You need to track calories completely rather than perfectly — the errors that matter are not the small ones you make every day but the days you only half log. A steady bias, like the splash of milk you never enter, is harmless to a trend-based estimate because it shifts the expenditure estimate and the target by the same amount. A single day logged at half its real intake is not harmless: it sits in the three-week window as if it were true and pulls the estimate down for as long as it stays there. This page works both cases through with numbers. It is arithmetic, not a trial result.
Two kinds of error
Everything on this page follows from one equation. Expenditure is what you ate plus the energy your weight change implies:
expenditure ≈ logged intake + (energy implied by the trend-weight change)
MacroDiet's weight-trend screen estimates your average daily deficit "based on your rate of weight loss over the past three weeks", and that estimate plus what you logged is what the expenditure card reads. The calorie calculator, by contrast, is static: it estimates TDEE from a formula and never looks at your log. The two react to logging error differently.
| Error | Example | Static calculator (formula TDEE) | Trend-based estimate (from your log and weigh-ins) |
|---|---|---|---|
| Steady bias | 200 kcal a day of never-logged items | Real deficit is 200 kcal smaller than planned; loss slower than expected | Expenditure estimate reads 200 kcal low, so the target it produces is 200 kcal low — the bias is priced in |
| Random noise | A portion 10% heavy one day, 10% light the next | Averages out over a week | Averages out over the three-week window |
| Partial day | Breakfast and lunch logged, dinner (1,500 kcal) never entered | No effect on the formula | The day is trusted as complete; the window average falls; the estimate is too low for about three weeks |
| Blank day | Nothing logged at all | No effect | The day is skipped; the window is one day shorter |
The first row is the one people worry about and the one that matters least. Take the 75 kg example: TDEE 2,269 kcal, target 1,857 kcal, planned deficit 413 kcal. If 200 kcal a day goes unlogged, the real deficit against the formula is 213 kcal and the loss comes in at about 0.19 kg a week instead of 0.38. That is a real cost with a static number. But a trend-based estimate watches the scale move at 0.19 kg a week while the log says 1,857, and concludes that expenditure is about 2,069 kcal — 200 below the truth. The next target it sets is 200 kcal lower than a perfect log would produce, which is exactly the correction the under-logging needed. The estimate is wrong as a fact about your body and right as an instruction for your log.
In plain terms: a mistake you make every day is a mistake the system learns.
The partial-logging trap (worked example)
A partial day is different because it is not consistent. Suppose someone eats 3,000 kcal a day and logs it faithfully, except for one evening when dinner never gets entered and the day closes at 1,500 kcal. Body weight does what 3,000 kcal a day does; the log claims less.
Over a 21-day window:
honest average = 3,000 kcallogged average = (20 × 3,000 + 1,500) ÷ 21 = 61,500 ÷ 21 = 2,929 kcalerror = 71 kcal per day
The expenditure estimate is built on the logged average, so it comes out about 71 kcal a day too low, and it stays too low until that day drops out of the window three weeks later. One partial day is a small distortion. Two a week is a 142 kcal distortion that never clears, and the targets built on it will be tighter than they should be — a slower diet that looks like a plateau. Why weight loss stalls lists this among the causes for exactly that reason.
| Days logged at half intake in the window | Logged average | Estimate error |
|---|---|---|
| 0 of 21 | 3,000 kcal | 0 |
| 1 of 21 | 2,929 kcal | −71 kcal/day |
| 3 of 21 | 2,786 kcal | −214 kcal/day |
| 6 of 21 | 2,571 kcal | −429 kcal/day |
The fix is not more precision — it is completeness. The order of preference on a messy day is: estimate the missing meal roughly (a restaurant main as 800 or 1,200 kcal is closer to the truth than 0); if the whole day is unrecoverable, leave the whole day blank so it is skipped; never leave a day half filled. A blank day costs one day of data. A half day costs three weeks of accuracy.
The minimum that keeps the math honest
Precision on the days you log is less important than the number of complete days and the regularity of weigh-ins. MacroDiet's expenditure screen states its own floor: "Continuous updates require 3 days of food logging and one day of weight logging per 7-day period." Below that, the estimate holds rather than updating on thin data.
| Habit | Why it matters to the estimate |
|---|---|
| Log complete days — at least 3 in every 7 | The window average is only as honest as the days in it |
| Weigh in at least once a week, ideally most mornings | The trend needs points to move; one weigh-in a week is the floor, not the target |
| Estimate rather than omit | An estimate within a wide margin beats a missing meal by a long way |
| Leave a write-off day fully blank | Blank is skipped; partial is trusted |
| Be consistent about the small stuff | Always log the milk or never log the milk — either way it becomes a steady bias the estimate absorbs |
| Weigh dense foods, eyeball leafy ones | Error scales with calories per gram; 10% on 600 kcal of rice is 60 kcal, 10% on a salad is nothing |
Notice what is not on the list: gram-perfect scales, second decimals, logging every leaf. The cooked-versus-raw question belongs here too — it is a steady bias if you always do it one way, and a random one if you mix.
For the static calorie deficit calculator the lesson is the mirror image. Because it cannot learn your bias, its output is only as good as the log you check it against: three or four weeks of trend weight tell you whether the deficit you planned is the one you are getting, and the honest response to a gap is to log completely for two weeks before touching the rate.
The short version
Track completely, not perfectly. A steady bias — 200 kcal a day you never log — is learned by a trend-based estimate and priced into the next target. A partial day is not: one day logged at 1,500 in three weeks of 3,000 drags the estimate down by about 71 kcal a day until it leaves the window. MacroDiet keeps updating on three complete food days and one weigh-in a week; estimate what you cannot weigh, leave a lost day blank, and never log half of one.
Educational overview only — not medical advice. Zyra is a training and nutrition tracking tool; the rates, caps and floors on this page describe how the calculator behaves, not what any individual should do.