Why Every TDEE Calculator Is an Estimate

A calculator returns the average of everyone who shares your four inputs#

Type your height, weight, age and sex into a TDEE calculator and it answers a question you did not ask. It reports roughly what people with those four numbers burn, on average. Your own daily expenditure is one draw from the spread around that average, and no arrangement of the same four inputs can tell you where in the spread you sit. That is why the output is an estimate in the strict statistical sense, not in the hedging sense.

You can see the honesty of the field in what its best equation ships with. When a consortium pooled 6,497 doubly labelled water measurements in people aged 4 to 96 — the largest such dataset assembled — and fitted a prediction equation to it, they published the equation together with a 95 percent predictive interval, the band inside which an individual's real expenditure is expected to fall. Tested against an independent 598-person validation set, the model correlated well (R² = 0.67) and 94.6 percent of measurements landed inside that band, but the average absolute gap between prediction and measurement was 11.2 percent¹. That is the state of the art, built on isotope measurements rather than self-report, with more data than any consumer tool will ever see. It still hands you a band. TDEE explained covers what the number is made of; this article is about why the band cannot be closed.

The 26 percent that has no name#

The obvious explanation — that human metabolism is erratic, that your burn bounces around day to day — turns out to be wrong, and the study that shows it is the most useful thing in this literature.

Researchers took 150 adults in northeast Scotland, measured basal metabolic rate and body composition, and then partitioned the variance in BMR into within-person and between-person components. Only 2 percent of the observed variability was within-subject, and 0.5 percent of that was analytic error from the equipment. Essentially all the variation was between people. Of that between-person variance, fat-free mass explained 63 percent, fat mass 6 percent, and age 2 percent. Sex and bone mineral content were not significant. Twenty-six percent remained unexplained — and it was not accounted for by circulating leptin or triiodothyronine; thyroxine explained a quarter of the residual in men and nothing in women².

Your resting metabolism barely moves — 2 percent of its variation is within-person. What is unstable is not your body. It is the resolution of any map drawn of it.

Read that as an upper bound on what a calculator could ever do. Johnstone's team measured fat-free mass directly and still left a quarter of the differences between people unexplained by anything they could name. A consumer calculator does not measure fat-free mass at all — it infers it from height, weight, age and sex, and then multiplies the result by a self-assigned activity category. It is working several steps back from the data that produced the 63 percent, chasing a residual that the people with the DXA scanner could not chase either.

Two uncertain numbers, multiplied together#

The second reason the output is a band is arithmetic, and it is easy to underrate because both terms look precise on screen.

A TDEE is a product: an estimated resting rate times an activity factor. Suppose the resting estimate is 1,500 kcal and you allow it a modest individual error of ±10 percent, which is generous relative to what validation studies report. Suppose you are also genuinely unsure whether your life is a 1.4 or a 1.6 — two neighbouring rows on the same dropdown. Run the corners:

	Activity factor 1.4	Activity factor 1.6
RMR 1,350 (estimate −10%)	1,890 kcal	2,160 kcal
RMR 1,500 (point estimate)	2,100 kcal	2,400 kcal
RMR 1,650 (estimate +10%)	2,310 kcal	2,640 kcal

That table is our arithmetic, not a published result, and it is deliberately conservative on both inputs. It still spans 750 calories. Multiplying two uncertain quantities does not average their uncertainties; it compounds them, and the second one is the worse offender because nobody validated it — you chose it about yourself, from a menu of adjectives, which is the subject of activity multipliers explained. A single four-digit output is the geometric centre of that table with the table deleted.

The quantity itself is not a constant#

There is a third problem, subtler than the other two: a flawless measurement taken today would still be describing something that will not hold still.

Daily expenditure includes spontaneous movement that responds to how much you eat and how much you train, and it shifts as body composition changes and as sustained weight loss trims expenditure below what body size predicts. None of that is speculative and none of it is this article's to prove — what NEAT is and why it matters has the variability, and the post-weight-loss adaptation is a known, modest, well-documented effect. The point for a calculator is narrower. A calculator estimates a parameter; a parameter that drifts with behaviour has a value only relative to a period of time. "Your TDEE" in the sense of a fixed personal constant is a category that does not exist, which is a more fundamental limit than any equation's error bar.

Even the reference standard has a band#

If prediction equations are estimates, what would a measurement look like? Doubly labelled water — drinking water enriched with heavy hydrogen and heavy oxygen and tracking how fast each washes out — is the reference method for free-living expenditure, and it is worth knowing that it too reports with uncertainty.

Sixteen subjects spent seven consecutive days in a whole-room indirect calorimeter with doubly labelled water measured concurrently, so the isotope result could be checked against a physical measurement of the same days. Fitting the isotope elimination curves from daily urine samples gave a precision of 4.5 percent and an accuracy of −0.5 percent; the conventional two-point sampling protocol gave 6.0 percent and −3.0 percent on the same subjects³. The choice of sampling protocol moved the answer by about two and a half percentage points of accuracy.

So the hierarchy is not "estimates versus the truth." It is a several-percent band at the top, an 11 percent average deviation for the best population equation, and something wider than that by the time a free tool has inferred your lean mass from your height and asked you to describe your own week. Nothing in the chain is a point value. The intake side of the ledger behaves the same way, for the same reason, in why calorie counts are ranges.

Using a number that is really a band#

None of this makes the calculation useless. It makes it a starting hypothesis with a known shape, which is a genuinely useful object as long as you treat it like one.

Three practical consequences. Read the output as a range — take the number, allow it something like 10 to 15 percent either way, and notice that this range is wide enough to contain both "losing slowly" and "gaining slowly" at the same intake. When an input is ambiguous, take the lower branch, because an inflated estimate produces an invisible failure: a deficit that is actually maintenance, contradicted by nothing for weeks. And replace the estimate as soon as you can afford to, because a fortnight or two of steady logging read against a weekly weight average stops being a claim about the population and becomes a claim about you, by the method in finding your maintenance calories.

The calculator's job is to get you close enough that your own data can do the rest. Judged on that job it works fine. Judged as an oracle it was never going to, and the people who built the best version of it said so in the paper, in the form of an interval printed next to the answer. How far off it lands in practice, and for whom, is measured in how accurate TDEE calculators really are.

FAQ#

If my TDEE is only an estimate, is it worth calculating at all?#

Yes, as a starting point you intend to overwrite. The estimate tells you whether your maintenance is nearer 1,800 or 2,800, which is the difference that matters when you are setting a first target from nothing. What it cannot do is settle a 200-calorie question, because its own error is larger than that: the best equation in the field, fitted to 6,497 isotope measurements, deviated from measured expenditure by an average of 11.2 percent in validation.

How wide should I treat my TDEE range as being?#

Roughly 10 to 15 percent either side of the number is a defensible working assumption for a general adult, and wider if you are outside the populations these equations were built on. Worked through as a product rather than a single figure, a resting estimate allowed ±10 percent and an activity factor of 1.4-versus-1.6 spans about 750 calories on a 1,500-calorie resting rate. That is our arithmetic, not a measured interval, but the shape is right: the uncertainty in the two terms multiplies rather than averaging out.

Why do two TDEE calculators give me different numbers for the same person?#

Because they are running different equations and, more importantly, attaching different multipliers to the same words. The resting-rate formulas disagree with each other modestly — a few percent — while two tools can assign "moderately active" values of 1.45 and 1.55, which on a 1,500-calorie resting rate is a 150-calorie gap created entirely by wording. Neither tool is wrong; both are reporting a population average with a band around it, and only one of them is showing you the band.

Sources#

This article was researched and drafted with AI assistance and reviewed for accuracy by the BurnWeek team. It is general information, not medical advice. How we research and correct our articles →