Nutrition and Exercise Metabolism 21(2):97–104.
Why does the target never go below resting expenditure?
The target never drops below resting expenditure, whatever rate you ask for. That is our own choice, and it is built as a refusal rather than as advice: the application does not tell you what you ought to eat, it stops proposing less than a certain amount. Declining to advise is not advice, which is precisely why the line sits here rather than somewhere that would start a conversation about your health. Hitting it is stated in words.
Where does the protein target come from?
Protein is set by body mass rather than as a share of the calories: 1.8 g per kilogram when the goal is losing weight, and 1.6 otherwise. It does not depend on the calorie figure at all, and that matters exactly in a deficit, where protein is what holds muscle: a proportional split would drag it down with the shrinking number, taking it away precisely when it is needed most.
Morton RW, Murphy KT, McKellar SR, et al., 2018. A systematic review, meta-analysis and meta-regression of the effect of protein supplementation on resistance training-induced gains in muscle mass and strength. British Journal of Sports Medicine 52(6):376–384. On deficits: Helms ER, Zinn C, Rowlands DS, Brown SR, 2014. A systematic review of dietary protein during caloric restriction in resistance trained lean athletes. International Journal of Sport Nutrition and Exercise Metabolism 24(2):127–138.
Why does protein hit a ceiling?
Protein has a ceiling: no more than 35% of the calories. Our own choice, with no source under it. It earns its place on the combination of a high body mass and a low target, where the per-kilogram figure eats the whole plate and leaves nothing for carbohydrates. Hitting the ceiling is named in words rather than applied quietly: clamping without saying so swaps your decision for ours.
Why does fat have a floor as well?
A share of the calories, but never below 0.8 g per kilogram of body mass. Our own choice, and this particular number is ours entirely: on a low target, a plain share drops fat to where it stops being food. If even that floor will not fit inside the target, the application lowers it again and says so — bars that cannot be met by construction are worse than an honestly tight plate.
How do grams turn into calories?
Through the Atwater factors: 4 kcal in a gram of protein, the same in a gram of carbohydrate and 9 in a gram of fat. The same factors take the target back apart, which is why the macros always add up to the calorie figure, give or take the rounding of grams to whole numbers. You can check that on any target with a calculator, and our own checks do exactly that.
FAO, 2003. Food energy — methods of analysis and conversion factors. FAO Food and Nutrition Paper 77.
Why does the portion weight need your eye?
From a photo the model returns, for each item on the plate, a name, the numbers per 100 g and an estimate of the serving in grams. The app multiplies one by the other: the portion’s calories, protein, fat and carbohydrates are the numbers per 100 g times the grams. So the grams are a multiplier on everything — a serving read as 150 g instead of 250 g takes 40% off every number at once.
A better model does not cure this, because the photo does not hold the answer. A picture shows the area of the food, not its depth or how tightly it is packed; the same plate holds very different amounts of rice; oil and butter soaked into the food and a sauce underneath are invisible; and nothing in the shot tells the size of the plate. What is on the plate the model reads well from the photo; how much of it there is, it has to guess. That is why it is asked to rate its confidence in the serving weight specifically, and when that confidence is below 50% the item says so: “Confidence in the weight estimate: … — the grams almost certainly need fixing”.
What to check, then, is the weight — not the name and not the numbers per 100 g, unless the item itself was recognised wrongly:
- “Portion weight, g” of each item — from a kitchen scale or the weight on the pack. The numbers per 100 g stay as they are, and the portion is recounted from the new grams.
- “Total weight”, if you weighed the whole dish: the grams of all items are scaled to it in proportion.
- Anything the photo cannot show — oil in the pan, a sauce, sugar in the tea — add with “Add item manually”.
- If an item was recognised as the wrong food, rename it and press “Recalculate”: the numbers per 100 g are worked out again from the description, and the weight is left to you.
Describing a meal in words takes part of the guess away where you know the amount: the model is told to take a weight or a volume named in the text (“rice, 200 g”) as given, and to make the items add up to a total weight given under the description. It is an instruction to the model rather than a calculation of ours, so the grams are still worth a glance.
When do the two estimates of expenditure agree?
The application works out your expenditure twice, by two separate routes: from the formula, and from the slope of your weigh-ins. The two count as agreeing if they differ by no more than 12%. Our own choice: that is roughly one rung of the activity ladder, and a gap smaller than a rung means nothing. When the measurement interval is wider than the threshold itself, the application says nothing at all — calling something agreement when it simply was not measured is worse than silence.
Where does the margin before the under-logging remark come from?
Before saying that some of the food went unlogged, the application takes a margin: a one-sided 95% interval on the error of the formula, which is a multiplier of 1.645. There is no source for that number and there cannot be — it is a definition, read off the normal distribution. One-sided, because there is only one direction in which being wrong here costs anything: accusing somebody who logged everything honestly is dearer than staying quiet.
Why does the forecast stop at 1,095 days?
The forecast of when you reach your target is not shown beyond that. Our own choice: past that point it is fortune-telling rather than a forecast, and naming a date nobody will meet is worse than naming none. It is computed from the achievable rate rather than the requested one, for the same reason as the warning about the constant above.
When does the app offer to change your target?
Only after at least 28 days of weigh-ins, and only as an offer. If your weight has been moving clearly slower than the plan, and was a week earlier too, while the food you logged stayed near the target, the app suggests a step of up to 10% of the target, and no more than 20% in all. Nothing changes until you press “Apply”; “Not now” pauses the question for 14 days. When weight moves much faster than planned, the same rule gives back what an adjustment took. Our own choice, and every threshold in it is ours. There is no offer to cut at all under 18, with a body mass index below 20, when you asked for nothing about weight loss, or with a health mark that rules it out; and the target never goes below resting expenditure. It is an estimate from your own weigh-ins and entries, not medical advice.
Where the check-up list comes from
Prevention is a reference to published recommendations for adults without symptoms, selected by age, sex, country of care, answers about smoking, pregnancy plans and the organs a check applies to, and the foods you avoid; eating patterns also by your goal. It does not diagnose or prescribe.
Sources name the organisation, edition, jurisdiction, status and check date. Texts were summarised with AI and checked against the source; they have not been reviewed by a doctor. Dates mean the edition was checked then, not that it is current today. Re-check is planned at least annually.
The source table is prepared for the release of this section and will appear here when it opens. European guidance remains reference information, without national or personal due dates. The scope is described in terms of use.