Why your watch, the gym machine and the app never agree on calories

You finish a spin class sweating and proud. The bike console says 620 calories. Your Apple Watch says 410. The app you logged it in splits the difference at 520. Three numbers, one workout, a 200 calorie spread, and a nagging question about which one earned you dinner.

The answer is that none of them knows. They are all estimating a quantity that is genuinely hard to measure, and they estimate it from different starting points.

Nobody in the room measured your calories

Real energy expenditure is measured with a metabolic cart or a respiratory chamber, tracking the oxygen you consume and the carbon dioxide you produce. That equipment lives in labs. Everything on your wrist, on the machine and in the app is a model standing in for it.

The models disagree because they know different things about you and the work you did. A Stanford study that put seven wrist devices against lab-grade metabolic measurement found they were excellent at heart rate, within a few percent, but poor at energy, with error rates that made the calorie readout the least trustworthy number they produced.

Three estimators, three sets of inputs

Line up what each source actually feeds into its guess and the spread stops being mysterious.

The machine often runs highest because it knows the least: many consoles apply a formula based on their resistance and your weight, with no age, sex or actual heart rate. The app applies a MET value, a table of typical intensities from the Compendium of Physical Activities, to whatever you logged, which is fine on average and blind to how hard you personally worked. The watch usually has the most to go on, which is why it is the one to trust if you must choose.

Where the estimates go most wrong

Two situations blow the numbers up. The first is strength training and stop-start work, where heart rate rises from effort and tension but your actual oxygen cost is lower than that heart rate implies, so wrist devices tend to overstate lifting. The second is anything the app has no good template for, where a generic MET value is doing all the work.

Steady cardio is the friendly case. On a continuous run or ride, heart rate tracks effort closely and the estimates tighten. This is the same fault line as VO2 max: heart-rate-driven models are at their best in steady aerobic work and shakiest everywhere else.

What to do with this

1. Pick one source and stick to it. Your own watch, read against its own history, is the only fair comparison. Mixing the console and the app just adds noise.
2. Trust the device that knows you. Between a bare machine and a watch with your profile and heart rate, the watch is the better guess.
3. Do not eat back the full number. For weight goals, treat the burn as an overestimate, follow your weight trend over weeks, and adjust from reality.
4. Expect strength days to be the least accurate. If a lifting session shows a big burn, be skeptical.

Because calories are a personal estimate rather than a shared unit, they are the wrong thing to compete on (the fitness challenge ideas that last use active minutes, distance or streaks instead), which is why Fit, the cross-brand fitness competition app, does not rank a crew on them. Fit compares the numbers that survive the jump between an Apple Watch, a Garmin and a Fitbit, like active minutes and distance, and leaves the calorie estimate on your own wrist where it can still guide you. Where Fit loses: it is not a nutrition or calorie-tracking tool, so if managing intake is your goal, a dedicated food and energy app will serve you better.