An app estimates your energy expenditure from your logs by running the calorie equation backwards. A calculator guesses expenditure from your height, weight and a dropdown; an adaptive app instead takes what you actually ate and what your weight actually did, and solves for the only unknown. It is the difference between predicting the weather and reading a thermometer — slower to start, but it measures you rather than the average person.
Intake in, weight change out
Energy balance is bookkeeping: calories in − calories out = change in stored energy. Rearranged:
calories out = calories in − change in stored energy
"Calories in" is your food log. "Change in stored energy" is read off the weight trend: if trend weight is rising, energy is being stored; if it is falling, energy is being drawn down. The rate of change, multiplied by an energy value for the tissue involved, gives kcal per day. Subtract that from average intake and what is left is expenditure.
The worked example every adaptive app uses in some form: you log 3,000 kcal a day, and the trend-weight slope says you are storing about 200 kcal a day. Then 3,000 − 200 = 2,800 kcal is your expenditure. No equation, no activity multiplier, no guess about your job — just arithmetic on your own numbers.
| Input | Where it comes from | What it feeds |
|---|---|---|
| Average daily intake | food log | "calories in" |
| Trend-weight change rate | smoothed weigh-ins | "change in stored energy" |
| Energy value of the tissue changing | a model: fat ≈ 9,400 kcal/kg, lean ≈ 1,800 kcal/kg, mixed by rate | converts kg/day to kcal/day |
| Starting estimate | a BMR equation × activity multiplier | the first value, before data exists |
The starting estimate is the same population formula a calculator uses, with the same error — MacroDiet's own tutorial says it "may be higher or lower than your true expenditure". Its only job is to give the model somewhere to begin.
Why it takes about three weeks
Three reasons, all about noise:
- Daily weight is noisy. Water, glycogen, salt and what is in your gut move the scale by a kilogram or more overnight; that is 7,700 kcal of apparent "energy" that never existed. A trend line over many days averages the noise away; a few days cannot.
- The tissue is inferred from the rate. A slow loss is mostly fat; a fast one includes more lean tissue and water. Because fat carries around five times the energy per kilogram of lean tissue, the model needs enough days to be confident about the rate before it can convert kilograms to kcal.
- The estimate is deliberately cautious. Early on the model allows larger steps and then tightens; it can overshoot by a few dozen kcal and settle back. Most adaptive models use roughly a three-week window of nutrition data, and MacroDiet's guidance is that after 20–30 days of consistent tracking you should have a refined, personalised estimate.
After that first month, a genuine change in expenditure — a new job, a training block, NEAT falling during a diet — shows up in the estimate within a week or two, because it shows up in the weight trend.
In plain terms: the app needs enough mornings to tell a trend from a splash.
What incomplete logging does to it
This is the one way to get a genuinely wrong answer. The model trusts your log. A day you logged at half its real intake is not flagged as "probably incomplete" — it is taken as a low-intake day, and it drags the average down.
The arithmetic: twenty days at 3,000 kcal and one day logged as 1,500 when it was really 3,000. Average intake in the window becomes (20 × 3,000 + 1,500) ÷ 21 ≈ 2,929, so expenditure reads 71 kcal lower than it should — for the whole three weeks that day stays in the window, and every target built on it is 71 kcal too low. Two such days, roughly double.
The fix is counter-intuitive but simple: a blank day does less damage than a half day. Blank days are skipped; half days are believed. If you cannot log a day properly, either estimate the whole day roughly (an estimate within about 30% of reality is fine) or leave it empty. Consistently untracked small items — the same splash of milk every morning — are harmless: the estimate simply becomes expenditure net of the milk, and every target is correct relative to it.
How MacroDiet does it (3 of 7 days)
MacroDiet's expenditure screen states its rules on the page, so here they are verbatim in substance:
| Stage | What MacroDiet needs | What it does |
|---|---|---|
| Day 0 | Your onboarding answers | Starting estimate from validated equations — may be higher or lower than your true expenditure |
| Start updating | 3 consecutive days of food logging | Begins revising the estimate from your nutrition and weight data |
| Keep updating | 3 days of food logging and 1 weigh-in in every 7-day period | Continuous updates; the estimate follows your lifestyle, activity and intake as they change |
| Too little data | Fewer than 3 food days or no weigh-in in a 7-day period | Holding — the estimate is kept at its last value rather than guessed from thin data |
| Dialled in | 20–30 days of consistent tracking | A refined, personalised expenditure estimate |
Exercise is not an input. The tutorial's answer to "do I need to track exercise?" is no: your weight and nutrition data give the algorithm all the information it needs, because the weight trend already contains the net effect of every session and every step. That is also why the app does not add a tracker's "calories burned" back to your target — see why exercise burns fewer calories than the tracker says.
The 3-of-7 floor is lenient on purpose: a missed weekend does not break the estimate. What does break it is the half-logged day above, which is why "log fully or leave it blank" is the one habit worth building.
This is a description of arithmetic and of what the app's own screen says, not medical advice. The energy value of tissue is a model constant, not a measurement of you; the estimate is only as honest as the log it is built from.
The short version
Adaptive apps solve expenditure = intake − stored energy with the stored-energy term read from your trend weight — 3,000 kcal in and a 200 kcal surplus means 2,800 out. It takes about three weeks because daily weight is noisy and the tissue behind a change is inferred from its rate. Half-logged days are the one thing that fools it; blank days are safer. MacroDiet starts updating after 3 consecutive logged days and keeps updating with 3 food days and 1 weigh-in per week, otherwise it holds. Educational overview only — not medical advice.