How to measure your maintenance calories

17 August 2026·7 min read

Every calorie calculator asks the same four things: height, weight, age, sex. Then it multiplies by an “activity level” you picked from a dropdown. Out comes a confident number like 2,340, usually with no indication of how much confidence it deserves.

That number is an average of people who share your measurements. You are not the average of those people, and the activity multiplier, the part you guessed at, swings the answer by hundreds of calories on its own. The arithmetic on how far off it runs is worth seeing once.

The better news is that you do not need the formula, because you are running a live experiment on yourself every single day.

The measurement that beats the estimate

Energy balance is not a theory you have to trust. It is bookkeeping. If you ate a certain amount and your body mass changed by a certain amount, the difference went somewhere, and rearranging that gives you your burn:

Worked through: over 14 days you averaged 2,000 calories a day, and your trend weight fell 0.8 kg. That 0.8 kg represents about 6,160 calories, or 440 a day across the window. Your actual burn is therefore about 2,440, not the 2,100 the calculator told you. Every day you have been eating at what you believed was a 100 calorie deficit, you were in fact 440 down.

Here are three more people, to show how differently it can land.

The middle row is the interesting one. That person is eating 1,750 calories a day and barely moving, and the internet will tell them their metabolism is broken. It is not: they burn about 1,833, which is entirely ordinary for a smaller person, and their deficit is simply tiny. Knowing that turns an emotional problem into an arithmetic one, and an arithmetic one has a solution.

Notice what all three figures already include: workouts, fidgeting, jobs, terrible sleep, individual metabolism. None of it had to be estimated, because all of it showed up in the weight change.

Doing it properly

The method is simple. The ways to get it wrong are the interesting part, and every one of them produces a number that looks perfectly plausible.

• Use trend weight, not raw weigh-ins. Both ends of your window are single mornings, and single mornings carry a kilo of water noise. Over a fortnight, one kilo of endpoint error maps onto roughly 550 calories a day of error in your answer, which is more than most people’s whole deficit. Smooth first. (Why the scale lies most days.)
• Use at least two weeks. The same noise gets divided by more days as the window grows, so a longer window is a more precise one. Seven days halves your precision compared with fourteen. Anything under a week is dominated by whatever your water is doing.
• Log honestly, including the bad days. This is the one that quietly ruins the calculation. If you under-record by 300 calories a day, the arithmetic concludes you burn 300 less than you do, and it hands you a target 300 lower than it should be. You then have to eat at that lower target while still under-recording, which makes the next fortnight harder for no reason at all. The calculation is only as honest as its worst day.
• Do not count a day you forgot to log as zero calories. A missing day is missing data, not a fast. Averaging a zero into fourteen days at 2,000 drops your apparent intake by about 143 a day and inflates your apparent burn by the same amount. Either log the day approximately or leave it out of the average entirely.
• Try not to change everything mid window. Starting a new training block, a new job or a new sleep pattern halfway through means you have measured a person who no longer exists. It is not fatal, but start a fresh window afterwards.
• Recalculate as you go. Your burn falls as you get lighter, because a smaller body costs less to run and less to move. A figure measured 15 kg ago is describing someone else.

How precise is this, really?

Precise enough to act on, and worth being honest about. Two sources of error survive even a careful fortnight.

The first is your logging. Portion estimation is genuinely difficult, and almost everyone under-records to some degree rather than over-records. This does not invalidate the method, because the error is reasonably consistent from week to week: if you log the same way each fortnight, the change in your measured burn is still meaningful even if the absolute figure sits a little low.

The second is that not all weight change is fat. Early in a diet some of it is glycogen and water, which are far cheaper per kilogram than 7,700 calories, so the first window tends to overstate your burn. It settles down after the first two or three weeks. This is one more reason to keep recomputing rather than treating your first result as final.

What you end up with is a number that gets steadily better the longer you keep feeding it, which is the opposite of how a formula ages.

Why not just use your watch?

Because you would be importing its error into a measurement that did not have any.

Wrist-based active-energy figures are estimates built from heart rate and movement patterns. Heart rate tracking on modern wearables is generally good. Turning heart rate into calories is a much harder problem, and validation work has repeatedly found energy expenditure to be the least reliable output these devices produce.

More to the point, it is unnecessary here. The weight-change method already contains every calorie you burned, whether or not any device noticed you burning it. Adding your watch’s active-energy figure on top double counts the exercise you already accounted for, and replaces a measurement with a guess.

Movement data is still useful, just not as the source of your target. Use it as context for why a particular week was unusual: a fortnight with three long rides in it is a fortnight you would expect to look different.

What Cut does with it

Cut runs this calculation continuously on a rolling window of your own data, and sets your daily target from the result. The deficit is sized as a percentage of your bodyweight rather than as a fixed number, so it eases off as you get closer to your goal instead of staying brutally constant while you get smaller. There is a hard floor the target will never go below, whatever the arithmetic suggests.

Until it has enough of your data to measure honestly, which is about a fortnight of logging with weigh-ins spread across it, it uses the standard formula and labels it as a fallback. An estimate presented as a measurement is the exact thing this whole approach exists to avoid.