TDEE calculators are all wrong about you

21 August 2026·7 min read

Type “TDEE calculator” into a search engine and you will find hundreds of them. Enter the same height, weight, age and sex into five, and you will get five different numbers, sometimes spread across four hundred calories or more. They cannot all be right. In an important sense, none of them is.

That is not a scandal. It is what the formulas were built to do, and once you understand what they are actually computing, the fix becomes obvious.

What a TDEE calculator actually does

Total daily energy expenditure is everything you burn in a day: the cost of simply being alive, the cost of digesting your food, the cost of deliberate exercise, and the cost of all the incidental movement in between. Nearly every online calculator estimates the first of those with a regression formula, then multiplies the result to account for the rest.

The formula is usually Mifflin St Jeor, published in 1990 and still the common default. Sometimes it is Harris Benedict, originally from 1919 and revised in 1984, or Katch McArdle if the calculator asks you for a body fat percentage. Mifflin St Jeor looks like this:

Then the calculator multiplies that resting figure by an activity factor you selected from a dropdown. The conventional set runs roughly 1.2 for sedentary, 1.375 for lightly active, 1.55 for moderately active, 1.725 for very active.

Watch the error appear, with actual numbers

Take a 35 year old man, 82 kg, 178 cm. Run him through the three formulas above and you get three different resting figures before the activity guess even enters the picture.

An 80 calorie spread at rest. Annoying, but survivable. Now apply the activity multiplier to the Mifflin St Jeor figure and watch the spread explode.

Six hundred and seventeen calories separate the top and bottom of the range that this one person could plausibly have selected. One rung of the dropdown is worth roughly 300 calories a day. And nothing on that dropdown is defined precisely enough for you to know which rung is yours: is a desk job plus three gym sessions “lightly” or “moderately” active? Ask ten people and you will not get one answer.

So two errors stack:

• Individual spread around the average. Even the resting-burn stage is a prediction of what the average person of your dimensions burns. Real people scatter around that average, commonly by a couple of hundred calories a day in either direction at identical size. The formula cannot see your muscle mass, your thyroid, your genetics or your history, because it was never given them.
• The dropdown you guessed. It is the largest term in the calculation and the least well specified. People are famously poor at classifying their own activity level, and the classification moves the answer by hundreds of calories.

The number you get is therefore a population average multiplied by a guess. It is a perfectly reasonable starting point. It is a poor thing to run months of dieting on, because if it lands 300 calories high, your “deficit” is a maintenance diet with extra paperwork, and you will spend six weeks concluding that your body is broken.

In fairness: when a calculator is good enough

Honesty cuts both ways here. For a first week, a formula is fine. You need some starting target, and the formula’s error does not matter yet, because you are about to find out what it is. If your weight does roughly what the formula implied it would, the formula was close enough for you. If it does not, you have learned something no calculator could have told you.

The mistake is not using a formula. The mistake is never checking it against what actually happened, and then treating the unchecked number as a fact about your body.

The version that is actually yours

Your real TDEE is sitting in data you can collect yourself: what you ate, and what your weight trend did while you ate it. Energy balance is the one piece of arithmetic in this field that is not an estimate.

Take our 35 year old. He logs for a fortnight, averages 2,150 calories a day, and his trend weight falls 0.55 kg. That loss represents about 4,235 calories, or 303 a day across the window. His measured TDEE is therefore about 2,453: close to the “lightly active” guess, and nearly 600 below the “very active” one he might just as easily have ticked.

Notice what that 2,453 already contains. His workouts. His commute. His fidgeting. His desk job. His bad sleep in week two. His particular metabolism. He did not have to estimate a single one of them, because every one of them showed up in the weight change. That is the entire appeal of the method: the things that are hardest to estimate are exactly the things it never asks you to estimate.

The full method, including the four ways people get it wrong, is its own post, and the ways to get it wrong are the part worth reading.

Why this matters more the longer you diet

A static calculator number has a second flaw that only shows up later: it goes stale.

As you lose weight your burn falls. A smaller body has less tissue to maintain and less mass to move, so the same walk costs fewer calories than it did. On top of that, spontaneous movement tends to drift down during a sustained deficit: you take the stairs less, fidget less, sit more. Neither of these is damage, and neither is mysterious, but together they mean the number that was right in January is wrong by April.

The formula only ever knew your starting weight. You can of course re-run it at your new weight, but you will just get a fresh population average multiplied by the same guess. A measured TDEE, recomputed on a rolling window of recent data, tracks the drift on its own. The target eases down because the measurement moved, not because you failed at anything, and that distinction matters more than it sounds when you are eight weeks in and looking for a reason to quit.

What Cut does

Cut is this article as an app. It runs the measurement continuously against your own logs, sets the daily target from the result, and recomputes as your body changes so the target follows you down. For the fortnight before it has enough data to measure honestly it uses a standard formula, and it labels that clearly as a fallback rather than dressing an estimate up as a measurement.