How Accurate Are Nutrition Labels? The 20% Rule and What It Means for Your Deficit

TL;DR

• Labels are estimates with a legal margin. In the US, calories and fat can run up to 20% over the label before a product is out of compliance; protein and fibre can fall up to 20% short.
• The UK and EU use a similar idea: for a food with 10-40 g of protein, carbs or fat per 100 g, the true value can sit 20% either side of the label.
• Rounding causes more trouble than tolerance. A US serving under 5 kcal can say 0, and under 0.5 g of fat can say 0 g, which is how cooking spray becomes “free”.
• Most single foods land close to their label, and random errors in both directions largely cancel across a week of mixed eating.
• The fix is not distrust. Log consistently, then judge your target by your weekly weight trend, which absorbs any steady label error automatically.

You are eating 1,900 kcal a day by your log. If every label you relied on were understating by the full 20%, you would really be eating about 2,280 kcal, and a 500 kcal deficit would shrink to around 120. That is the worry behind the question, and it is worth answering properly, because the real picture is far less alarming and the fix is the same one we use for plateaus.

Our guide to reading nutrition labels covers what each line means. This post covers how much you can trust the numbers on it.

What the rules actually allow

Food companies do not measure every batch. They calculate values from recipe databases or send samples to a lab, then print a figure that has to fall within a tolerance when a regulator checks it.

The US: the 20% rule

The FDA sorts nutrients into groups, and the direction of the allowance depends on whether the nutrient is one you want more or less of.

• Calories, fat, saturated fat, sugar, sodium: the measured value can be up to 120% of the label. Coming in under the label is never a problem for the manufacturer.
• Protein, fibre, added vitamins and minerals: the measured value must be at least 80% of the label for naturally occurring nutrients, and at least 100% where the nutrient is added.

So the rule is lopsided by design. A label that understates calories by up to a fifth, or overstates protein by up to a fifth, can still pass.

The UK and EU: tolerance bands

UK and EU guidance gives tolerances by how much of a nutrient there is per 100 g. For protein, carbohydrate, sugar and fat:

• Under 10 g per 100 g: plus or minus 2 g (1.5 g for fat).
• 10-40 g per 100 g: plus or minus 20%.
• Over 40 g per 100 g: plus or minus 8 g.

The UK band runs both ways, so it is less tilted than the US one. A chicken breast labelled 24 g of protein per 100 g could legally measure anywhere from about 19 to 29 g.

What this does not mean

A tolerance is the outer edge, not the typical error. The rules exist because natural foods vary: one tub of yoghurt is not identical to the next, and a chicken breast from one bird differs from another. Most products sit well inside the band.

Where label errors actually hide

Independent lab testing of packaged foods has tended to find most products close to their labels, with a lean towards slightly more calories than stated rather than fewer. The products that drift furthest share a few traits.

Rounding and tiny servings

Rounding is legal, predictable and often bigger than the tolerance in practice. On a US label, anything under 5 kcal per serving can be listed as 0, and fat under 0.5 g per serving can be listed as 0 g. Manufacturers set the serving size, so a one-second spray of oil becomes a zero-calorie food. Spray for five seconds and you have used several servings of “nothing”. Our guide to tracking cooking oil, butter and sauces shows how quickly that adds up.

UK labels list values per 100 g, which makes this trick harder, but the per-portion column still uses a portion size the brand chooses.

Snack foods, bakery and anything hand-portioned

A crisp packet or bar made on an industrial line is consistent. A muffin from the in-store bakery, a deli salad or a café sandwich is portioned by a person. If one sandwich has an extra slice of cheese, the label does not change. Expect wider variation on anything assembled by hand.

Restaurant and takeaway menus

Calorie counts on menus are calculated from recipes, then cooked by staff on a busy shift. A generous ladle of sauce or an extra pour of oil can move a dish by 100-200 kcal. Our eating out guide covers how to pad these entries sensibly.

Fresh food without a barcode

Loose fruit, meat from the counter and fish all vary naturally. The generic database entry is an average. For these, your bigger risk is usually the weighing, not the data; logging cooked weight against a raw entry can hide around 300 kcal a day, as our raw vs cooked guide explains.

Why your log still works

If labels can be off by a fifth, it is fair to ask why anyone bothers counting. The answer is in how errors behave across a week.

Random errors mostly cancel

Your yoghurt might run 5% high, your bread 3% low, your chicken 4% high and your rice spot on. Over 30-40 different items a week, the overs and unders largely offset each other. You are not exposed to the worst case on every food at once. The full 20% scenario in the introduction would need every product you eat to be wrong in the same direction at the legal maximum, which does not happen.

A steady error is still a steady number

Even if your overall intake did run, say, 5% higher than logged, it would run 5% higher every week. That makes your log a consistent measuring stick, just one with a slightly off zero. A consistent measuring stick is all you need to steer, because you are not using it alone.

The scale corrects for the gap

Your weekly weight trend is the real test of your intake. Set a starting target with the calculator, log for two to three weeks, then compare the trend to what you expected. If you expected 0.5 kg a week and are losing 0.3 kg, you adjust down by around 150-200 kcal. You never need to know whether the gap came from labels, your metabolism or the oil. The trend absorbs all of it. Our post on how often to recalculate covers the timing.

How to reduce label error without obsessing

You cannot audit a manufacturer, but you can avoid the places where error piles up.

Weigh, do not trust the serving size

“Serving: 30 g” is a guide, not a description of your bowl. Weigh cereal, nut butter, oils and cheese for a week. Most people find their real portions run 20-50% over the stated serving, which dwarfs any label tolerance.

Prefer the barcode to a typed search

A typed search for “Greek yoghurt” offers dozens of entries, some user-made and wrong. Chowdown’s barcode scanner pulls the exact product you are holding, which removes database error and leaves only the manufacturer’s own margin. Our logging methods comparison found barcode plus scale the most accurate combination.

Check zero-calorie claims once

If something says 0 kcal and contains oil, butter, sugar or cream, it is a rounding artefact. Weigh how much you actually use, look up the base ingredient per 100 g and log that instead.

Pad the uncontrolled meals, not the whole day

If one meal a day comes from a restaurant or bakery, add 10-20% to that entry. Leave the packaged and weighed food alone. Padding everything just shifts your numbers and teaches you nothing.

When labels genuinely matter more

For most people chasing fat loss, label error is background noise. It deserves more attention in a few cases.

Very small deficits

If you are running a 200-250 kcal deficit, a 5% steady error is a meaningful share of it. Rely on the weight trend rather than the log alone, and give it longer, three to four weeks rather than two.

High-protein targets

US protein claims can legally read up to 20% higher than reality for naturally occurring protein. If you are aiming for 1.6-2.2 g per kg and leaning on one or two products for most of it, aim for the top half of the range so a shortfall does not drop you below the useful minimum.

Narrow-range eating

If you eat the same five products every day, errors do not cancel the way they do across a varied diet. Your log will be consistently off in one direction. That is still fine for steering by the trend, but it is a good reason to rotate a few staples.

The short version

Labels are allowed to be wrong, and the US rule leans towards understating calories, but most foods sit close to their label and the errors mostly cancel. The bigger leaks are serving sizes, rounding to zero and hand-portioned food, and all three are under your control. Log consistently, weigh the calorie-dense items, scan barcodes rather than searching, and let a few weeks of weight data tell you whether your number is right.

A log that is consistently slightly wrong will still get you where you are going.

Track your weekly trend free with Chowdown →