Home › Blog › Why Calorie Trackers Disagree
Why Every Calorie Tracker Gives You a Different Number
You run 5 kilometres. Your watch says 412 calories. The treadmill says 340. Strava says 385. An online calculator says 350. All four are describing the same run — and all four are estimates built on different assumptions.
None of them is measuring calories. Calories cannot be measured directly outside a laboratory. What every device is actually doing is estimating energy expenditure from proxies, and the numbers diverge because the proxies differ.
What a calorie actually is, and why it is hard to measure
A kilocalorie is the energy required to raise one kilogram of water by one degree Celsius. In the body, energy is released when carbohydrate and fat are oxidised, and that process consumes oxygen in a fixed and predictable ratio.
The reference method for measuring energy expenditure is indirect calorimetry: you breathe into a mask, and the analyser measures oxygen consumed and carbon dioxide produced. From the ratio between them, energy expenditure can be calculated to within a few percent, along with the proportion of fuel coming from fat versus carbohydrate.
This is accurate and it is also impractical. You cannot run a Sunday long run wearing a metabolic cart. So every consumer device substitutes something measurable for oxygen consumption, and the substitution is where the error enters.
Method 1: MET-based estimation
The approach used by most online calculators, including this one, and by many apps:
MET values come from the Compendium of Physical Activities, which assigns each activity an energy cost based on measured oxygen consumption across study populations. Running at 9.7 km/h is rated at 9.8 METs. Multiply by body mass and time and you have an estimate.
Strengths: transparent, reproducible, and correctly weights the two variables that matter most — body mass and duration. Anyone can check the arithmetic.
Weaknesses: it is a population average. It cannot know that you have unusually good running economy, that you are running into a headwind, or that today's route was on sand. Expect ±10 to 15 percent for an individual.
Method 2: heart-rate-based estimation
Fitness watches with optical or chest-strap heart rate use algorithms that map heart rate, and often heart rate reserve, to oxygen consumption, adjusted for age, sex, weight, and sometimes measured VO₂max.
Strengths: genuinely personalised. It responds to your individual cardiovascular response rather than an average, and it handles varied terrain — hills raise heart rate, and the algorithm sees that.
Weaknesses: heart rate is affected by many things that do not change energy expenditure. Caffeine raises it. So do heat, dehydration, poor sleep, stress, and illness. On a hot day your heart rate might be 10 beats higher at the same pace, and the watch will interpret that as extra work when the actual energy cost is unchanged.
Wrist-based optical sensors add a second layer of error. They are reliable at steady pace, less so during intervals, and they can lose the signal when the band is loose or in cold weather. Published validation studies of wrist optical heart rate during running commonly find mean absolute errors in energy expenditure in the range of 15 to 30 percent.
Why your watch is more accurate on some runs than others
Heart-rate-based estimates are at their best during steady, moderate, aerobic running in stable conditions. They degrade during intervals, in extreme heat, and at very high intensities where the relationship between heart rate and oxygen consumption becomes non-linear.
Method 3: accelerometer and power-based estimation
Some devices estimate expenditure from motion data alone, or from running power calculated from accelerometry. These handle cadence and vertical oscillation well and do not inherit heart rate's sensitivity to caffeine and heat.
They struggle with anything the accelerometer cannot see. Running uphill with a heavy pack looks identical to the sensor as running uphill without one, though the energy costs differ substantially.
Method 4: gym cardio machines
The treadmill readout deserves particular scepticism. Machines frequently overestimate, sometimes by 20 percent or more, and the reasons are structural rather than accidental:
- They often use a default weight. If you have not entered yours, most machines assume something around 70 kg. If you weigh 55 kg, every number is inflated by roughly a quarter.
- They report gross rather than net. Included in the total is the resting metabolism you would have burned sitting on a bench — 60 to 90 kcal per hour.
- They cannot detect handrail use. Holding the rails, particularly on an incline, reduces energy cost significantly while the display carries on as though you were not.
- There is a commercial incentive. A machine displaying a larger number is more satisfying to use. Regulatory pressure to be accurate is minimal.
If you use gym equipment as your reference, mentally discount the readout by 15 to 20 percent and you will usually land closer to reality.
The gross versus net problem
This one accounts for more of the discrepancy between devices than most people realise.
Gross calories is total energy expended during the activity, including baseline metabolism. Net calories is the additional energy above what you would have burned at rest during the same period.
For a 70 kg person, resting metabolism runs at roughly 70 to 75 kcal per hour. Over a one-hour run, that is the entire difference between a gross figure of 620 and a net figure of 550 — about 11 percent, from definitions alone, with no disagreement about the underlying physiology.
Most devices report gross. Some apps report net. Very few state which. If you are tracking a calorie deficit, gross figures will systematically overstate what your running contributes.
What the research says about accuracy
Multiple validation studies have compared consumer wearables against indirect calorimetry. The consistent finding is that these devices track heart rate and step count reasonably well but estimate energy expenditure poorly.
| Method | Typical error | Direction |
|---|---|---|
| Indirect calorimetry (lab) | ±2–3% | Reference standard |
| MET formula | ±10–15% | No systematic bias |
| Chest strap + algorithm | ±10–20% | Varies by device |
| Wrist optical HR | ±15–30% | Often overestimates |
| Gym cardio machines | ±20–40% | Usually overestimates |
Two points are worth drawing out. First, no consumer method is precise, and treating any of these figures as exact is a mistake. Second, the errors are usually consistent within a given device — which is what makes trend tracking viable even when absolute numbers are not.
The practical implication: track trends, not totals
If your watch overestimates by 15 percent, it will overestimate by roughly 15 percent every run. The absolute number is wrong; the comparison between last week and this week is still informative.
This leads to a straightforward approach:
- Pick one source and stay with it. Comparing your watch against the treadmill against an app produces confusion and no additional information.
- Treat the number as relative. "This week I burned more than last week" is reliable. "I burned exactly 2,847 calories this week" is not.
- Assume the figure is high. If you are eating back exercise calories, eating back 70 to 80 percent of what your device reports is a more realistic starting point.
- Let the scale arbitrate. Body weight measured over four to six weeks tells you far more about your actual energy balance than any device estimate.
Want a transparent baseline?
Our calculator shows the MET formula it uses — no hidden algorithm.
Where the variation between individuals comes from
Even with perfect measurement, two people of identical weight running identical distances at identical paces will not burn identical calories. The differences come from:
Running economy. The oxygen cost of running at a given pace varies by 20 to 30 percent between individuals with similar fitness. Elite runners are notable for exceptional economy, not just high VO₂max, and economy improves measurably with training.
Body composition. Muscle is metabolically more active than fat. Two people at 75 kg with different body composition will differ in both resting metabolism and exercise expenditure.
Gait mechanics. Excess vertical oscillation — bouncing — is wasted work. It raises energy cost without producing forward motion.
Environmental conditions. Heat raises energy cost through thermoregulation. Wind resistance rises with the square of relative velocity. Soft surfaces such as sand or deep snow can add 20 to 30 percent to the cost of a run.
Training status. Well-trained runners are more efficient and burn slightly fewer calories at a given pace than untrained runners of the same weight. This means your calorie burn per kilometre will decline slowly as you improve, which is a marker of adaptation rather than a loss.
So which number should you use?
For most runners, a MET-based estimate is a sensible default. It is transparent, has no systematic bias in either direction, and is roughly as accurate for individuals as a mid-range wearable.
If you have a chest strap and a device that has learned your VO₂max over months of training, that estimate is probably slightly better for varied terrain and interval sessions.
If your only source is a gym machine, discount it by 15 to 20 percent.
And in every case, treat the figure as approximate. The precision implied by a number like 412 is not real. The useful information is the order of magnitude and the direction of change over time.
Accuracy ranges cited reflect published validation research comparing consumer devices against indirect calorimetry. Individual device performance varies by model and firmware version.