How to Find Your Real Maintenance Calories (Not a Formula's Guess)
The short answer
Your real maintenance is the average number of calories you ate over a few weeks, adjusted by how much your trend weight moved in that time — roughly 7,700 kcal per kilogram gained or lost (about 3,500 kcal per pound). A calculator can only give you the average for people shaped like you, and the spread around that average is wide enough that it is a starting guess, not an answer.
So: log honestly for at least three weeks, weigh often enough to get a trend line rather than a number, and do one piece of arithmetic at the end. That number is worth more than any formula, because it is measured on the only person you are trying to feed.
What a TDEE calculator is actually doing
Every online maintenance calculator runs the same shape of sum. It takes your height, weight, age and sex, puts them through an equation fitted to a study population — Mifflin-St Jeor and Harris-Benedict are the usual ones — and gets a resting metabolic rate. Then it multiplies that by an activity factor you chose from a dropdown: sedentary, lightly active, moderately active, and so on.
Two things are worth noticing about that. The first is that the equation was fitted to a group, so it describes the middle of that group well and the edges badly. Some people's resting rate sits meaningfully above what the equation predicts for them, and some sit below, and nothing about your height or age tells you which you are.
The second is the multiplier. You are being asked to compress everything you do in a day — your job, your commute, your children, your fidgeting, whether you take the stairs — into one of five words. The gap between "lightly active" and "moderately active" can be several hundred calories, and you are guessing which side of it you fall on.
Why individuals deviate from the formula
Non-exercise movement varies enormously. All the walking, standing, shifting and gesturing you do without calling it exercise is one of the largest sources of difference between two people of the same size — and it varies within one person too, week to week. A fortnight with a deadline and a fortnight with a garden to dig are not the same fortnight, and neither is captured by a dropdown.
Body composition is invisible to the equation. Two people who weigh 80 kg but carry very different amounts of muscle will have different resting rates, and the standard formulas take only total weight as input. Formulas that use lean mass exist, but they need a body-fat figure that most people can only estimate.
A long deficit changes the target. If you have been eating below maintenance for months, your maintenance now is not what it was at the start. Some of that is simply being a smaller body; some is reduced spontaneous movement, which tends to drop quietly when energy is short. The direction is predictable. The size is not, which is exactly why measuring beats predicting.
The measurement, step by step
Pick a window of three to four weeks. Shorter than three and normal fluid movement will swamp the signal.
Log everything, every day, the same way. Precision matters less than consistency here — more on that below.
Weigh yourself most mornings, same conditions: after the loo, before food, similar clothing. You are collecting a cloud of points, not a verdict.
Compare averages, not days. Take the average weight of the first week and the average of the last week, or use a weighted-average trend line if your app draws one.
Do the sum.
Worked example. Over 21 days you averaged 2,400 kcal a day. Your first-week average weight was 82.4 kg; your last-week average was 81.8 kg. That is 0.6 kg lost, times about 7,700 kcal, is roughly 4,600 kcal — spread over 21 days, about 220 kcal a day of deficit. Add that back: your maintenance over that period was around 2,600 kcal a day.
In pounds: if you ate 2,400 a day and dropped 1.3 lb over three weeks, that is about 4,550 kcal, again roughly 215 a day, giving the same answer.
The 7,700-per-kilogram conversion is itself an approximation — it assumes the weight moving is mostly fat, which over a few weeks is roughly but not exactly true. Treat the result as a number with a margin around it, not a decimal-point fact.
What breaks this method
Inconsistent logging, not imprecise logging. If you estimate your olive oil as "a tablespoon" every day, and it is really more like one and a half, the error is a constant — it shifts your derived maintenance by the same amount every day, and your calculated number still works as a target in your own units. What ruins the calculation is logging carefully on weekdays and not at all at the weekend. That is why describing a meal in a sentence and accepting the estimate is often better practice than abandoning the log because you cannot weigh a restaurant portion; Day So Far exists largely because the meals people skip logging are the ones that decide the answer.
Too short a window. Two weeks can work if your logging is tight and your weight is stable. Three to four is safer.
Single-day weights. Comparing this Monday to last Monday is comparing two pieces of noise. Use averages.
Starting mid-fluctuation. Do not begin the window the morning after a very salty meal, a long-haul flight, a heavy training session or, if it applies to you, at a point in your cycle when fluid retention is typically high — and do not end it there either. Both ends want to be ordinary days.
Changing things mid-window. Starting a new training programme halfway through means you have measured a blend of two different lives.
Treat the number as perishable
Your measured maintenance is a description of the weeks you measured, not a permanent property. It drifts as your weight changes, as your training changes, as the seasons change what you do with your evenings, and as a long deficit does its quiet arithmetic. Re-measure every couple of months, or whenever the number stops predicting what the scale does.
That rolling comparison — what you logged against what the scale did — is the whole method, and it is the same calculation whether you run it yourself on the back of an envelope or let an app keep it updated in the background. The envelope works perfectly well. What matters is that you are measuring your own energy balance rather than inheriting someone else's average.