Body Fat Calculator Caliper
Enter seven skinfolds and this runs the Jackson–Pollock 7-site and 3-site equations at the same time on the same folds, then tells you something no caliper chart does: how wide the band around your number really is. You get your percentage, the spread between the two protocols, the reason they disagree, and the smallest change you are entitled to call real. Everything runs in your browser; nothing is uploaded.
Seven-site total 120 mm, body density 1.0591 g/mL. The 3-site equation on the same folds returns 17.1% from 57 mm, a gap of -0.29 points.
Your three sites carry 47.5% of your seven-site total. For your age and totals the two equations agree at 48.4% — your three sites carry less than the equilibrium share, so the 3-site equation reads low. That gap is arithmetic, not a mistake you made.
Noise band ±0.69 points at 95% with 1 mm of per-site scatter, so the honest reading is 16.7–18.1%. One millimetre at a single site is worth 0.134 points here; the same millimetre at every site is worth 0.94.
Smallest change you can call real: 0.98 points — that is 7.3 mm off your seven-site total, or 0.92 kg of fat at 78 kg. Anything smaller than that is your pinch, not your body. On the 3-site protocol the threshold is looser, 1.36 points, and you would need 4 independent sets to bring this one under half a point.
3-site density (men) = 1.10938 − 0.0008267·Σ₃ + 0.0000016·Σ₃² − 0.0002574·age
then body fat % = 495 ÷ density − 450Skinfolds in millimetres, Σ is the sum of the sites, women use the corresponding female constants. These are published prediction equations, not inventions of this page. Every table below is arithmetic this page performs on them: derivatives taken numerically, per-site error propagated as an independent random error of σ millimetres, and the 3-site-versus-7-site equilibrium solved by bisection. None of it is copied from anywhere, and none of it claims to say how close these equations are to your true body fat — only how much the protocol and the instrument move the number before the equation ever sees it.
The seven sites, exactly
A caliper is only as good as the place you put it. These are the seven landmarks the equation above expects, with the direction the fold runs at each one. Two rules apply to all seven: measure the same side of the body every time — the right side by convention — and mark the spot with a skin pencil rather than re-finding it by eye.
The three sites flagged in the calculator are the ones the 3-site equation uses: chest, abdomen and thigh for men; triceps, suprailiac and thigh for women. Everything else on this page follows from those seven numbers.
The pinch and the read
Technique is where most of the error lives, and it is systematic — it pushes every reading the same way, which is worse than random scatter because averaging will not remove it.
- Measure at the same time of day, under the same conditions, before training. Skin hydration and temperature change how a fold behaves, which is why protocols fix the conditions rather than pretending they do not matter.
- Grasp the fold with thumb and forefinger roughly a centimetre apart and lift it away from the muscle underneath. If you can feel muscle inside the pinch, you do not have a skinfold — let go and take it again.
- Put the jaws on the fold about a centimetre below your fingers, at the marked line, with the jaws perpendicular to the direction the fold runs.
- Release the trigger and let the spring close. Do not squeeze the caliper shut; the spring tension is the whole point of the instrument.
- Read at the same count every time. Two seconds after release is the usual convention. The number is still falling at that point, which sounds like a flaw and is not: as long as you always read at two seconds, the drift is identical at every visit and cancels out of the comparison. Read at four seconds once and you have moved the result by the amounts in the drift table below.
- Take two readings per site. If they differ by more than one or two millimetres, take a third and record the middle value. Never discard the high one because you did not like it.
- Total the seven. Do not round the total to the nearest five — the equation is close to linear in the sum, so rounding the sum throws away information the caliper actually gave you.
Why 3-site and 7-site disagree
This is the question every caliper user eventually asks, and it has a clean answer. The two equations were fitted on different people with different site sets, so they only return the same number when the three sites happen to carry a particular share of the seven-site total. Call that share ρ. Solving for the ρ that makes the two agree gives a number that barely moves:
Men. Across every age and total computed here the equilibrium sits between 0.451 and 0.500 — your three sites have to carry roughly 47 to 48 percent of the seven-site total. The same table for women, over the identical grid, runs from 0.466 to 0.504:
Now the useful part: how far apart the two protocols land when ρ is somewhere else. At age 30, difference in percentage points, 3-site minus 7-site:
A five-point swing in ρ — which is a completely ordinary difference in where someone stores fat — moves the answer by roughly five points. That is larger than anyone's technique error and larger than the difference between a cheap caliper and an expensive one. It is the single biggest reason two honest measurements of the same person come out different, and it is invisible unless you run both protocols on the same folds, which is what the calculator above does.
What one millimetre is worth
Both equations are near-linear in the skinfold sum over the range most people occupy, so the sensitivity has a simple meaning: percentage points per millimetre of total. It is not a constant — it falls as the total rises, because of the small positive quadratic term.
Two things follow. First, the 3-site equation is roughly twice as steep as the 7-site one, so each millimetre of error hurts about twice as much there. Second, a millimetre matters far less than people assume: at a seven-site total of 140 mm, a man would need about 8 mm of real change to move one percentage point. A single bad site is not your problem — eight millimetres of distributed change is.
Two kinds of error, and only one of them averages away
Random error is the site-to-site wobble in your own hand: you pinch slightly differently each time. If each site carries an independent error of σ millimetres, the sum carries σ·√n, and that propagates through the sensitivity above. At a 180 mm seven-site total for men and a 90 mm three-site total:
Seven sites win, but by less than the arithmetic of “more data” suggests. They have √7 against √3 — 1.53 times more accumulated noise — and only claw that back because the 7-site equation is about half as steep. The net gain at σ = 1 mm is 0.427 versus 0.285 points, about a third. Worth having, not transformative.
Technique bias is the other family, and it behaves completely differently. If every site is off by the same amount — because you pinched shallow, or read late, or your landmark drifted — the error adds linearly, and the number of sites stops helping:
Look at the last two rows: one millimetre at every site costs 0.74 points on the 7-site protocol and 0.73 on the 3-site — identical to two decimal places, and the same near-equality holds for women. Adding four more sites does nothing whatsoever against a systematic error, because four more sites also means four more places for it to happen. Averaging does not help either; a bias is not noise. The only defence is the protocol itself: mark the landmark, read at the same count, lift fat rather than muscle.
Timing the read is the easiest bias to acquire by accident. If your reading decays by d millimetres per site between the two-second mark and whenever you actually look, the whole sum drops by n·d:
The size of d is a modelling parameter here, not a measured constant — it depends on your caliper, your tissue and how patient you are. What is not a modelling assumption is the shape of the result: drift is multiplied by the number of sites and then by the sensitivity, so it lands in exactly the same place as any other systematic error.
The smallest change you are allowed to believe
Comparing two measurements means comparing two noisy numbers, so the noise adds: the standard deviation of the difference is √2 times the single-visit figure. At 95% confidence the change has to exceed 1.96·√2·σBF before it means anything. Both columns below are the same calculation, expressed once in percentage points and once in millimetres of the total.
With ordinary steadiness — 1 mm per site — a man using seven sites needs 0.79 points, or about 7 mm off his total, before the drop is real rather than his own hand. Note the millimetre figure is the same for men and women while the point figure is not: the millimetres are pure noise geometry, the points go through each equation's own sensitivity.
Averaging whole sets is the way to buy precision back, and it is expensive. To drag the 95% threshold below a given figure at 1 mm of per-site scatter, the number of independent complete sets you need is:
Twenty-three complete three-site sets to resolve a quarter point is not a practical protocol, and the honest conclusion is that monthly tracking should be read to the nearest point, not the nearest tenth. Anyone showing you a caliper chart with one decimal place is reporting their arithmetic, not your body.
Siri or Brozek: the other choice nobody mentions
The caliper gives you a sum of millimetres. The equation turns that into a body density, and a second equation turns the density into a percentage. There are two standard ways to do the second step, and they do not agree. This page uses Siri throughout; here is what the alternative would have given you.
They cross at a density of 1.0614, about 16.4% body fat, and diverge in opposite directions on either side of it — Siri reads higher on fatter bodies, Brozek higher on leaner ones. Below about 12% or above about 30% the choice is worth more than half a point, which is comparable to your per-site technique error. Whichever you pick, pick it once and write it down next to your log, because half the “my numbers jumped” stories are someone switching conversion equations between visits without knowing there was a choice.
What a fold actually is
One piece of geometry explains why skinfold millimetres feel oddly large. A skinfold is a double layer of skin with the subcutaneous fat between them: the caliper is measuring down one side and back up the other. A 20 mm reading is two layers of roughly 10 mm each, and only part of that is fat — the skin itself is in there twice as well.
The halving is exact arithmetic. The skin share is deliberately left qualitative here because it varies by site and by person, and any specific millimetre figure would be invented. The practical consequence is real though: on lean bodies, where a single layer may only be a few millimetres, the skin is a meaningful fraction of what you measured, which is one reason caliper methods behave worst at the lean end. Conversely a large fold is almost entirely fat, so the same absolute error in millimetres represents a smaller proportional error — the tables above already account for this through the falling sensitivity.
Choosing a caliper
Every proper skinfold caliper does the same thing: a spring closes the jaws at a nominally constant pressure and a scale reports the gap. That constant pressure is the whole design — it is what makes your reading comparable to the readings the equations were built on, and it is why you release the trigger rather than squeezing.
The differences that exist are narrow and worth being honest about:
- Graduation. Plastic entry-level calipers are commonly marked to 1 mm or 0.5 mm; metal dial and digital models commonly read to 0.5 mm or 0.1 mm. A finer scale lets you record finer differences. It does not steady your hand, and the tables above show the hand is the limiting factor long before the scale is.
- Frame. A body that flexes under the spring adds its own error to every reading. A rigid frame does not. This is the one structural difference that plausibly matters, and it is why metal frames are standard in settings where readings are compared over time.
- Dial versus digital. Both report the same jaw gap. Digital removes parallax and the rounding you do when you glance at a needle; a dial never needs a battery and lets you watch the needle settle, which some people find makes their read timing more consistent. Read timing consistency is worth more than the display, per the drift table above.
- Cost. Nothing on this page suggests an expensive caliper buys accuracy. Per the bias table, a systematic one-millimetre error at every site moves the answer by about three quarters of a point whether you spent a little or a lot, and per the equilibrium table the choice of protocol can move it five times as far.
The one test worth doing, and it costs nothing: pinch the same site five times in a row at the same landmark, releasing fully between each. Work out the standard deviation of those five numbers. That is your σ, and it is the value the calculator at the top of this page asks for. If it comes out near 2 mm, no caliper will make your monthly comparison tighter than about 1.6 points — and if it comes out near 0.5 mm, the instrument was never your problem and you should spend your effort on landmarks instead.
Caliper questions
How do I use a body fat caliper?
Pinch a fold of skin and fat about a centimetre above the landmark, lift it away from the muscle, place the jaws a centimetre below your fingers perpendicular to the fold, release the trigger and read at the same count every time — two seconds is the usual convention. Take two readings per site and a third if they differ by more than a millimetre or two. Total seven sites and run them through a Jackson–Pollock equation, which is what the calculator above does.
How accurate is a skinfold caliper?
This page can only speak to precision, not to accuracy against a laboratory reference. What the arithmetic shows is that with a typical 1 mm of per-site scatter, seven sites carry a 95% band of about ±0.56 points around a single reading, and a change has to exceed 0.79 points to be distinguishable from noise. Systematic technique error is the larger hazard: one millimetre at every site moves the result by 0.74 points and no amount of averaging removes it.
Should I use 3 sites or 7?
Seven, if you are willing to take them, but know what you are buying. Against random per-site scatter the 7-site result is about a third quieter — 0.285 points versus 0.427 at 1 mm. Against systematic error it buys almost nothing: 0.74 points versus 0.73 for the same one-millimetre bias. Seven sites also give you the ρ check, which tells you whether a disagreement between protocols is your fat distribution or a bad pinch.
Why do my 3-site and 7-site numbers disagree?
Because the two equations were fitted on different site sets, they only agree when your three sites carry a specific share of the seven-site total — between 0.45 and 0.50 for men and 0.47 and 0.50 for women across every age and total computed here. Outside that band the gap opens fast: at a 180 mm total, moving that share from 0.40 to 0.55 swings the difference by about 6.8 points for men. The calculator above reports your share and the equilibrium for your numbers.
How many millimetres is one percent of body fat?
It depends on your total. At a seven-site sum of 140 mm a man needs roughly 8 mm of change for one point; at 100 mm the sensitivity is 0.142 points per millimetre, so about 7 mm. The sensitivity falls as the total rises, so leaner people move further per millimetre than the table implies at the top end.
How often should I measure?
Monthly is the sensible cadence. With 1 mm of per-site scatter the smallest believable change at 95% is 0.79 points, and losing 0.79 points of body fat takes a realistic amount of time; measuring weekly mostly measures your hand. Measure at the same time of day, before training, under the same conditions.
Can I measure myself, or do I need a partner?
Most sites are reachable alone — chest, abdomen, thigh, suprailiac and triceps all are. The subscapular and midaxillary folds are the difficult ones, because reaching for them moves the skin you are about to pinch. If you measure alone, do it alone every time; switching between self-measured and partner-measured introduces a systematic difference that will look exactly like progress or exactly like a plateau.
My caliper and my smart scale disagree. Which is right?
Neither is a reference. A caliper measures the thickness of a pinch at specific landmarks and infers the rest from an equation; a smart scale sends a current through you and infers composition from an electrical signal that moves with hydration. They fail in different ways, which is why the spread between them is not evidence that one is broken. Our BMI and smart scale page covers what those devices actually measure, and our fat percentage page runs several methods side by side.
Where exactly does the caliper go on the abdomen and thigh?
Abdomen: a vertical fold about two centimetres to the right of the navel, not at the navel itself. Thigh: a vertical fold on the front midline, midway between the hip crease and the top of the kneecap, with your weight on the opposite foot so the muscle underneath relaxes and the fold lifts cleanly. Both are listed with their failure modes in the seven-site table above.
Do I need an expensive caliper?
The arithmetic here does not support that. Graduation finer than 0.5 mm records detail your pinch does not contain, and the dominant errors — landmark drift, reading timing, lifting muscle with the fat — are identical on any caliper. A rigid frame and a scale you can read consistently are worth paying for; extra digits are not.
Related tools
body fat calculator · fat calculator with error budget · measure body fat percentage · how to measure body fat at home · fat percentage across three methods · Navy tape method · body fat percentage chart
Not medical advice. Every figure on this page is arithmetic this page performs on published prediction equations. It describes precision lost in the instrument and the protocol; it says nothing about how close those equations are to the truth for you. See our disclaimer.