Every diet that works is running the same engine#
A calorie deficit means your body takes in less energy from food than it spends, averaged over days and weeks rather than over one evening. That is the whole mechanism of fat loss, and it is what keto, fasting, points, portion control and Mediterranean eating all have in common on the days they work. The cleanest demonstration is the POUNDS LOST trial: 811 overweight adults were randomly assigned to one of four reduced-calorie diets spanning 20% to 40% of energy from fat, 15% to 25% from protein and 35% to 65% from carbohydrate. Two years later, weight loss landed between 3.0 and 3.6 kg in every arm, with no significant difference in any comparison (P > 0.20). The authors' own conclusion: reduced-calorie diets produce clinically meaningful weight loss "regardless of which macronutrients they emphasize"1.
So the arithmetic is settled. What is not settled — and what most calorie advice quietly gets wrong — is the schedule. A deficit is not a number you set on Sunday and collect on Friday. It is the gap between two quantities you can only estimate, and both of them move when you push either one. The workable version of "eat less than you burn" is closer to this: pick a plausible intake, hold it long enough to read a trend, and let your own weight tell you afterwards what the deficit actually was. Below is why the physiology forces that reframing — and where each specific decision (how big, how fast, how to count) is worked out in its own article.
The first law is true, and mostly unhelpful#
Energy cannot be created or destroyed, so what you store has to equal what you eat minus what you burn. That is not a diet philosophy; it is bookkeeping, and no serious researcher disputes it. The trouble is that a true identity is not a prediction. It guarantees the books balance. It says nothing about what any one term will do when you move another.
The concrete failure is the rule everyone was taught. A pound of fat is called 3,500 kcal, so a 500-kcal daily deficit is supposed to yield a pound a week, week after week. It does not, and the field said so in public: a working group of energy-balance researchers concluded that "the 3500 kcal per pound rule should no longer be used," because it is routinely misapplied in a way that implies a linear weight change nobody actually experiences3. Worth flagging the provenance, since this blog holds one standard for everyone: that panel was convened by the American Society for Nutrition together with ILSI North America, a food-industry-funded body, though the authors declared no conflicts and the conclusion — that dieters should expect less than the folk rule promises — is not one a food sponsor has any obvious use for. It also matches what Hall's independent NIH modelling found on its own. The rule's arithmetic was never the error; extrapolating it forward in a straight line is. That distinction gets unpacked in is a pound of fat really 3,500 calories.
What a deficit actually does over time#
Replace the straight line with a model that lets the body push back, and the predictions change shape entirely. The NIH's dynamic model of adult metabolism — the engine behind the agency's own Body Weight Planner — simulates expenditure adapting as weight comes off. Its headline result is that for an average overweight adult, every sustained change in energy intake of 100 kJ per day leads to an eventual body-weight change of about 1 kg, and that "the bodyweight response to a change of energy intake is slow, with half times of about 1 year"2. The consensus panel put the same timeline in plain numbers: roughly 1 year to reach 50% of the eventual loss and about 3 years to reach 95%3.
(100 kJ is about 24 kcal — that conversion is mine, not a figure either paper prints. Hall reports the rule in kilojoules.)
| The question | The 3,500-kcal straight line | The dynamic model |
|---|---|---|
| Cut ~100 kJ/day and hold it forever | weight keeps falling, without limit | about 1 kg lost, then a new steady weight |
| How fast does it arrive? | the same amount every week, indefinitely | ~50% of it in a year, ~95% in three |
| Does your starting body fat matter? | no — a pound is a pound | yes: more adiposity means more loss for the same cut, and a longer wait |
Read the middle row twice, because it is the row that ruins diets. The straight line predicts a constant weekly payout, so anything less than that reads as failure. The real curve is steepest at the start and flattens as your smaller body burns less — which means a slowing scale is the model working as designed, not evidence that you broke. The same physics runs in reverse, and this is the unsettling part: Hall's group calculated that a persistent average gap of only about 30 kJ per day between intake and expenditure accounts for the observed average weight gain of the US adult obesity epidemic2. Nothing in energy balance happens fast, in either direction.
Both sides of the equation move when you push one#
The deepest reason a calculated deficit is not a delivered deficit is that "calories in" and "calories out" are not independent dials. As the consensus panel put it, "all of the components of energy balance interact with each other." Eat less and you spend less: less food to digest, less mass to carry, and — the part nobody plans for — less spontaneous movement.
How much of your deficit does that erase? A systematic review and meta-analysis of 28 randomized controlled trials compared what participants actually did against models that account only for metabolic compensation — so any remaining shortfall is behaviour, not chemistry4.
| Intervention | Result vs. a no-behavioural-compensation prediction |
|---|---|
| Dietary restriction | up to 12–44% less weight loss than expected |
| Exercise | 55–64% less weight loss than expected |
| Overfeeding | 96% less weight gain than expected |
That exercise row is the one to sit with, and it is why adding training to a diet so often does less than the treadmill's readout promised — the reasons are laid out in whether exercise burns as many calories as you think. The overfeeding row is the same defence pointed the other way: the body resists gain about as stubbornly as it resists loss. And on top of all this behavioural compensation sits a genuine metabolic one — hold a weight 10% or more below your maximum and 24-hour expenditure runs measurably below what your remaining tissue predicts, an effect that is real, persistent, and covered in what metabolism actually is.
A calorie deficit is not a setting you choose. It is a result you find out, weeks later, that you had.
Your deficit is the gap between two estimates#
Here is the arithmetic problem that no app can solve for you, and it deserves stating plainly because it explains most of the frustration in this whole topic. A deficit is a subtraction, so it inherits the error of both inputs — and the smaller the answer is relative to the terms, the worse that gets.
Work it through. Say your calculated maintenance is 2,400 kcal, carrying roughly ±10% of real individual error — that is a genuinely good estimate, better than most people have, and the reasons it cannot be tighter are in TDEE explained. Say you log 2,000 kcal, and you log carefully, so call it ±20%; the layers behind that figure are audited in how accurate calorie counting really is. On paper you have a 400-kcal deficit. But combine the two error bars and the plausible truth runs from a 240-kcal surplus to a 1,040-kcal deficit. (That worked example is my own arithmetic on two published error magnitudes — neither source prints it, and note that ±20% means twenty percent in either direction, not a hidden discount.)
This is not an argument against counting. It is an argument about what the count is for. The number you set is a hypothesis about your body; the trend on the scale over weeks is the measurement that tests it. Consistency beats precision here, because a consistent error cancels out of a comparison of you against yourself — which is exactly why the deficit is worth reading as a band rather than a target you either hit or missed.
What to set, and how to know it worked#
With the model in hand, the practical decisions are narrower than the internet suggests, and each one has its own evidence:
- How big. Size it against your own maintenance rather than a flat 500, and keep it modest enough to survive: how big should your calorie deficit be.
- How fast. The sustainable band scales with your bodyweight, not with a universal 1–2 lb: what a safe rate of weight loss looks like.
- Over what window. Your body integrates energy over days, so a single day means much less than the week it sits in: does your body count calories daily or weekly.
- What the deficit is made of. The deficit decides how much you lose; protein has a lot to say about whether it comes off as fat or muscle — see protein for weight loss.
- How to actually count. Method matters more than app choice: how to count calories.
And one finding from POUNDS LOST outranks all of them. Across every diet arm, attendance at the counselling sessions was strongly associated with weight loss, at about 0.2 kg per session attended1. Showing up predicted the result better than the macros did. Treat that as the design brief: the best deficit is not the largest one you can calculate, it is the one you are still running in month six, when the curve has flattened and the model says you are exactly on schedule.
FAQ#
What is a calorie deficit, exactly?#
It is the state of taking in less food energy than your body expends, averaged over a meaningful stretch of time — not a food, a protocol, or a product. Every diet that produces fat loss produces it this way. In the 811-person POUNDS LOST trial, four diets with very different macronutrient splits all landed within 3.0–3.6 kg of loss at two years, with no significant difference between them1.
Does it matter which foods the deficit comes from?#
For the amount of weight you lose, far less than the marketing implies — POUNDS LOST found no significant separation between 65% and 35% carbohydrate at two years. For everything else, it matters a lot: which foods you choose drives how full you feel, how well you stick to the plan, and how much of the loss is fat rather than muscle.
How long does a calorie deficit take to show up?#
Longer than a straight-line rule suggests. Modelling of adult metabolism puts the half-time of the body-weight response to an intake change at roughly a year, with about 95% of the eventual change arriving over about three (Hall et al., 2011; Hall et al., 2012). Early weeks move faster than that average because of glycogen and water shifts, which is also why the first fortnight flatters every diet.
Sources#
- Sacks FM, et al. Comparison of weight-loss diets with different compositions of fat, protein, and carbohydrates. N Engl J Med. 2009;360(9):859-73.
- Hall KD, Sacks G, Chandramohan D, Chow CC, Wang YC, Gortmaker SL, Swinburn BA. Quantification of the effect of energy imbalance on bodyweight. Lancet. 2011;378(9793):826-37.
- Hall KD, Heymsfield SB, Kemnitz JW, Klein S, Schoeller DA, Speakman JR. Energy balance and its components: implications for body weight regulation. Am J Clin Nutr. 2012;95(4):989-94.
- Dhurandhar EJ, Kaiser KA, Dawson JA, Alcorn AS, Keating KD, Allison DB. Predicting adult weight change in the real world: a systematic review and meta-analysis accounting for compensatory changes in energy intake or expenditure. Int J Obes (Lond). 2015;39(8):1181-7.
- Hall KD. What is the required energy deficit per unit weight loss? Int J Obes (Lond). 2008;32(3):573-6.
- NIDDK. Body Weight Planner. National Institute of Diabetes and Digestive and Kidney Diseases.
