This is not medical advice. General information about food tracking and calorie targets, not a treatment plan. If you are under 18, pregnant or breastfeeding, have ever had an eating disorder or a difficult relationship with food, train at a competitive level, or are managing a medical condition, talk to a doctor or a registered dietitian before you start counting calories or change your intake. The full list is here.

Almost every calorie app that offers to credit a workout back is solving a problem it created. The daily target it handed you is not a resting number. It is an estimate of everything you do in a day, training included, so adding Tuesday's run on top of it counts Tuesday's run twice.

That part is arithmetic and it is not really in dispute. Sitting on top of it are the measurement problems: what a wrist device actually knows about energy expenditure, how far out people are when they guess instead, and how much of a workout's burn survives the rest of the day.

Two related questions have their own posts. How the resting half of the estimate is derived, and how wrong it can be about one person, is covered in how accurate are calorie tracking apps under the section on the target you are counting toward. Finding the number your weight holds steady at is maintenance calories. This one is only about what exercise should and should not be allowed to do to that number.

Your activity multiplier is a whole day, not a resting number

Every mainstream calorie target is built in two steps. An equation estimates your resting metabolic rate from height, weight, age and sex. That figure is then multiplied by a physical activity level, or PAL, and the product is what you would eat to hold your weight where it is.

The equation is the half people know about. MyPlate uses Mifflin-St Jeor, which a 2005 systematic review of the four equations in common clinical use found the most reliable of them, predicting resting metabolic rate within 10% of the measured value in more people than any alternative and with the narrowest error range. That same review flagged noteworthy errors when the equation is applied to an individual rather than a group.1

The multiplier is the half almost nobody reads, and it is where this question is decided. The joint FAO, WHO and UNU expert consultation that defines PAL for the rest of the field does not derive it from your job, or from your resting state, or from anything else that exercise could sensibly be added to afterwards. It builds a PAL by writing down a full 24 hours and adding up the energy cost of each block.2

Their worked example for an active lifestyle is a table you can check line by line: 8 hours sleeping, an hour of personal care, an hour eating, 8 hours standing and carrying light loads, an hour commuting by bus, an hour walking at varying paces, 3 hours of light leisure, and one hour of low intensity aerobic exercise costed at 4.2 times basal rate. The column totals 42.2, divided by 24 hours, which is where the PAL of 1.76 comes from.2

The workout is a row in the table rather than something added to the table afterwards.

The consultation is explicit that this is deliberate. It rejected the older practice of classifying people by their occupation alone, on the grounds that "there are people with light occupations who perform vigorous physical activity in their spare time, and people with heavy work who are quite sedentary the rest of the day."2 What replaced it classifies whole lifestyles: sedentary or light activity is a PAL of 1.40 to 1.69, active or moderately active is 1.70 to 1.99, and vigorous is 2.00 to 2.40, with a footnote that values above 2.40 are difficult to hold for long.2 The description of the sedentary band includes people who "do not exercise or participate in sports regularly", which tells you what the higher bands are carrying.

So should a calorie counter app add your exercise calories back?

No, if the activity level you picked describes the training you actually do. And if it does not, the answer is still not to eat back a session: it is to move the level, which is a different action with a different size.

MyPlate asks the question once, during setup, and asks it in workouts per week: low is 0 to 1, moderate is 2 to 3, high is 4 to 6, very high is 6 or more. Those map onto PAL values of 1.375, 1.55, 1.725 and 1.9. Most calorie calculator apps do the same two steps, and you can run this one yourself in the calorie calculator.

Take a 34-year-old woman, 72 kg and 168 cm, who wants to lose weight. Mifflin-St Jeor puts her resting rate at 1,439 calories. The app multiplies that by her activity band and then by 0.85 for the weight-loss goal, which gives a daily target of 1,681 on low, 1,895 on moderate, 2,109 on high and 2,323 on very high. Every step up that ladder is worth 214 calories a day, because every step is the same 0.175 of PAL.

Now put a workout next to it. Three sessions a week that a watch scores at 450 calories each come to 1,350 a week, or 193 calories a day spread across seven. One band step is 214. The band is not roughly gesturing at her training, it is accounting for it to within a rounding error. Credit the sessions on top and she has approximately doubled them.

She has also moved them. The band spread that training evenly over seven days, which is the honest way to hold an average. A per-session credit puts the whole of it back on the day she trained, which is also the day she is least inclined to argue with an app offering her 450 extra calories.

The genuine exception is a change in habits, not a change in a day. If you have gone from one session a week to five, the band you picked at setup is describing somebody else, and the fix is to change it. In MyPlate that lives in settings, behind the same activity question setup asked, and changing it rewrites the plan. What it is not is a reason to add Tuesday's run to Tuesday's target.

What a watch actually knows about calories burned

Suppose you wanted to eat the burn back anyway. The number you would be eating back is the least reliable figure on the screen.

Stanford researchers strapped seven wrist-worn devices onto 60 volunteers, selected to vary in age, height, weight, skin tone and fitness, and compared them against indirect calorimetry while the volunteers sat, walked, ran and cycled. Six of the seven measured heart rate to within a median error of 5% during cycling. Not one achieved an error in energy expenditure below 20%. The Apple Watch was the best of the group and the Samsung Gear S2 the worst, and error ran higher for men, for higher body mass index, for darker skin tone, and for walking.3

Their own conclusion is the sentence worth carrying: most wrist-worn devices adequately measure heart rate in laboratory-based activities but poorly estimate energy expenditure, "suggesting caution in the use of EE measurements as part of health improvement programs."3

Note "laboratory-based". Sitting, walking, running and cycling under supervision on a treadmill and an ergometer is the friendly case for a wrist device. A day containing housework, a commute, carrying a toddler and 40 minutes of circuits is not that case.

The broader picture is no kinder. A meta-analysis pooled 60 validation studies and 104 effect sizes, comparing wrist and arm-worn monitors against indirect calorimetry, room calorimeters and doubly labelled water in healthy adults. Accuracy varied by activity type, and heterogeneity was large and significant for many devices. Adding heart rate or heat sensing to accelerometry cut the error for most activity types, and research-grade devices did better on total expenditure but worse than commercial ones during ambulatory and sedentary tasks. The authors end by saying the estimates need improving, which is not the tone of a solved problem.4

The figure on your wrist is not random. It is just not accurate enough to settle a 450-calorie question about dinner, and its error stacks on top of every error already inside the target.

Without a device, people guess three to four times high

Guessing is worse, and somebody has measured exactly how much worse.

Willbond and colleagues took 16 moderately active, normal-weight adults aged 20 to 35 and had them complete treadmill sessions calibrated by indirect calorimetry to burn exactly 200 or 300 calories, both at the same intensity. Afterwards each person estimated what they had burned, then ate that many calories from a buffet. The 200-calorie session was estimated at 825 calories. The 300-calorie session was estimated at 897. Asked to eat back precisely what they had spent, they took 557 and 607 calories: less than they had estimated, and still two to three times what the session actually cost.5

Sixteen people, one laboratory, and the spread is enormous. That 825-calorie estimate carries a standard deviation of 1,062. This is a small study measuring something crude, and I would not quote the multiplier to two decimal places. The authors' own summary is that normal weight individuals overestimate the energy cost of exercise by three to four fold, and that when asked to compensate for it precisely with food, they still eat two to three times the measured cost.5

Only about 72% of the burn survives the day

There is a second reason the credit is too generous, and it has nothing to do with measurement error.

Careau and colleagues used the largest dataset assembled on adult total energy expenditure and basal energy expenditure, 1,754 people measured by doubly labelled water while living normal lives, to ask how much of the energy spent on extra activity actually reaches the day's total. Energy compensation in a typical human averages 28%, through a reduction in basal expenditure, which means roughly 72% of the extra calories burned in activity translate into extra calories burned that day.6

Compensation also varied considerably between people of different body composition, and the authors are careful about what that does and does not mean. They set out two readings and say deciding between them will be key: either people who compensate more are more likely to accumulate body fat, or getting fatter makes the body compensate more strongly, which would make losing fat progressively harder. They do not claim to know which.6

That 28% is an average across 1,754 people rather than a figure for you, but apply it to a perfectly measured 450-calorie session and about 324 of it is left by the end of the day. Apply it to a watch figure of the kind no device in the Stanford comparison estimated to within 20%, and there is not much left to build a meal on.

Why exercise moves the scale less than the arithmetic predicts

All of this shows up where you would expect it to: in exercise trials, where people lose less weight than the prescribed dose says they should.

Thomas and colleagues reviewed the studies that both monitored compliance with the exercise prescription and measured body composition, then worked backwards from each study's weight change to the energy deficit it implied. Their conclusion was that the small weight loss seen in most of the interventions was primarily due to low doses of prescribed exercise energy expenditure, compounded by a concomitant increase in caloric intake.7 The eating back happens whether or not an app suggests it.

King and colleagues put 35 overweight and obese sedentary adults through 12 weeks of supervised exercise, five sessions a week. The group lost an average of 3.7 kg, close to what the prescription predicted, which taken alone would suggest nobody compensated for anything. Sorting people by whether their own loss matched their own prediction breaks that apart: the non-compensators lost 6.3 kg and the compensators lost 1.5 kg, across a range running from minus 14.7 kg to plus 1.7 kg on the same programme.8

What separated them was mostly intake. Compensators ate 268 calories a day more than at baseline and were hungrier by the end; non-compensators ate 130 calories a day less, with no change in their appetite ratings.8 Those intake figures come with standard deviations near 455 and 485 calories a day and rest on self-report over 12 weeks, so read them as a direction rather than a dose.

The useful part is what it says about averages. A group mean hid a 16 kg spread, and there is no way to tell from the outside which half of it you are in. A food intake tracking app is the cheapest instrument available for finding out, and that is the one argument for tracking this entire section supports.

What the evidence does not settle

Three questions I went looking for answers to and could not settle.

Whether exercise prevents weight regain. The ACSM position stand on physical activity for weight management found that moderate-intensity activity of 150 to 250 minutes a week provides only modest weight loss, that more than 250 minutes a week is associated with clinically significant loss, and that cross-sectional and prospective studies link more than 250 minutes a week to better maintenance after loss. It then says, plainly, that "no evidence from well-designed randomized controlled trials exists to judge the effectiveness of PA for prevention of weight regain after weight loss."9 That was written in 2009 and I could not find a later randomised trial that closes it.

Whether compensation is something you are or something that happens to you. Careau's dataset establishes that compensation tracks with adiposity and states outright that the direction of causality is undetermined.6 If it is a trait, some people should expect less from added activity than others and there is currently no way to find out which you are except by measuring yourself for a couple of months.

How wrong a wrist device is across a whole week of ordinary life. Nearly every validation study above is minutes to hours of defined activity under supervision. The number an app would actually credit you is a free-living daily total, and the heterogeneity in that meta-analysis is wide enough that I would not attach a single percentage to it.34

How to set the activity band, and then leave it alone

If you are counting calories to lose weight, the activity band is the one setting that moves your target by hundreds of calories, and it takes about ten seconds to get approximately right.

Count the sessions you did over the last month and divide by four. That is your ordinary week, which is the thing the band is asking about, rather than your best week or the week you intend to have. Where you fall between two bands, take the lower one. Being wrong low shows up as losing faster than you planned, which is visible within a fortnight and easy to correct upward; being wrong high shows up as a month of nothing happening, which is the version people quit over.

Then leave it alone and let the log do the work. Track what you eat for two or three weeks, weigh yourself on a fixed schedule, and compare the trend against what the target predicted. That comparison is the only measurement in this entire post taken on you rather than on a population, and it makes every equation above redundant the moment you have it.

Change the band when your habits change, not when your week does. A fortnight of illness or one heroic Saturday is not a band change. Five sessions a week for a month, when you used to do one, is.

Two limits of my own before you act on any of this. MyPlate has no exercise logging and no Apple Health connection, which in this specific argument is a design choice I would defend, but it also means the app has no way to represent a week that is genuinely unlike your others. And logging a weight in MyPlate does not recompute your target: your expenditure falls as you get lighter, and the app keeps showing the number it worked out at setup until you re-run it. That is on my list and it is not fixed yet.

One thing this post is not an argument for is eating less. The lowest target MyPlate will show is 1,200 calories a day for women and 1,500 for men, taken from the NIH and NHLBI clinical guidelines, and that floor overrides the arithmetic whatever the arithmetic says.10 If you train seriously, the number that should move is the band, upward. Eating well under what your training costs has a name and a literature: the IOC's 2023 consensus statement defines low energy availability as inadequate energy intake in relation to exercise energy expenditure, and describes a syndrome of health and performance consequences in both female and male athletes.11 That is a conversation for a doctor rather than an app.

And if tracking has turned into something you are trying to win, where a missed session or an unlogged meal spoils the day, stop and talk to someone about it. That is a better use of the next hour than anything above.

Common questions

Should you eat back your exercise calories?

No, not if the activity level in your app describes the training you actually do. A daily calorie target is your resting metabolic rate multiplied by a physical activity level, and the joint FAO, WHO and UNU consultation that defines those levels builds each one from a full 24 hours with exercise as one of the rows: its worked example for an active lifestyle includes an hour of low intensity aerobic exercise inside a PAL of 1.76.2 Crediting a session on top counts that session twice. If your training has genuinely changed, raise the activity level instead. In MyPlate that is a setting you change when your habits change, not a daily adjustment.

How accurate is the calories burned number on my watch?

Not accurate enough to eat against. Stanford researchers tested seven wrist-worn devices on 60 volunteers against indirect calorimetry and found that none achieved an error in energy expenditure below 20%, while six of the seven measured heart rate to within a median error of 5% during cycling.3 Error was worse for men, for higher body mass index, for darker skin tone and for walking, and those were supervised laboratory activities rather than a free-living day. A meta-analysis of 60 validation studies found accuracy varies by activity type with large heterogeneity between devices.4 If you are shopping for the most accurate calorie tracker, the useful thing to know is that no consumer device measures either side of the equation: intake is looked up or estimated, and expenditure is modelled from movement and heart rate.

Does MyPlate track exercise or connect to Apple Health?

No to both. MyPlate asks how many workouts you do in a typical week during setup, uses the answer to pick a physical activity level between 1.375 and 1.9, and does not ask again unless you change it in settings. There is no workout log and no calories-burned credit, so the app has nothing to double count. The cost of that is real and worth stating: if your activity swings from week to week, a single band cannot express it, and logging a weight does not recompute your target either.

Which activity level should I pick in a calorie counter app?

The one that matches your ordinary month, and the lower of the two if you are between them. Count the sessions you actually did over the last four weeks and divide by four rather than describing the week you intend to have. Picking too low shows up as losing weight faster than you planned, which is visible within a fortnight and easy to correct; picking too high shows up as a month in which nothing happens, which is the version people quit over. For reference, the FAO, WHO and UNU consultation classes a whole sedentary or light-activity lifestyle at a PAL of 1.40 to 1.69, an active one at 1.70 to 1.99, and a vigorous one at 2.00 to 2.40, noting that anything above 2.40 is hard to sustain.2

I have started training five days a week. What should I change?

The activity band, once, and then nothing for a fortnight. For a 34-year-old woman of 72 kg and 168 cm on a weight-loss goal, moving up one band in MyPlate is worth 214 calories a day, because each band step is the same 0.175 of physical activity level applied to the same resting rate. That is close to what three 450-calorie sessions a week come to when spread over seven days, which is 193 a day. After the change, log your food for two or three weeks and compare the weight trend against what the new target predicted. That comparison measures you, where every equation in the app is measuring a population.

Move the band, not the day

The target your app gives you already contains the training you habitually do, because that is how a physical activity level is constructed.2 A credit for today's workout is therefore not extra information, it is the same information entered twice, sized by a device of the kind that the Stanford comparison could not get within 20% on energy expenditure.3

What is left after you stop adjusting daily is duller and more useful. One band that matches your ordinary month, one target, and two or three weeks of logging to find out whether the estimate was right about you. King's trial is the reason that last step matters: the same 12-week programme produced everything from a 14.7 kg loss to a 1.7 kg gain, and nothing on the outside of a person predicts which they will be.8

None of which is an argument against training. The same ACSM review that could not find trial evidence for exercise preventing regain reports that endurance activity or resistance training improves health risk even without weight loss.9 That is the better reason to do it, and it does not require the app to pay you back in calories.

Sources

Every figure above traces to one of these. If you find a number that doesn't match the source it claims, tell me and I'll correct it.

  1. Frankenfield D, Roth-Yousey L, Compher C (2005). Comparison of predictive equations for resting metabolic rate in healthy nonobese and obese adults: a systematic review. Journal of the American Dietetic Association, 105(5):775-789. Read on PubMed →
  2. FAO/WHO/UNU (2004). Human energy requirements: report of a Joint FAO/WHO/UNU Expert Consultation. FAO Food and Nutrition Technical Report Series, 1. Chapter 5, Energy requirements of adults. Read chapter 5 →
  3. Shcherbina A, Mattsson CM, Waggott D, et al. (2017). Accuracy in wrist-worn, sensor-based measurements of heart rate and energy expenditure in a diverse cohort. Journal of Personalized Medicine, 7(2):3. Read on PubMed →
  4. O'Driscoll R, Turicchi J, Beaulieu K, et al. (2020). How well do activity monitors estimate energy expenditure? A systematic review and meta-analysis of the validity of current technologies. British Journal of Sports Medicine, 54(6):332-340. Read on PubMed →
  5. Willbond SM, Laviolette MA, Duval K, Doucet E (2010). Normal weight men and women overestimate exercise energy expenditure. Journal of Sports Medicine and Physical Fitness, 50(4):377-384. Read on PubMed →
  6. Careau V, Halsey LG, Pontzer H, et al. (2021). Energy compensation and adiposity in humans. Current Biology, 31(20):4659-4666. Read on PubMed →
  7. Thomas DM, Bouchard C, Church T, et al. (2012). Why do individuals not lose more weight from an exercise intervention at a defined dose? An energy balance analysis. Obesity Reviews, 13(10):835-847. Read on PubMed →
  8. King NA, Hopkins M, Caudwell P, Stubbs RJ, Blundell JE (2008). Individual variability following 12 weeks of supervised exercise: identification and characterization of compensation for exercise-induced weight loss. International Journal of Obesity, 32(1):177-184. Read on PubMed →
  9. Donnelly JE, Blair SN, Jakicic JM, et al. (2009). American College of Sports Medicine position stand: appropriate physical activity intervention strategies for weight loss and prevention of weight regain for adults. Medicine & Science in Sports & Exercise, 41(2):459-471. Read on PubMed →
  10. National Heart, Lung, and Blood Institute, NIH (1998). Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults. NIH Publication 98-4083. Read the guidelines →
  11. Mountjoy M, Ackerman KE, Bailey DM, et al. (2023). 2023 International Olympic Committee's (IOC) consensus statement on Relative Energy Deficiency in Sport (REDs). British Journal of Sports Medicine, 57(17):1073-1097. Read on PubMed →

The formulas and limits behind MyPlate's own numbers, with their citations, are on the health sources page.