Exercise adds fewer calories to your day than your tracker says for two separate reasons: the tracker's number is wrong to begin with, and even the right number does not all land in your daily total. Wrist devices miss exercise energy by a fifth or more, and the body compensates for a hard session by doing a little less elsewhere. Neither is a reason to skip training; both are reasons never to eat the tracker's number back.
Energy compensation in numbers
The additive model says a 400 kcal run adds 400 kcal to the day. The evidence says the day adjusts. In a study of 332 adults across five populations, total energy expenditure measured with doubly labelled water rose with physical activity at low activity levels but plateaued at the upper range, supporting what the authors called a constrained model of total expenditure2. More movement did not keep buying more expenditure; the body held the total within a narrower band than simple addition predicts.
A larger dataset put a number on it. Pooling doubly labelled water and basal measurements from 1,754 adults living normal lives, Careau and colleagues estimated that energy compensation averages 28%, through a lower basal rate: only about 72% of the extra calories burned in activity showed up in the day's total3. Compensation varied between people and was linked to body composition3. A 2026 review of exercise trials found total expenditure rose by only about 30% of what the additive model expects in aerobic exercise programmes, with less compensation during resistance training and more when exercise was paired with eating less6.
Share of extra activity energy that showed up in daily expenditure · %
Not every study agrees. A 2025 observational study found a straight-line link between activity and total expenditure, with no sign of a plateau or compensation5. So the size of the effect is debated; its existence in many datasets is not.
Where does the missing energy go? Three places, none of them mysterious:
- Less movement elsewhere. After a hard session people sit more, fidget less and take the lift — NEAT falls (see what is NEAT). The session is real; the rest of the day quietly pays for part of it.
- Better economy. A trained body does the same work for less energy: the same 5 km costs a practised runner less than it cost them as a beginner. How much less is individual, so this article puts no number on it.
- Physiological adaptation. Some non-movement processes appear to run a little lower when activity is high — the plateau in the data above is not explained by sitting alone.
An illustrative band, as arithmetic rather than a measurement of you: a session that costs 400 kcal on the machine might raise the day's total by 250–300 kcal while eating at maintenance, and by 150–200 kcal while in a deficit, when compensation is stronger. The exact split is individual; the direction is not. For scale, the 28% average above would leave about 290 kcal of a 400 kcal session in the day's total3.
| What the session cost | What a tracker might show | Illustrative net addition to the day, maintaining | Illustrative net addition, in a deficit |
|---|---|---|---|
| 400 kcal | 480–520 kcal if it reads 20–30% high (it can also read low) | ≈ 250–300 kcal | ≈ 150–200 kcal |
Read across the row: a tracker might display 500, the true cost might be 400, and the day's total might rise 250. Eat the 500 back and the "deficit" is gone.
In plain terms: the run happened, the calories were spent — but the day around it shrank a little to make room.
Why wearable calorie counts are unreliable
Now the tracker column of that table. Wrist-worn trackers estimate energy from heart rate, motion and your profile, through a model that has never met you. Shcherbina and colleagues tested seven commercial wrist devices, bought retail, on 60 volunteers of varied age, size, skin tone and fitness while sitting, walking, running and cycling, against continuous telemetry and indirect calorimetry. Six of the seven measured heart rate to within a 5% median error during cycling; no device achieved an error in energy expenditure below 20 percent1. Error was higher for men, for higher body mass index, for darker skin tone and for walking.
That is the key asymmetry: the pulse is measured, the calories are modelled. Heart rate at a given effort varies with fitness, heat, sleep and stress, and the conversion from pulse to energy is a population regression with the same weakness as a BMR equation — right on average, wrong for the person wearing it, and wrong in a direction that is hard to predict. The Stanford authors' conclusion was that devices adequately measure heart rate in the lab but poorly estimate energy expenditure, and suggested caution in using the EE numbers1.
Twenty percent was the smallest error in that study. Across tasks, the median error in energy ranged from 27.4% for the best device to 92.6% for the worst, while heart-rate error ranged from 2.0% to 6.8%1. On a session the device calls 500 kcal, that is at least 100 kcal of uncertainty before compensation is even considered.
Median error of seven wrist devices: heart rate vs calories burned · %
Best device Worst device
Those are sizes of error, not directions. The study measured how far each reading was from the lab's figure, not whether it was high or low1, so a tracker can miss in either direction. The authors could not say why energy estimates were so poor: the calculations are proprietary, and people differ in how much energy the same activity costs them1.
One study is one lab. A 2022 systematic review of 65 articles found the mean absolute percentage error (the average miss, ignoring direction) for energy expenditure was above 30% for every brand of wrist device tested, and concluded that none proved accurate for it4.
Why we do not add exercise calories back
Given both effects, adding a tracker's figure to a calorie target is the wrong operation twice over:
- It double-counts. The activity multiplier in our calorie calculator already includes your training frequency — "4–6 sessions a week" is priced into the
1.55. Adding the sessions again on top is counting them twice. - It imports the tracker's error. A number that can be 20–30% off, added to a target that is itself a 100–200 kcal estimate, makes the target worse, not better.
- It ignores compensation. Even a perfect session figure overstates what the session adds to the day.
Zyra Nutrition is designed around the other route. Its expenditure tutorial answers "do I need to track exercise?" with a plain no: weight and nutrition data are meant to give the algorithm everything it needs. The reasoning holds — the weight trend already contains the net effect of every session, after compensation, after economy, after the lift you took instead of the stairs — so an expenditure read from the trend captures what a tracker cannot. Status, plainly: the expenditure Zyra Nutrition shows today is the onboarding equation; the update that reads it from your logs is coming, not live (see how apps estimate energy expenditure). Either way, no exercise figure is added back to the target.
What to track instead
- Intake, honestly logged. Intake decides the trend; exercise nudges it. Three hundred kcal of unlogged food undoes a session's real contribution every day.
- Trend weight over three to four weeks. That is the window in which the net effect of your training plus your eating becomes legible through the water noise.
- Training itself, in training units. Sets, reps, load, distance and time are measured; "calories burned" is modelled. If a session went well, the log of the session says so — in Zyra Workouts or on paper — without a calorie figure attached.
- The tracker's number as an activity index, if at all. It is fine for "was today a big day or a small one?" It is not fine as an input to a calorie target.
What the studies found
| Study | Who was studied | Key result | Year |
|---|---|---|---|
| Pontzer et al., *Curr Biol*2 | 332 adults in five populations; doubly labelled water and accelerometers | Total expenditure rose with activity at low levels and plateaued at high levels | 2016 |
| Shcherbina et al., *J Pers Med*1 | 60 adults, seven wrist devices, lab tasks vs ECG and indirect calorimetry | Heart-rate error 2.0–6.8% by device; no device got energy within 20% (27.4–92.6%) | 2017 |
| Careau et al., *Curr Biol*3 | 1,754 adults living normal lives; doubly labelled water | Energy compensation averaged 28%: about 72% of extra activity energy reached the daily total | 2021 |
| Germini et al., *J Med Internet Res*4 | Systematic review, 65 articles on wrist trackers | Energy-expenditure error above 30% for every brand; none accurate for it | 2022 |
| Howard et al., *PNAS*5 | Observational study of activity and doubly labelled water expenditure | Straight-line link between activity and total expenditure; no evidence of compensation | 2025 |
| Pontzer & Trexler, *Curr Biol*6 | Review of exercise trials and free-living studies | In aerobic exercise trials, total expenditure rose by about 30% of the additive prediction | 2026 |
Limits worth knowing
- Devices age fast. The lab study tested the devices and software of 2017; newer models may do better or worse1.
- Lab, not life. Short, controlled treadmill and bike stages are not a day of mixed activity.
- Compensation is an average. It varied between people3, and one 2025 study found none5.
- Cross-sectional data. The 1,754-person dataset compares different people, which cannot show that the same person compensates when they add exercise.
- Abstracts only for the 2021, 2022, 2025 and 2026 papers; we read the 2017 lab study in full.
Latest research
Two 2026 papers sharpen both halves of this article. A systematic review of 47 studies of wrist trackers found heart rate and step counts have become much more accurate over two decades, but energy expenditure improved least: even recent devices showed 10–40% error against indirect calorimetry7. And a review of exercise trials and field studies concluded that, in aerobic exercise programmes, total expenditure rose by only about 30% of what simple addition predicts6. The counterweight is the 2025 study that found no compensation5; the debate is open, and none of it changes the practical point that the weight trend, not the watch, shows what training added.
This is arithmetic and cited measurement, not medical advice. The compensation band is illustrative; the tracker error is what one laboratory study found for seven specific devices in 2017 — newer devices may differ, and the way to know is to compare them against your own trend, not against their own marketing.
The short version
Your tracker's exercise energy figure is unreliable — no wrist device in a seven-device lab study got within 20%1 — and your body compensates for a session by spending a little less elsewhere, so a 400 kcal run might add 250 to a maintenance day and less in a deficit. We never add exercise calories back: the activity multiplier already contains training, and the weight trend already contains its true net effect.
- Wrist devices measure heart rate well but miss calories burned by 20% or more, in either direction14.
- In a large dataset, about 72% of extra activity energy reached the daily total on average3; exercise trials suggest less, one study found no offset.
- The weight trend over a few weeks shows the net effect of training; a watch's calorie figure does not.
Educational overview only — not medical advice.
