Setpoint

Are TDEE calculators accurate?

For a group, yes, reasonably. For you specifically, often not — and the design of the calculation explains why. It stacks two separate estimates on top of each other, and the second one is a guess you make yourself.

What the calculator is actually doing

Every TDEE calculator on the internet runs the same two steps.

Step one: predict resting metabolic rate

An equation — usually Mifflin-St Jeor, sometimes Harris-Benedict — takes your height, weight, age and sex and returns the calories you would burn lying still all day. These equations were built by measuring a few hundred people and fitting a line through them.

Step two: multiply by an activity factor

You choose a description of your life from a dropdown — sedentary, lightly active, moderately active, very active — and the calculator multiplies. That multiplier is typically somewhere between about 1.2 and 1.9.

Both steps carry error, and the second is far worse than the first.

Where step one goes wrong

A prediction equation describes the average person with your measurements, and you are not the average person with your measurements. Two people of identical height, weight, age and sex can have genuinely different resting metabolic rates, and the equation has no way to tell them apart because it was never given anything that would.

The main thing it is blind to is body composition. Muscle and fat do not cost the same to maintain, and the equation only sees total weight. So at the same 85 kg, a lifter and a sedentary person of the same height and age receive the identical answer, which cannot be right for both.

Typical error here is on the order of ten percent for most people, and larger at the extremes — very lean, very heavy, very muscular, very old. On a 1,700 calorie resting rate, ten percent is 170 calories before you have chosen anything from a dropdown.

Where step two goes badly wrong

This is the real problem, and it is structural. The activity multiplier is not measured, not derived, and not checked. You pick it, from four or five vague labels, describing your own behaviour.

Three things make it unreliable:

Why the errors do not cancel out

The two steps are multiplied together, not added. So an overestimate of resting rate gets amplified by whatever multiplier you chose, rather than being softened by it.

A worked example

Suppose the equation returns 1,700 calories but your true resting rate is 1,550 — a 9% overestimate, well within normal. You select “moderately active”, a multiplier of 1.55, when your real figure is closer to 1.4.

CalculatorReality
Resting rate1,7001,550
Activity multiplier× 1.55× 1.4
TDEE2,6352,170

Neither error was outlandish. Together they produced a figure 465 calories too high — and if you set a 500 calorie deficit against it, you are eating at roughly maintenance and wondering why nothing is happening.

Why it feels like the app is lying to you

This is the practical consequence, and it is where most people give up on tracking altogether.

You do everything right. You weigh your food, you hit the target every day, you stay honest for six weeks. And the scale barely moves. The obvious conclusions are that you are doing it wrong, that you cannot be trusted to log accurately, or that something is broken about your metabolism.

Usually none of those are true. The target was set from a number that was several hundred calories too high on day one, and nothing in the app ever went back to check.

What to do instead

The information you need is already in the data you are producing. If you know what you ate and you know how your weight trended, you can work backwards to what you burned, because energy balance is arithmetic rather than prediction.

That is what adaptive TDEE does, and it is the approach Setpoint takes. It costs you a fortnight of consistent logging before it can say anything useful, which is a genuine drawback. What you get back is a number that describes you rather than a population.

In fairness to the calculators

  • A starting point has to come from somewhere. Before you have logged anything there is no data to measure from, so an equation is the correct tool. Setpoint uses one too, on day one.
  • They are unbiased on average. Across a population the errors do cancel. That is what they were built for. The problem is applying a population tool to one person and then never revisiting it.
  • Direction usually survives. Even a wrong number, held consistently, will tell you whether you are heading up or down. It just will not tell you how fast, or what to change.

The failure is not the equation. It is treating its output as a measurement and never checking it against what actually happened.

The short version

A TDEE calculator multiplies an estimate with about ten percent error by a multiplier you guessed from a four-item dropdown. It is a reasonable place to start and a poor place to stay. After a fortnight of logging you have something better available, and the only question is whether your app bothers to use it.

See how Setpoint works