How to count calories without weighing your food

Putting a number on a plate of food means answering four separate questions, and only one of them is about weight.

1. What is it? Chicken thigh or chicken breast. Greek yoghurt or the 0 per cent version. Whether the sauce is tomato-and-stock or tomato-and-cream. This is worth more calories than any portion question and no scale addresses it.
2. How much of it is there? The portion. This is the only one a scale answers, and it is also the one you can learn to estimate well, because you eat the same twenty foods most weeks.
3. What went into cooking it? The oil in the pan, the butter under the fish, the glaze, the fat the meat was roasted in. A photo cannot know how much oil went into the pan, and neither can a scale you used before cooking started.
4. How dense is it? Two visually identical bowls of soup can differ threefold depending on whether the base is stock or cream. Density is a question about the recipe, not the portion.

This is not an argument against kitchen scales. It is an argument about where the error lives. Someone who weighs 145 grams of chicken to the gram and then free-pours the oil has bought precision on the cheapest term in the sum while leaving the expensive one untouched. The same asymmetry is why restaurant and takeaway meals are so hard to log — in someone else’s kitchen, questions one, three and four are all unanswerable at once, and a scale on your table would not help with any of them.

Energy density spans a factor of about twenty-five across ordinary food. That single fact tells you where to aim your effort: the cost of a mistake is the size of the mistake multiplied by the density, so a small error on a dense food beats a large error on a light one.

There is a second reason to watch the dense end. Errors on light foods are symmetrical — sometimes you take more salad than you thought, sometimes less, and across a week it averages out. Errors on fats are not, because fats are usually added, and the failure mode is forgetting an addition rather than imagining one. Nobody has ever logged a tablespoon of oil they did not use. That is what makes the systematic drift downward.

Hands are the best portable reference for one dull reason: yours is always with you, at the table, in a canteen, at somebody else’s house. It also scales roughly with the rest of you, so a larger person’s palm holds more — which is loosely appropriate, since a larger person usually needs more. Treat that as a convenience, not a law of physiology.

The thumb is the one that matters

Most people who use this system remember the palm and the fist, because protein and vegetables feel like the important parts of a meal. They are the two units where an error is cheapest. The thumb — oil, butter, nut butter, dressing, cream — is the one carrying 125 calories a unit, and it is the one that gets skipped, because a splash of oil does not feel like an ingredient. Count the thumbs first and the rest of your estimate can be sloppy.

What hands cannot do

A hand measures volume, badly, and what you want is mass. That is tolerable for solids of a familiar shape and useless for anything poured, piled or hidden: a bowl of soup, a smoothie, a curry over rice, a salad that has already met its dressing. For those, the next section is more use than your hand is.

For foods a hand handles badly, borrow an object whose size you already know precisely. The value of a reference object is not that it is more accurate than a hand — it is that it is fixed, so it does not quietly grow on the days you are hungry.

Why piles fool you

Volume grows with the cube of length. A serving of pasta that is 20 per cent taller and 20 per cent wider in both directions than your usual is not 20 per cent more food — it is 1.2 × 1.2 × 1.2, about 73 per cent more. “A bit more than normal” is very often close to double, and this is the single most common way an eyeballed portion goes wrong.

Bowls hide the portion

Bowls are the worst case for every estimator, human or otherwise. The rim conceals how deep the food goes, two bowls of similar diameter can hold wildly different volumes, and in a soup or a stew the fat is dissolved out of sight. When we tested three AI vision models on 30 meal photographs, the widest disagreements were all bowls: one photograph of a bowl of ramen drew estimates of 250, 631 and 739 kcal from the three models, and across all the food photos the median gap between the highest and lowest estimate was 49.2 per cent. If you eat a lot of soups, stews and curries, your log is noisier than a plate-eater’s, however you produce it.

There is a one-minute fix worth doing. Fill the bowl you actually use with water to the level you actually fill it, tip that into a measuring jug, and write the number down. You now know your bowl holds, say, 450 ml, permanently, and every future bowl of anything is a fraction of a known volume rather than a guess about a shape.

The fastest route to never weighing your food is to weigh it for one evening. Published hand-size guidance is a population average applied to your hand; twenty minutes converts it into your own numbers, after which the scale can go back in the cupboard.

1. Serve the way you normally serve. Do not make a careful portion for the exercise. The number you want is what you habitually put on the plate.
2. Weigh it, then look at your hand next to it.Note the grams beside the hand unit — “my cupped hand of cooked rice is 165 g”, not “a cupped hand is 120–160 g”.
3. Do the oil test. Pour into a tablespoon the amount you would normally put in the pan, and see how many tablespoons it fills. This is the highest-value sixty seconds in the whole exercise, because each tablespoon you were not counting is about 125 calories.
4. Cover your regulars, not the long tail. Six to eight foods usually account for most of what a person eats: their rice or pasta serving, their cereal bowl, their usual meat portion, the milk in their coffee, the knife-load of nut butter, the cheese on the pasta.
5. Weigh cooked if you serve cooked. Rice roughly triples in weight with water. A calibration in dry grams is useless at the table.
6. Redo it when your food changes. A new job, a new flatmate, a new cuisine, a new bowl. Once a year otherwise.

Write the results somewhere you will find them — a note on your phone is fine. The output of this exercise is about ten lines of text, and it is the difference between using someone else’s hand and using yours.

Chicken, rice, some vegetables, cooked in olive oil. Estimated with the hand units above, then compared against what the plate might actually have held.

The estimate came in 151 calories low, about 20 per cent. Look at where that came from. Seven grams of oil cost 63 calories. Twenty-five grams of chicken cost 43. The vegetables, wrong by twenty grams, moved the total by seven. The smallest quantity on the plate produced the largest error, which is the whole argument in one line.

Two honest observations about the total. The first is encouraging: errors in different directions partly cancel, so a day of six estimates is proportionally more accurate than any one of them. The second is not: that only holds while the errors are independent, and the oil error is not. It runs the same way at breakfast, lunch and dinner, so it does not cancel — it accumulates. Fixing your oil estimate is worth more than fixing everything else combined.

Some meals are not estimable, and pretending otherwise is how a tracking habit quietly stops meaning anything.

• Anything cooked by someone else. Restaurant kitchens use more fat than home cooks, and you cannot see any of it. This is a different problem with its own set of tactics for takeaways and eating out.
• Mixed dishes where you cannot see the components. A lasagne, a curry, a pie, a smoothie. You are not estimating a portion, you are guessing a recipe, and the range is enormous.
• Food eaten by hand from a container. Nuts, crisps, chocolate, cheese while cooking. The failure is not that a handful is hard to size — it is that nobody counts the handfuls. The only real fix is to portion out before eating rather than estimate after.
• Drinks. Juice, lattes, alcohol, the milk that goes into four coffees a day. Easy to measure, easy to forget, and they go down without registering as food.
• Packaged food. This one is not a failure so much as a category error. The answer is printed on the box, so read it or scan the barcode. Estimating something that has a label on it is effort spent to become less accurate.

For the genuinely unknowable ones, the useful move is to estimate deliberately high rather than to skip the entry. A meal logged at plus-30 per cent keeps the day’s total roughly honest. A meal not logged at all takes the whole day out of the record, and a week with three of those tells you nothing.

A calorie target is not a fact about your body; it is a starting hypothesis you test against the scale you stand on. That changes what you need from your food log. You do not need the level to be right. You need the differences to be right — this week against last week — and a log that is consistently 12 per cent low gives you that perfectly well.

Which is why estimating the same way every day beats weighing accurately on the days you can face it. An inconsistent log is worse than a biased one, because you cannot tell a change in your eating from a change in your measuring.

When the bias breaks

The catch is that a stable bias only stays stable while your habits do. Change cuisine, start eating out more often, switch from grilling to frying, or move in with someone who cooks differently, and your error changes size without announcing itself. The tell is simple: when your logged intake and your weight trend disagree for two or three weeks straight, believe the weight trend. It is a measurement, and your log is an estimate.

At that point the fix is to adjust the target rather than to buy a better scale — the number you are aiming at came from an equation fitted to a population, so it was always going to need tuning against your own results. How that target is arrived at, and why the usual recommendation is around half a kilo a week, is covered in our guide to setting a calorie target for weight loss. Chasing the last 50 calories of measurement precision, meanwhile, buys you nothing, because the target itself is not precise to 50 calories.

A photo-based tracker is doing exactly what the rest of this article describes, faster and without asking you to remember the density of rice. Biteline reads the plate, names the dish, estimates the portion and does the nutrition arithmetic — which removes the tedium of the method, not its uncertainty. The camera cannot see the oil either. It is subject to the same four errors, in the same order.

What it does change is who does the correcting and how quickly. The portion slider sits at the top of the result because portion is where the error concentrates, and everything you learned by calibrating your hand is exactly what you use to move it: this is not 150 grams of rice, it is closer to two cupped hands, so drag it and watch the calories and macros rescale. An estimate you correct in four seconds is more accurate than one you accept, and far more accurate than a weighed log you gave up on in week three.

A reasonable division of labour looks like this:

• It has a label. Scan the barcode or type the numbers. Never estimate a known quantity.
• You cooked it. Calibrated hands, and count the thumbs of oil. You know the recipe, which is the information nothing else has.
• Someone else cooked it. Photograph it, correct the portion, accept a wider error bar, and do not spend ten minutes on a number that was never going to be exact.

You can read more about how Biteline turns a photo into a meal log on the home page, including where it is weakest and what it does not claim.