Your maintenance calories are the average daily intake at which your weight average holds steady. To find them, log what you eat and weigh yourself every morning. Two weeks narrows the number to roughly ±300 kcal a day; four weeks, to about ±105. A calculator estimates a population. Your log measures you.
TL;DR
- A calculator gives you a population estimate. Your own log and scale give you a measurement.
- One morning's weight means very little. Day-to-day swings of around half a percent of body weight are normal.
- With daily weigh-ins, 14 days pins maintenance to about ±300 kcal a day and 28 days to about ±105. I ran the arithmetic; the table is below.
- ±300 kcal is the size of one activity-level step in a calculator, so day 14 is a coarse check and day 28 is an answer.
- Average whole weeks, keep festive weeks out of the window, and keep the logging method fixed for the whole run.
What are maintenance calories?
Maintenance calories are the intake at which your body weight holds steady on average: what you eat matches what you burn. You will also see it called TDEE, total daily energy expenditure. A maintenance calorie calculator estimates it: an equation predicts your resting metabolic rate from age, sex, height and weight, and a multiplier scales it up for activity. Your log measures it: you record what you eat, weigh yourself, and watch what the scale does.
A 2005 systematic review compared the four equations most used in practice (Harris-Benedict, Mifflin-St Jeor, Owen and WHO/FAO/UNU) and found Mifflin-St Jeor the most reliable: it predicted resting rate within 10% of the measured value in more people than any other, with the narrowest error range. The same review adds that noteworthy errors and limitations exist when it is applied to individuals (Frankenfield et al., J Am Diet Assoc, 2005). You are an individual.
Why does the calculator give a different number from my log?
Because the equation predicts the average for people who share your inputs, and you sit somewhere around that average, not on it. Then comes the activity level. Calculators use conventional multipliers: 1.2 for sedentary, 1.375 for lightly active, and so on upward. Take a resting rate of 1,700 kcal. Sedentary gives 2,040 kcal a day; lightly active gives 2,338. One step on a menu you choose by feel moves the answer by about 300 kcal a day, and most people could argue for either box. Quick rules that multiply your body weight by a fixed number are the same estimate with fewer inputs. A calculator gives you a sensible number to start from. When your log disagrees with it for long enough, the log wins.
How do I find my maintenance calories from my own log?
- Weigh yourself the same way every morning. Copy the protocol from the longest weigh-in record I know of: one healthy man weighed himself on 9,521 days between December 1993 and April 2023, always after the overnight fast, undressed, after using the toilet and before any food or drink (Schneditz et al., Renal Failure, 2023). It is one subject, and that is why the protocol is worth copying. In practice: same scale, same spot on the same floor, after the toilet, before the kopi.
- Log everything with one method, drinks included. Which method matters less than keeping it. I compared six in the best ways to track calories, and the one that wins is the one you are still using on day 30. A log that runs high every day is fine; one whose method changes halfway is not.
- Average the intake. Add up the logged calories and divide by the number of days you logged.
- Take the weight trend, not the gap between two mornings. Fit a straight line through every morning weight; a spreadsheet's SLOPE function gives it in kilograms per day. Or compare the average of the first half of the run with the second.
- Convert the trend to calories. The conventional figure is about 7,700 kcal per kilogram, the metric form of 3,500 kcal per pound. Kevin Hall traced it to a 1958 calculation that assumed the weight lost is 87% fat, and calls it an oversimplification: it roughly holds for people starting with more than about 30 kg of body fat and overstates the energy per kilogram for leaner people. In his example of a leaner person it comes to about 5,900 kcal per kilogram (Hall, International Journal of Obesity, 2008). So the conversion is itself uncertain by roughly a quarter. Treat the result as a direction and a size, not a decimal.
- Add the drift back. If the trend went down, you ate below maintenance, so maintenance is your average intake plus the daily gap. If it went up, subtract.
A made-up example, to show the arithmetic only. Over 28 days someone logs an average of 2,100 kcal a day, and the line through their morning weights falls 0.8 kg. That is 0.8 × 7,700 = 6,160 kcal, or 220 kcal a day below maintenance: maintenance, in the units of their log, is about 2,320 kcal a day, give or take the ±105 in the table below. At Hall's leaner 5,900 kcal per kilogram it is about 2,270. At this resolution, the same answer.
How long does it really take?
Longer than two weeks, if you want an answer rather than a hint. I ran the arithmetic. The input is the day-to-day wobble in that thirty-year record. From one morning to the next, the change in weight had a standard deviation of 0.53% of body weight, with a mean and median of zero. If each morning's error is independent, a single morning is off by about 0.375%: 0.30 kg for an 80 kg person. I simulated 20,000 runs of daily weigh-ins for such a person with a true drift of 300 kcal a day below maintenance, fitted a straight line to each run, converted the slope at 7,700 kcal per kilogram, and measured how far the answers spread. Each band below holds 95% of the runs.
How precisely daily weigh-ins pin down maintenance, in kcal a day. “Straight line” is the fitted trend; “two halves” compares the average weight of the first half of the run with the second.
- 7 days
- Straight line: about ±859
- Two halves: about ±990
- 14 days
- Straight line: about ±301
- Two halves: about ±345
- 21 days
- Straight line: about ±165
- Two halves: about ±189
- 28 days
- Straight line: about ±105
- Two halves: about ±121
- 42 days
- Straight line: about ±58
- Two halves: about ±66
Two things follow. Over 14 days, the smallest weight change the straight line can tell apart from noise is 0.56 kg, which is about 306 kcal a day. That is one activity-level step in a calculator. Day 14 can tell you whether the calculator was roughly right or a whole step out; it cannot tell a 2,200 from a 2,400. By day 28 the band is about ±105 kcal a day, and that is an answer. To run the same fit on your own weigh-ins, measure your maintenance calories from your log.
The second column is what most people actually do: compare this week's average weight with last week's. It costs a little precision, ±345 at 14 days against ±301, and the gap narrows as the run gets longer.
The assumptions, plainly: one man's noise level, errors independent from one day to the next, and a fixed 7,700 kcal per kilogram. The record itself suggests independence is optimistic, because the spread of the change grows from 0.53% over one day to 0.69% over seven: some of the wobble lasts for days. If I use 0.53% directly as the daily error, every band widens by about 40%, to ±427 at 14 days and ±151 at 28. If your mornings wobble more than his did, use those. This is arithmetic on a published noise figure, not a study of people finding their maintenance.
What ruins the measurement?
- The weekend. In 80 adults with 4,657 weigh-ins, weight was highest on Sunday and Monday; increases began on Saturday and decreases on Tuesday (Orsama et al., Obesity Facts, 2014). A larger study of 1,421 people, about 220 weigh-ins each, found a within-week swing of 0.35%, up over the weekend and down through the week (Turicchi et al., PLoS One, 2020). Compare Monday with Monday, or average whole weeks.
- Festive weeks. In that same study, weight rose by a mean 1.35% over Christmas, and in March it was still at least 0.35% above the pre-Christmas weight. Our Christmas is Hari Raya, Chinese New Year and Deepavali open-house season. Don't run the window across one, and don't read the fortnight after one as your new maintenance.
- Changing how you log halfway. A new app, a new portion habit, starting to count drinks in week three: each changes the units of the log.
- Reading one morning. At 80 kg a single morning is typically 0.3 kg off the trend for no reason at all.
- A late, heavy or salty dinner. The next morning is often up. Read it against the day before, not as fat gained overnight.
Which of this have I actually done?
Part of it. I tracked for about four weeks by describing each day to an AI chatbot, then built the quick-log in Habitus Mind to reproduce that. One rule from those weeks is this post's rule too: when a weigh-in jumps, the quick-log reads it against the previous days (a heavy, salty or late meal) instead of as fat gained overnight.
The simulation above is arithmetic, not my data. I'm not publishing my own weight or maintenance number, and wouldn't offer them as evidence.
Do I need an app to find maintenance calories?
No. A notebook, a scale and a calendar are enough. The from-your-log calculator does the arithmetic in your browser and keeps nothing.
If you log in Habitus Mind, the Fitness module shows part of this. The dashboard draws a food intake trend over 7, 30 or 90 days or a custom range, with the average kcal a day for that range. The Weight card shows your last 30 weigh-ins as a sparkline with the change from one day to the next; the weigh-in is one field on Daily Add. The quick-log's estimates lean high on purpose and flag oily, sauced or vaguely portioned food as uncertain.
What it does not do: it does not compute maintenance calories, and it does not average your weight by week. Its range average also counts a day with nothing logged, today included, as zero, so over a range with gaps it reads lower than the average of the days you did log. The averaging in this post is yours to do.
Habitus Mind is free to start. The quick-log lives in its Fitness module; to try it, create a free account.
FAQ
What are maintenance calories?
The average daily intake at which your weight holds steady: what you eat matches what you burn. It is also called TDEE. A calculator estimates it from your age, sex, height, weight and activity level. A consistent food log and daily morning weigh-ins measure it for you.
How long does it take to find your maintenance calories?
About four weeks of daily morning weigh-ins and a consistent food log. By my arithmetic on published day-to-day weight noise, 14 days pins maintenance to about ±300 kcal a day and 28 days to about ±105. Day 14 is a coarse check; day 28 is an answer.
Are maintenance calorie calculators accurate?
As a starting point, yes. A 2005 systematic review found Mifflin-St Jeor the most reliable of four common equations, with noteworthy errors for individuals. The activity level adds more: one step, from sedentary to lightly active, moves the answer by about 300 kcal a day.
How much does weight fluctuate from day to day?
In one man's nearly thirty-year record of standardised morning weigh-ins, the change from one day to the next had a standard deviation of 0.53% of body weight, averaging zero. Across 1,421 people, weight also swung about 0.35% within the week, higher after the weekend. One morning tells you very little.
Should I weigh myself every day?
If you are trying to find maintenance, yes: daily weigh-ins are what make four weeks enough. Weigh the same way each morning, after the toilet and before eating or drinking, on the same scale and floor, and judge the trend or weekly averages, never a single morning.
