BMI is height and weight, arranged into a single number. It is free, instant, and requires no equipment, which is why it is everywhere.

It also cannot tell muscle from fat, or where fat sits, and when researchers have checked it against a proper body composition scan the disagreement is larger than most people expect.

Key takeaways

• In a study comparing BMI with DXA scans, BMI classified 26% of adults as obese while the scan found 64% (Shah and Braverman 2012, PLOS ONE).

• 39% of people BMI called non-obese were obese by body fat measurement.

• Misclassification hit 25% of men and 48% of women.

• It remains genuinely useful for populations and for tracking your own change. It is weakest as a verdict about one person at one moment.

What is BMI actually measuring?

Your weight divided by the square of your height. That is the entire calculation.

It says nothing about what the weight is made of, or where it is carried. A muscular person and a sedentary person of identical height and weight get identical scores, as do two people with the same total fat distributed very differently.

It was developed in the nineteenth century to describe populations, not to diagnose individuals, and that origin explains most of its strengths and all of its weaknesses.

How often does BMI get it wrong?

More often than its ubiquity suggests, and unevenly by sex.

A cross-sectional study compared BMI against DXA scans, the reference method for body composition, in adults with a mean age of 51 (Shah and Braverman 2012):

39% of participants were called non-obese by BMI but were obese by body fat measurement. Misclassification ran at 25% in men and 48% in women.

That sex gap is the part worth carrying. Women in that study were misclassified nearly twice as often as men, and the errors increased with advancing age. Women naturally carry a higher proportion of body fat at the same BMI, and BMI has no way to account for it.

Note the direction, too. These were false negatives: people reassured by a normal-looking BMI who had a body fat percentage that said otherwise. The popular version of this critique is usually about muscular athletes being wrongly labelled overweight, which is real but far rarer.

So is BMI useless?

No, and the case against it is often overstated in the opposite direction.

It works well for what it was built for: describing populations, tracking trends across large groups, and providing a cheap first-pass screen. Across thousands of people, BMI and body fatness correlate strongly enough to be informative.

It is also perfectly reasonable as a personal trend line. If your height is constant, a BMI that moves is a weight that moved, and comparing your own BMI over time avoids nearly all of the between-person problems. That is the compare-you-to-you principle again, and it rescues a surprising number of imperfect metrics.

Where it fails is exactly where people most want to use it: as a single verdict about one individual on one day.

What should you look at instead?

Anything that distinguishes what the weight is made of, or where it sits.

Waist measurement. A tape measure costs almost nothing and captures where fat is carried, which BMI cannot. Abdominal fat behaves differently from fat elsewhere.

Body fat percentage, with realistic expectations about the measurement. Home scales estimate it by passing a small current through you, and the error is substantial: bioimpedance underestimated body fat by around 4.6 percentage points in men and 5.1 in women against DXA, with limits of agreement spanning more than 14 points (Li et al. 2013, PLOS ONE). We go into what that means in body fat percentage.

Your own trend in anything. A stable measurement moving in one direction over months tells you more than a precise measurement taken once.

How you actually function. Fitness independently predicts outcomes, and people who were overweight but fit had risk similar to normal-weight fit people in pooled analysis.

What about the BMI categories?

Treat the boundaries as soft, because they are.

The cutoffs are round numbers chosen for convenience, and nothing physiological changes as you cross 25.0 or 30.0. Several bodies now recommend different thresholds for some populations, since the relationship between BMI and body fat varies by ancestry as well as by sex and age.

The practical reading: a BMI near a boundary tells you very little, and a BMI deep into a category is more informative than one that is close. Either way it is a prompt for a better measurement rather than a conclusion.

Frequently asked questions

Is BMI accurate?

For populations, reasonably. For individuals, often not. Compared with DXA scans, BMI identified 26% of adults as obese while the scan found 64%, and 39% of those BMI called non-obese were obese by body fat (Shah and Braverman 2012).

Why is BMI worse for women?

Because women carry a higher proportion of body fat at the same height and weight, and BMI cannot distinguish tissue types. In the same study, BMI misclassified 48% of women against 25% of men, with errors increasing at older ages.

Does BMI work for muscular people?

Poorly, though this is the less common failure. Muscle is denser than fat, so a muscular person can score as overweight while carrying little fat. The more frequent error in the data is the opposite: people with normal BMI and high body fat.

What is a better measure than BMI?

Waist measurement is the cheapest useful addition, since it captures where fat sits. Body fat percentage is more informative in principle but is measured imprecisely at home. Fitness is a strong independent predictor. Your own trend in any of these beats a single reading of all of them.

Should I ignore my BMI?

No, but read it as a rough screen rather than a diagnosis. Its most defensible personal use is as a trend: since your height is fixed, changes in your BMI over months are simply changes in your weight, and comparing you against yourself sidesteps most of what makes the measure unreliable between people.

Knit shows patterns, not diagnoses. Talk to a clinician about readings that concern you.