The number on a food label and the number a tracker reports for an hour of exercise are not the same kind of measurement. A label figure is calculated from the Atwater factors and carries a legal tolerance of around 20 per cent; a burn figure is read from a table of population averages and is rarely better than 25 per cent for an individual. Subtracting the second from the first compounds both errors, which is why the daily balance is the least reliable number either one produces.
That does not make either figure useless. It makes them good at comparing things measured the same way — this cereal against that one, a run against a swim — and bad at the arithmetic people most want to do with them.
Where does the number on the label come from?
Almost never from burning the food. It is calculated: grams of protein, carbohydrate and fat multiplied by fixed energy factors and added up. Those factors are the Atwater system, derived in the 1890s from bomb calorimetry corrected for what the body fails to digest and what it excretes.
| Component | Factor | Notes |
|---|---|---|
| Protein | 4 kcal/g | Corrected downward for urinary nitrogen loss |
| Carbohydrate | 4 kcal/g | Available carbohydrate, not total |
| Fat | 9 kcal/g | The densest, and the one worth weighing |
| Alcohol | 7 kcal/g | Declared in energy but not in the macros |
| Fibre | 2 kcal/g | EU figure; some regimes count it as zero |
| Polyols | 2.4 kcal/g | Sugar alcohols, partially absorbed |
A calorimeter measures the heat a food releases when it is burned completely. A person is not a calorimeter: some of the energy leaves undigested, and some is spent on digestion itself. The Atwater factors are an average correction for that gap, and an average is exactly what they remain.
How far out can a label be?
Further than most people assume, and in a direction that is not random. In the United States a declared value is treated as compliant if the food contains no more than 20 per cent more than the label claims; the EU works to comparable tolerance guidance. The figure is also an average of a production run rather than a measurement of the packet in your hand.
On top of that tolerance sits a systematic error. The Atwater factors overstate the energy actually absorbed from foods whose structure resists digestion — whole nuts are the studied case, where measured absorption has come in meaningfully below the labelled figure because intact cell walls pass through. High-fibre foods behave the same way to a smaller degree.
The error runs the other way for heavily processed food. Cooking gelatinises starch and denatures protein, and both make more of the energy reachable. The same 100 grams of oats delivers more absorbed energy cooked than raw, and the label does not distinguish between the two.
And the figure for what you burn?
A different method with a wider band. Most activity figures come from MET values — multiples of resting metabolic rate, where one MET is defined as 3.5 millilitres of oxygen per kilogram per minute. The MET arithmetic covers how the multiplication works.
The problem is not the arithmetic, it is the input. A MET value is a population average taken from a modest number of subjects, and real people vary around it by a quarter or more depending on fitness, technique and body composition. Two people of the same mass doing the same activity for the same time do not spend the same energy, and the table cannot know that.
A wrist device adds its own error rather than removing any. Heart rate is measured reasonably well; the conversion from heart rate to energy is where the accuracy goes, because that relationship is individual and the device has not measured yours. Validation work has repeatedly found energy-expenditure error far larger than heart-rate error on the same device.
What happens when you subtract one from the other?
The uncertainties add rather than cancel, so the difference is less certain than either figure that went into it. That is the part the arithmetic hides: two numbers that each look precise to the calorie produce a balance that is not.
| Quantity | Typical figure | Plausible range |
|---|---|---|
| Intake, as labelled | 2,400 | 1,900 to 2,900 |
| Expenditure, as estimated | 2,900 | 2,200 to 3,600 |
| Implied deficit | 500 | Anywhere from a surplus to 1,700 |
A deficit of 500 with a band that wide is not a plan, it is a hypothesis. The honest reading is that the ledger sets a direction and the scale over several weeks decides whether the direction was right.
What are these numbers good for, then?
Comparison, which is where a shared bias stops mattering. Two cereals calculated with the same Atwater factors can be ranked against each other even though neither figure is exact, and the per 100 g column is what makes that comparison fair. The same applies to activities: a MET table ranks a swim against a walk correctly even where it prices neither well.
Tracking a trend is the other real use. A logging error that repeats every day is a constant, and a constant subtracts out of a week-on-week change even when it never subtracts out of the daily total.
What none of it supports is treating a session as a licence — the idea that an hour of exercise cancels a particular food. Both sides of that trade carry error bars wider than the food, and the reasons a deficit stalls have more to do with the total moving than with any single evening.
Questions people ask
Is 3,500 calories per pound of fat right? It is a static approximation and it overstates loss over time, because maintenance falls as body mass does. It is a reasonable first estimate for a week and a poor one for a year.
Does the body absorb every calorie on the label? No, and the shortfall varies by food and by how it was prepared. Whole nuts and high-fibre foods deliver less than the label states; refined and well-cooked foods deliver close to it.
Are two labels on similar products comparable? Usually yes, because they were calculated the same way. That shared method is what makes the comparison sound even where the absolute figures are not.
Should I bother logging at all? As a calibration exercise it is genuinely useful, because most people are surprised by where their intake actually sits. As a permanent daily balance to the calorie it is precision the underlying numbers cannot support.
Does cooking change the calories? It changes how many are available to you, which is the figure that matters, while leaving the number printed on the packet exactly where it was.
Two numbers, two methods, two error bars, and only one of them printed as though it were exact. The calories per serving calculator turns a recipe into the per-portion and per-100 g figures a label would carry, the calories burned calculator puts an activity in MET terms so the estimate is at least explicit about where it came from, and the TDEE calculator shows how small the exercise share of a day usually is.