How accurate are food scales? Resolution, minimum load and the five things that make one read wrong

The short version

For weighing food, a $20 digital scale is more precise than the food is consistent. A 1 gram step on a 150 gram portion is 0.7%; the USDA entry you are logging that portion against is itself an average across samples of that food, and the breast you cooked is not the average. Chasing a more accurate scale is almost always the wrong optimization — but there are three cases where a scale genuinely misleads you, and they are worth knowing.

Resolution is not accuracy

Three different specs get collapsed into the word "accurate", and only one of them is printed on the box.

• Resolution (graduation). The smallest step the display will move in. Nearly every scale on this site publishes 1 g: Etekcity's ESN00 at "0.1 oz / 1 g / 1 mL", the Ozeri at "graduations of 0.1 oz (1 gram)", the Luminary at a "1g increment", COSORI in "1-gram increments". A scale with 1 g resolution displays 149 or 150, never 149.4.
• Accuracy (tolerance). How far the reading may sit from the true mass. Most kitchen-scale makers do not publish one at all. OXO is the exception on this site, stating readings "within 1/8 oz or 1 g" — in other words, roughly one graduation. Greater Goods describes "one gram accuracy" and says the scale "calibrates and shuts off automatically".
• Repeatability. Whether the same 150 g gives 150 g again five minutes later, on the same spot, at the same temperature. Nobody publishes this, and it is the one you can test yourself: weigh something, lift it, put it back, repeat.

The practical consequence: a scale advertising 0.1 g resolution is not necessarily more accurate than a 1 g scale. It just displays a smaller step — which on a cheap load cell often means displaying noise. For food, 1 g resolution with a stated tolerance beats 0.1 g resolution with no stated anything.

Minimum load: the spec nobody reads

Every digital scale has a weight below which it does not pretend to read. Most makers hide it; two on this site publish it, and the numbers are higher than people expect:

Each figure is the maker's published wording except where the row says otherwise. "Not published" means the page does not state it, so this page does not guess it.

A 3 to 5 gram floor is why a kitchen scale is the wrong tool for a 5 g scoop of creatine, a gram of salt or a 3 g dose of a supplement: you are at or below the bottom of the instrument's stated range, where the reading means very little. For food portions — 30 g of oats upwards — it never comes into play.

Five things that make a good scale read wrong

All five of these produce errors larger than any kitchen scale's own tolerance.

1. An unlevel or soft surface. A load cell measures the force pressing straight down on it. On a tea towel, a cutting board that rocks, or a counter seam, part of the load is carried by the edge of the housing instead of the sensor. Bare, flat, level counter, every time.
2. Off-centre loading. A single-sensor scale reads lowest in the corners. Four-sensor designs are better at this — Greater Goods publishes four sensors — but centring the bowl costs nothing and fixes it on any scale.
3. Overhang. A plate wider than the platform is the most common real-world error, because it rests partly on the counter without you noticing. It is also the reason OXO's pull-out display exists: a big plate hides the numbers on every other scale here.
4. Tare drift across a long session. Portioning five containers means five tares, and short auto-off timers compound it: the ESN00 and EK6015 both shut the display off after 120 seconds, while the Luminary's auto-off is adjustable to 300 seconds in the app. A scale that sleeps mid-portion and comes back zeroed on the full bowl is the classic way to log a meal at half its real weight.
5. A dying battery. Low cells are the most common cause of a scale that reads consistently but wrongly. If a known weight suddenly reads 5% off, change the batteries before you blame the scale.

When 1 gram actually matters

Put the error in proportion to the thing being weighed. One gram of error is:

• 0.7% of a 150 g chicken breast portion — invisible against the natural variation between two breasts, and against the fact that the database entry is an average.
• 2% of a 50 g serving of oats — still far inside the noise of how level you scraped the cup.
• 7% of a 14 g tablespoon of peanut butter — now it is worth centring the jar properly, because fats are calorie-dense and this is where sloppy weighing shows up in a week's total.
• 20% of a 5 g creatine dose — out of scope for a kitchen scale, and at or below some makers' stated minimum anyway. Use the scoop that came with the tub.

So the honest answer to "is my scale accurate enough?" is: yes for food, no for supplements, and the bigger error in your tracking is almost certainly whether you weigh cooked or raw. That one is worth more than any upgrade: do you weigh food cooked or raw, and the method itself is on how to use a food scale for macros.

Scales that publish the numbers