The problem with a formula
Open any calorie tracker and it will tell you your maintenance calories before you have logged a single meal. It gets there with a formula: your age, your height, your weight, and a dropdown where you guess how active you are.
The number looks authoritative. It is a population average with your guess stapled to the front of it.
Then reality shows up. You eat what the app told you to eat, the scale does not move, and the app has no way of being wrong. The only variable left is you. So the deficit you were promised quietly turns into a question about whether you are logging honestly.
The arithmetic
Burnin does not guess. It waits for evidence, then does subtraction.
If you are losing weight, you burned more than you ate, and the difference is sitting in the tissue you lost.
TDEE = average daily intake − (daily weight change rate × 7700 kcal/kg)7700 kcal is the accepted energy content of a kilogram of body mass.
None of that is clever. It is the energy balance identity, it has been in textbooks for decades, and the trackers worth using already work this way. Anyone selling the equation itself as an innovation is selling you something.
The equation is the easy part. Feeding it numbers worth trusting is not.
Where the work actually is
Both inputs are noisy. Your weight on any given morning is mostly water, glycogen and whatever has not finished moving through you. Your logging is honest and incomplete. Run the equation on raw daily figures and your metabolic rate appears to swing by 800 calories between Tuesday and Wednesday.
So the subtraction is not really the job. Deciding which numbers deserve to go into it is.
It uses the longest window your data can support. There are four: 7, 14, 21 and 28 days. Each one has a threshold it has to clear before it is allowed to run, and the 28 day window needs 16 days of logged calories and 9 weigh-ins. More evidence means a steadier answer, so it reaches for the longest window it can justify and steps down when the data is not there.
It fits a line rather than subtracting endpoints. First weight minus last weight is hostage to two arbitrary mornings. A regression across every weigh-in uses all of them.
It discards weigh-ins that cannot be true. Two kilograms overnight is a scale on a carpet, a different time of day, or last night’s dinner. Those go before the line is fitted.
Discarding one makes it less confident, not more. A rejected weigh-in stops counting toward the data requirement, so the interval widens. Cleaning your data should never make an estimate look sturdier than the evidence underneath it.
It refuses to return nonsense. Estimates are clamped between 1000 and 5000 kcal. Outside that range something is wrong with the inputs, and a wrong number delivered confidently does more damage than no number at all.
How sure it is
Every estimate carries an interval, built from the standard error of your weight trend and the spread in your intake. Keep logging and it narrows.
Burnin shows that as a confidence tier instead of burying it. Two weeks in, you should be told the estimate is provisional. Six weeks in, you have earned a tighter number. A single figure with no error bar implies a precision that does not exist, and you are going to make decisions with it.
What this does not claim
There is no validation study behind any of this. No comparison against indirect calorimetry, no accuracy figure to quote at you.
What there is: a method you can read, arithmetic you can check, and every input sitting in your own log. If the estimate looks wrong to you, you can go and find out why.
That is a smaller claim than validated. It is also the only one that is currently true.