Key takeaways
- Adults under-report their intake by roughly 20–30% on average when measured against doubly labelled water — and a scale is not the only fix.
- Almost all of the error sits in six calorie-dense foods: oil, butter, nut butter, cheese, dressings and dry grains.
- One tablespoon of olive oil is about 120 calories, so a three-tablespoon drizzle adds 360 without looking like a portion at all.
- Hand portions are accurate enough to act on: a palm is 20–30 g of protein, a thumb about 10–12 g of fat.
- A consistent bias can be corrected from two to three weeks of weight data. Random error cannot.
You can track calories usefully without ever owning a food scale. Scan barcodes and labels for packaged food, use hand-size references or an AI photo scan for everything else, and measure only the six or so calorie-dense foods that carry most of the error. The goal is not a perfect number — it is an estimate that is wrong in the same direction every day, because a consistent bias can be corrected and random error cannot.
How wrong are eyeballed portions, really?
Wrong enough to matter, and reliably in one direction. When researchers compare what people say they ate against energy expenditure measured with doubly labelled water — the reference method — adults under-report intake by roughly 20–30% on average. The gap widens with larger portions, with calorie-dense foods, and with foods people feel uneasy about logging.
20–30%
Typical under-reporting
Self-reported intake vs doubly labelled water, across adult populations
30–50%
Portion underestimation
How much smaller people judge an unmeasured serving to be than it is
10–40%
AI photo scan error
Typical range on a mixed plate — comparable to careful eyeballing, in seconds
The useful detail is where that error concentrates. It is not spread evenly across the plate. Misjudging a portion of broccoli by half costs you fifteen calories; misjudging the oil it was roasted in by the same proportion costs you a hundred and twenty. Tracking without a scale works when you spend your precision where the calories are dense.
Which foods actually need measuring?
Six categories carry most of the error, and they share a trait: they are added to food rather than eaten as a meal, so no natural portion size applies.
| Food | One serving | Calories | Why it goes wrong |
|---|---|---|---|
| Olive oil, butter | 1 tbsp | ~120 | Poured, never portioned — a pan drizzle is often 2–3 tbsp |
| Peanut or nut butter | 1 tbsp | ~95 | A heaped spoon is closer to two level ones |
| Hard cheese | 30 g | ~120 | Grated cheese has no visual reference at all |
| Salad dressing | 2 tbsp | ~130 | Restaurant portions run to 4 tbsp on a side salad |
| Dry rice, pasta, oats | 100 g dry | ~350–380 | Cooked volume triples, so the cooked portion tells you nothing |
| Cream, oat milk in coffee | 60 ml | ~30–120 | Repeated three times a day without being logged once |
A set of measuring spoons and a dry measuring cup covers every row here and costs less than a scale.
Everything else — vegetables, fruit, lean meat, eggs, bread by the slice, yogurt by the pot — can be estimated by eye or read off the packaging without meaningfully damaging your log. That is the trade the no-scale approach is making, and it is a good one.
How do hand portions work?
Hand portions are a body-scaled reference: bigger people have bigger hands and generally need more food, so the measure adjusts itself. They are the fastest way to put a number on a plate you did not prepare.
| Reference | Food | Roughly equals |
|---|---|---|
| Palm (thickness of your hand) | Cooked meat, fish, tofu | 85–110 g — 20–30 g protein |
| Cupped hand | Cooked rice, pasta, potato | 20–30 g carbohydrate |
| Fist | Vegetables, salad | About 1 cup |
| Thumb (first joint to tip) | Oil, butter, nut butter, cheese | 10–12 g fat — about 100 calories |
| Tip of the thumb | Oil or butter on a pan | About 1 tsp — 40 calories |
A woman's hand and a man's hand give different absolute amounts, which is the point — the reference scales with the person using it.
Household objects do the same job for volume when hands are inconvenient: a baseball is about one cup, a tennis ball about half a cup, a deck of cards is roughly 85 g of cooked meat, and a golf ball is about two tablespoons.
Let the camera estimate the portion
2BIB reads a photo of your plate and returns calories and macros with a portion estimate you can adjust in one tap — plus barcode and nutrition-label scanning for anything that came in a package.
What do you use for packaged and restaurant food?
For anything with a barcode, no estimation is needed at all: the manufacturer has already weighed it. Scan the code, then check one thing — whether the panel is per serving or per package. A pint of ice cream is three to four servings, a bag of chips is two, and a bottled smoothie is often two and a half.
Restaurant meals have no weighable recipe in the first place, so a scale would not help even if you carried one. Chains publish nutrition data, which is the best source available; independent restaurants do not, so estimate by composition instead. Count the protein by palms, count the visible starch by cupped hands, and then add fat for the cooking you could not see — a restaurant portion typically carries one to two extra thumbs of oil or butter compared with the same dish cooked at home.
The repeat-meal shortcut
Most people eat the same fifteen or twenty meals on rotation. Estimate one carefully, save it, and reuse it. After two weeks of this, the majority of your logging is a two-tap job and the accuracy of those entries stops drifting, because you are copying a past decision rather than making a new guess.
How do you calibrate without ever weighing anything?
Use your weight trend as the measuring instrument. Your body integrates every calorie you actually ate, which no logging method can do, so it is the only honest feedback available outside a metabolic ward.
- 1Estimate the same way for two to three weeks. Consistency is the whole mechanism. Changing methods mid-stream makes the data unreadable.
- 2Compare weekly averages, not daily weights. Day-to-day swings of 1–2 kg are water, sodium and gut contents. The weekly average is the signal.
- 3Read the direction and the speed. Losing 0.5–1% of body weight per week means your estimate is working. Flat for three weeks means it is not.
- 4Adjust the target by 100–200 calories, then wait. Do not back-calculate an exact figure from the weight change — the 3,500-calorie-per-pound rule overstates what actually happens over time.
This loop corrects a consistent bias completely. If you under-report by 15% every single day, your target simply lands 15% lower than the theoretical one, and the result is identical. What it cannot correct is inconsistency: scanning some meals, guessing others, and skipping weekends produces random error that no amount of weight data can untangle. Weekends are the classic offender, and the other seven ways a deficit quietly disappears are worth knowing before you blame your estimates.
What does a no-scale day look like?
A worked example, with the measuring effort concentrated in the two places it pays.
Breakfast — scanned
Lunch — estimated
Dinner — photographed
Extras — remembered
Total measuring time: under a minute, spent on oil and granola. That is the whole discipline. If you want the method comparison behind these choices, AI scan, barcode, or manual entry ranks the four options on speed and accuracy, and how to calculate your macros covers what to do with the numbers once they are in.
When is a scale worth buying anyway?
A scale earns its place in three situations, and none of them is general weight loss.
- You cook in batches. Weighing a whole recipe once and dividing by portions is faster than estimating each serving for a week.
- You are recomposing at low body fat. The closer you get to lean, the smaller the margin for error becomes.
- You want to recalibrate your eye. Two weeks of weighing teaches you what 30 g of cheese and 150 g of rice look like, after which you can put the scale away and keep most of the accuracy.
Outside those cases, the log that gets kept beats the log that is precise. A method you will still be using in week six — whether that means hand portions, an AI food scanner, or a barcode for everything — produces better results than a weighed diary abandoned on day nine. Pair it with a realistic target from the TDEE calculator and a plate built on balanced nutrition, and the missing decimal places stop mattering.
If tracking stops feeling neutral
Measuring food is a tool, not a moral exercise. If logging is becoming compulsive, if skipping a meal feels safer than estimating it, or if you have a history of disordered eating, step back from tracking and speak with your doctor or a registered dietitian. Balance, not perfection.
The short version
Scan what has a barcode. Measure oil, nut butter, cheese, dressings and dry grains. Estimate everything else with your hand or a photo. Keep the method the same for three weeks, watch the weekly average weight, and move your target by 100–200 calories when the trend disagrees with the plan.