Finding your maintenance calories: the 2-week real-world test

“What’s my maintenance calories?” — typed into a search box a thousand different ways, it is the most practical question in nutrition, because maintenance is the number every cut and every bulk is measured from. A calculator can predict it in seconds, and I use that prediction as the starting bid. But the only way to truly find my maintenance calories is to measure them: eat at the predicted number, watch the scale’s trend for two weeks, and let simple math correct the estimate. Here is the exact test I run, with a real worked example and the adjustment math shown.

Step 1: Calculate my maintenance calories — the prediction

Every test needs a starting number, and mine comes from the standard estimate: BMR from the Mifflin-St Jeor equation, multiplied by an activity factor. Our worked example is a 34-year-old man, 178 cm, weighing about 81 kg, with a desk job and three weekly gym sessions. His BMR is 10 × 81 + 6.25 × 178 − 5 × 34 + 5 = 1,758 calories. He picks “light” activity (×1.375) because his job is seated — a defensible, even conservative choice — giving a predicted maintenance of about 2,430 calories a day.

You can get your own prediction in seconds from our TDEE calculator. Write the number down. Then forget its authority — for the next two weeks it is a hypothesis, and your body is about to grade it.

Step 2: The 2-week test protocol

1. Eat at the predicted number — 2,430 a day for him — as consistently as you honestly can. Ordinary life accuracy is fine; restaurant-guess accuracy repeated daily is not.
2. Weigh yourself every morning, after the bathroom, before food and drink, and record it. Same scale, same spot on the floor.
3. Do not react to single days. Daily weight is mostly water noise. You are collecting data, not verdicts.
4. After 14 days, average week 1 and week 2 separately and compare the two averages. That difference is your trend.
5. Adjust by the math in step 3, then run one more week to confirm if the change was large.

His actual log

Fourteen morning weigh-ins at a steady 2,430 calories a day. Individual days swing by 0.7 kg; the weekly averages move by only 0.25 kg — which is exactly why averages, not days, drive the decision.

Look at day 9: 80.1, the lowest reading of the fortnight — followed immediately by 80.4 and 80.2. Anyone reacting to single weigh-ins would have declared victory on day 9 and disaster on day 3. The averages tell the calmer truth: his weight drifted down by a quarter of a kilo per week while eating 2,430.

Step 3: The adjustment math

A kilogram of body fat stores roughly 7,700 calories — the standard energy-density estimate used throughout nutrition science. That gives a clean conversion: a trend of 0.1 kg per week ≈ 110 calories a day (0.1 × 7,700 ÷ 7). Scale it linearly:

His trend was −0.25 kg per week at 2,430 eaten, implying a daily deficit near 275 and a true maintenance around 2,705 — his “light” activity pick had undercounted his training and weekend walking by roughly 11%. Note the size of that error: not a broken formula, just one activity level chosen one notch too low, which is the most common prediction mistake there is.

One discipline rule from the protocol: even when the math says 275, I adjust in steps of 100 to 200 calories, not the full amount. He moves to about 2,630, runs one confirmation week, and only adds the rest if the trend is still negative. Small steps protect against the noise the averages could not fully remove — water shifts can still bend a two-week trend by 50 to 100 calories a day.

When the scale lies to you

Calculating maintenance calories this way works because fat change is slow and steady. Water is neither, and it will happily impersonate fat for days at a time:

• Glycogen: your muscles and liver store carbohydrate with roughly 3 grams of water per gram of glycogen. A few high-carb days can add 300–500 g of scale weight that is not fat; a few low-carb days can strip it and look like miraculous loss.
• Sodium and big meals: a salty restaurant dinner plus food still in transit can add a full kilogram by morning, gone within two days.
• New training: starting or intensifying exercise causes temporary inflammation and water retention that can mask fat loss for one to two weeks.
• The menstrual cycle: shifts of 1–2 kg across a cycle are common, which is why comparing the same phase month-to-month — or simply running the test for three weeks instead of two — gives a cleaner read.

All four are reasons the protocol uses weekly averages and small adjustments rather than daily panic.

Running the test in reverse

The math is symmetric. Say a smaller eater — predicted maintenance 2,600 from the formula — eats exactly that and gains 0.2 kg a week on average. The table above reads that as a 220-calorie daily surplus, so her true maintenance is about 2,380, and her next test week runs at 2,450–2,500 after a cautious step down. Formula overestimates happen just as formula underestimates do, particularly for people whose jobs look active on paper but involve long seated stretches. Either direction, the test finds the number the formula missed — usually within one adjustment round.

The same test on two more bodies

The protocol is body-agnostic — here it is run twice more, quickly, to show the range of outcomes. A 32-year-old woman, 160 cm, 58 kg: BMR 1,259, light activity predicts maintenance of 1,731. She eats there, and her weekly averages rise by 0.1 kg — a 110-calorie daily surplus on the conversion table — putting her true maintenance near 1,620, and her retest at 1,650 after a cautious step down. The formula overshot her by about 7%, most likely because her “light” week contains fewer steps than the label implies. Contrast a 100 kg man, same equation, moderately active: predicted 3,069, eats there, averages flat for two weeks. No adjustment, no drama — his prediction was simply right, and he now owns a measured number instead of an estimate. Of the three people in this article, the formula was high once, low once and right once, each time by less than 300 calories — a fair summary of what to expect from any honest prediction, and exactly the gap the test exists to close.

The mistakes that ruin the test

• Weekend drift: eating 2,430 on weekdays and 3,200 on Saturday and Sunday averages about 2,650 — you tested a number you never intended to. The average intake is the intake; log the weekends.
• Changing two things at once: starting a new training programme mid-test adds water retention and appetite noise. Keep training, food and sleep as boring as possible for the fortnight.
• Tracking by memory: untracked cooking oil, milk in coffee and finishing a child’s plate routinely add 150–250 calories a day — the same size as the error you are trying to measure. Weigh and log for the test’s two weeks even if you never do it again.
• Quitting at day 6: the first week’s average still carries the water shift from the intake change itself. The signal only separates from the noise across the full fourteen days.

How accurate is the number you end up with?

Honest precision talk, because a measured number feels more certain than it is. Day-to-day, real maintenance wobbles: activity varies by ±5% or more across ordinary weeks, food labels themselves carry a legal tolerance, and absorption differs slightly between people and meals. After a clean two-week test and one adjustment round, I treat the result as accurate to about ±100 calories — call it 2,705 measured, 2,600–2,800 true. That is still three times tighter than the formula’s typical ±10–15% for an individual, and tight enough for every practical purpose: a cut built 15% below the measured number carries a deficit that survives the remaining error. What the test cannot do is freeze the number forever. Lose or gain 5 kg, change jobs, change seasons, and maintenance has moved — retest for a week and re-measure, the same way you would recheck any instrument against a changed load. Keep the original log when you do: comparing this year’s test against last year’s, at a different weight and a different job, is the most honest record of how your own energy balance actually behaves over time.

Sources: Mifflin et al., American Journal of Clinical Nutrition (1990) — the BMR equation used for the starting prediction; the 7,700 kcal-per-kilogram figure is the standard energy-density estimate for body-fat change (after Wishnofsky, 1958), used here as the approximation it is — real tissue change mixes fat, lean mass and water. This article is general information about a self-measurement method, not medical advice; anyone with a medical condition affecting weight or diet should run changes past their clinician.