BMI, BMR and TDEE: the error that piles up at each link of the chain

The BMI → BMR → TDEE chain and the error that piles up

The three most popular health-calculator indicators are not measurements, they are chained estimates, and the difference matters. A measurement looks at your body (a scale weighs you; indirect calorimetry measures how much oxygen you burn). An estimate looks at a table or an equation built from other people and assumes you behave like their average. BMI estimates weight-related risk from a geometric ratio; BMR estimates resting expenditure from a regression; TDEE estimates total expenditure by multiplying BMR by a factor. Each step inherits the error of the one before it and adds its own.

BMI (Body Mass Index)

The ratio of weight to height squared, used to screen weight ranges across populations. It does not measure fat.

BMR (Basal Metabolic Rate)

The energy the body spends at complete rest for vital functions. Estimated by regression from weight, height, age and sex.

TDEE (Total Daily Energy Expenditure)

BMR multiplied by an activity factor. It is the whole-day expenditure, including movement, digestion and exercise.

The point here is not to dismiss the calculators, they are an excellent starting point and teach the logic of weight and energy. The point is to show how wide the uncertainty band really is, so you do not treat a population number as an individual diagnosis. Throughout, we use the same coefficients that this site’s BMR calculator and TDEE calculator apply, so the guide and the tool tell exactly the same story.

BMI: a population index applied to one person

BMI is weight (kg) divided by height (m) squared. A person of 70 kg and 1.75 m has a BMI of 70 ÷ (1.75 × 1.75) ≈ 22.9. The ratio was born with the Belgian mathematician Adolphe Quetelet in the 1830s, as part of his project to describe the "average man", that is, it was a population statistic from the very start. The modern name came much later: in 1972, Ancel Keys and colleagues, in the paper "Indices of relative weight and obesity" (Journal of Chronic Diseases), compared several indices in 7,424 men across five countries and concluded that weight ÷ height² was the best of them for population use. That paper is what coined the term body mass index.

The World Health Organization consolidated the adult reference ranges in Technical Report 894 (2000). Notice that the overweight boundary (25) and the obesity boundary (30) are round cut-offs, chosen by epidemiological association with risk, not biological constants. The values below are exactly the ones the BMI calculator uses to classify the result:

Those boundaries do not apply the same way everywhere. In 2004, a WHO expert consultation, published in the Lancet, reviewed data from Asian populations and noted that the risk of type 2 diabetes and cardiovascular disease already rises at BMIs below the cut-off of 25. The consultation kept the international classification but flagged public-health action points at 23.0 and 27.5 kg/m² for these populations, a reminder that the same BMI means different body-fat levels depending on ancestry, sex and age. If you want the number BMI does not give, body composition, read the BMI vs body-fat percentage guide.

1. 1832Quetelet creates the ratio

   Adolphe Quetelet proposes weight ÷ height² while studying the "average man", a tool of population statistics, not of the clinic.

2. 1972Keys coins "body mass index"

   Keys et al. compare indices in 7,424 men, single out weight ÷ height² as the best, and name the "body mass index".

3. 2000WHO fixes the ranges

   Technical Report 894 standardizes the overweight (≥25) and obesity-class (30/35/40) ranges used worldwide.

4. 2004Asian cut-offs in the Lancet

   The WHO consultation flags action points at 23.0 and 27.5 for Asian populations, recognizing that BMI-related risk varies between peoples.

BMR: a regression fitted on one specific sample

Basal metabolic rate is the energy you would burn lying down, awake, fasted and in a neutral temperature, just to keep the heart, brain, kidneys and everything else running. Measuring it for real requires indirect calorimetry in a lab. Since almost no one has access to that, calculators use regression equations: someone measured the BMR of hundreds of people and fitted a line that predicts the value from weight, height, age and sex. The key word is "sample": the coefficients carry the traits of whoever was measured.

The first famous equation is Harris and Benedict’s, from 1919, built on a few dozen healthy young adults of the early twentieth century. Roza and Shizgal reassessed it in 1984 with more data and modern statistics, producing the "revised Harris-Benedict". In 1990, Mifflin and St Jeor published a new equation from 498 adults, today considered the most accurate for non-obese people. This site’s BMR calculator uses exactly the original Mifflin-St Jeor coefficients:

The major review by Frankenfield, Roth-Yousey and Compher (2005, Journal of the American Dietetic Association) compared the most-used equations against BMR measured by calorimetry and concluded that Mifflin-St Jeor was the most reliable: it predicted the value within ±10% of measured in more people, obese and non-obese, than any competitor, and with the narrowest error range. Note what "within ±10% for most people" means in practice: even the best equation is off by 10% or more in a meaningful fraction of individuals, and Harris-Benedict tends to overestimate, especially in heavier bodies. The chart below makes that divergence visible, for one hypothetical man of 30 years and 1.75 m, what each equation predicts as weight rises.

There is a third family of equations that changes the logic: those that use lean mass. The Katch-McArdle formula estimates BMR as 370 + 21.6 × lean mass (kg). Because it starts from the mass that actually "burns" energy, muscle and organs, not fat, it drops even the sex variable: two bodies with the same lean mass have similar BMR, whether the person is a man or a woman. That is why Katch-McArdle wins when body composition is known (via bioimpedance, skinfolds or DEXA): it swaps a population average for a fact about you. The downside is obvious, without a body-fat measurement it cannot be computed, and a guessed body-fat number becomes one more link of error.

The activity factor: a categorical guess

If BMR is already an estimate, TDEE stacks another one on top. The arithmetic is simple: TDEE = BMR × activity factor. The trouble lives in the factor. You pick one of five boxes ("sedentary", "light", "moderate"…), and each box becomes a fixed number. But real movement expenditure is continuous, not five steps. Two "moderately active" people can differ by hundreds of kcal. The factor is a pocket convention, not a measurement.

Those 1.2-to-1.9 numbers are a gym tradition, they are not the official physical activity level (PAL) bands from the literature. The FAO/WHO/UNU consultation (2001, Human energy requirements) defines PAL as total expenditure divided by BMR and sorts it into three bands: a sedentary or light lifestyle between 1.40 and 1.69; active or moderately active between 1.70 and 1.99; vigorous between 2.00 and 2.40. In other words, the calculators’ "1.2 sedentary" sits below the 1.40 floor the FAO treats as compatible with living. It is not a grave error, it is a different convention, but it shows how the same label can become quite different numbers depending on who built the table.

The good news is that the TDEE calculator itself does not fake certainty: besides the central value, it returns an estimate 10% lower and 10% higher, acknowledging the factor’s uncertainty. Try it here, change the activity level and watch the interval shift:

How much error piles up: two worked examples

Enough theory. Let’s take two real people and run the same person through two BMR equations, apply the activity factor, then add the uncertainty the tool itself assumes. The goal is to see, in kcal, the width of the band hiding behind "your TDEE is X".

1. Example 1, woman, 32, 68 kg, 165 cm, moderate activity (×1.55)Mifflin-St Jeor: 9.99×68 + 6.25×165 − 4.92×32 − 161 = 679.3 + 1,031.25 − 157.4 − 161 ≈ 1,392 kcal. Revised Harris-Benedict: 447.593 + 9.247×68 + 3.098×165 − 4.330×32 ≈ 1,449 kcal. Swapping the equation alone shifts BMR by 57 kcal (~4%). Multiplying by 1.55: TDEE of 1,392×1.55 ≈ 2,158 kcal (Mifflin) versus 1,449×1.55 ≈ 2,246 kcal (Harris-Benedict). Now apply the ±10% the tool assumes: the low end of Mifflin is 2,158×0.9 ≈ 1,942 kcal and the high end of Harris-Benedict is 2,246×1.1 ≈ 2,471 kcal. The honest band for "her TDEE" runs from ~1,940 to ~2,470 kcal, about 530 kcal wide, more than a quarter of the central value.
2. Example 2, man, 40, 90 kg, 180 cm, light activity (×1.375)Mifflin-St Jeor: 9.99×90 + 6.25×180 − 4.92×40 + 5 = 899.1 + 1,125 − 196.8 + 5 ≈ 1,832 kcal. Revised Harris-Benedict: 88.362 + 13.397×90 + 4.799×180 − 5.677×40 ≈ 1,931 kcal. Here, in a heavier body, Harris-Benedict overestimates by 99 kcal (~5.4%), exactly as the chart predicted. TDEE: 1,832×1.375 ≈ 2,519 kcal (Mifflin) versus 1,931×1.375 ≈ 2,655 kcal (Harris-Benedict). With the ±10%, the band runs from 2,519×0.9 ≈ 2,267 to 2,655×1.1 ≈ 2,921 kcal, about 650 kcal wide. And there is an even bigger lever: if he ticked "moderate" (1.55) instead of "light", his Mifflin TDEE would jump from 2,519 to 1,832×1.55 ≈ 2,840 kcal. Getting the activity level wrong by a single band moves ~320 kcal, more than the difference between the two equations.

The moral is not "calculators are useless". It is that the final number carries, added together, the BMR-equation error (some 5% between two reasonable choices, plus up to ±10% of individual accuracy) and the activity-factor error (which alone is worth hundreds of kcal). So the best strategy is to use the range as a starting point, pick the lower estimate as your initial target, and adjust to reality: if the scale does not move over two or three weeks eating "TDEE − deficit", your true TDEE was at the other end of the band. To turn kcal into amounts of food, see the macro calculator, and to size up each workout’s cost, read how many calories each activity burns.

Limits, water and safe use

Every estimate in this guide ignores things only your body knows: clinical history, medications, thyroid, sleep, recent diet, genetics. Formulas use population averages, they do not know you. Use the results as a reference range and track trends over weeks, not a single number on a single day.

• Compare trends (weeks/months), not isolated measurements.
• Cross-check BMI with other signals: waist circumference, body composition, lab tests.
• Start at the low end of the TDEE range and adjust slowly by how your weight responds.
• If you take medication that affects metabolism or heart rate, the formulas matter less, confirm with a professional.

Water intake usually shows up on the same screen and suffers the same "single number" problem. The popular rule of 35 ml per kg of body weight is one end of a range, not a law: this site’s daily water calculator uses 30 to 35 ml/kg and cross-checks it against the official references. The US Institute of Medicine (2004) sets an adequate intake of total water (drinks plus food) of about 3.7 litres/day for men and 2.7 litres/day for women; Europe’s EFSA (2010) recommends lower values, 2.5 and 2.0 litres/day. Since roughly 20% of water comes from food, the target for what you drink is lower than the total. And all of it varies with climate, exercise and sweat, one more case where the range beats the point.

Training heart rate, which many people use alongside TDEE to gauge expenditure, is the same kind of estimate. If you want to see why "220 − age" is also off by a dozen beats, read the guide on heart rate zones. It is the same principle as this guide, applied to the heart: a population value is not an individual diagnosis.