Square-Face Brow Chart

The bend at the peak set against the corner of the jaw underneath it. Six sections: the two angles side by side, peak rise at six brow lengths with the fractions to multiply by, the band from 126 to 144 degrees with the jaw marked inside it, the method for rounding the front corner, four ways to lower an arch that remove no hair, and a log to fill in before anything comes out. Print it or keep it on your phone.

Two sets of material are referred to below and neither is named here, only dated. The nine are the nine published positions on this face's brows that carry a date, running from 10 November 2013 to 23 March 2026, listed oldest first in the article this sheet belongs to. The pool is the lay advice — people answering other people, between 2016 and 2025 — which is where almost all of the method on this sheet comes from, and which the nine never cite. Where the two agree without contact, that is said. Where they contradict each other, that is said too.

1. The two angles, side by side

This is the only outline in the set whose brow advice reduces to a comparison between two angles, because it is the only one with a corner underneath the brow sharp enough to compare against. One angle is the bend at the peak of the brow, between the inner run and the tail. The other is the corner of the jaw below the ear, where the outline turns instead of curving. Both are interior angles, both can be drawn, and the entire instruction for this face is one inequality: the brow's bend has to be more obtuse than the jaw's corner. Drawn below at the same scale, so the difference in openness between them is the real difference.

the two ends level A = 135° the bend at the peak inner start tail end 127.9° the corner of the jaw up toward the ear forward to the chin the gap between them is 7.1 degrees
The two angles, what each one is, and what the record gives for each. Only one of the two is adjustable.
Angle Where it sits What is known about it Figure
The jaw's corner Below the ear, where the ascending part of the jaw meets the part that runs forward to the chin. The sharpest of the seven outlines this site classifies. It is bone, and nothing on this sheet changes it. It is the reference the brow is measured against. 127.9°
The brow's bend At the peak, between the inner run — the medial two thirds — and the tail, the lateral third. Measured exactly once in the published record: 60 people positioning their own ideal brow with a finger across five photographs each, in July 2017. The mean with its spread attached. 135° ± 9°
The difference Nowhere. It is not a place on the face; it is this sheet's arithmetic. No source in the twelve years of published advice on this face subtracts one from the other. The subtraction is what turns "soften it" into an instruction with a threshold. 7.1°

Two honest limits on the 135. It is not a square-face figure: the study never sorted its 60 by outline, and the words face shape do not appear anywhere in it, so the mean is what people asked for rather than what any shape needs. And the method was a finger on a photograph, which makes it a preference, not an optimum. What earns it a place here is the spread — it is the only brow angle in the record that comes with one, and a spread is the only thing that turns a single number into a band.

One honest limit on the 127.9. This site publishes that the jaw angle is the least reliable of its five features to read from a photograph, because a corner that turns is easy to misplace when the head is tilted or the lighting flattens the jaw, and that the rule set weights it below the ratio features for that reason. So measure it with a tape rather than trusting a photograph of yourself: three points — one below the ear at the corner, one up the ascending part, one forward along the jaw — and the angle between the two lines they define. If you cannot get it, use 127.9 and treat the answer as approximate.

2. Peak rise at six brow lengths

The model, stated so it can be checked. Take the brow as two straight runs meeting at the peak, with the two ends level — which is what every source assumes without saying, since the tail is forbidden from dropping below the inner start and the level case is the boundary. Take the inner run as two thirds of the length and the tail as one third, which is the split the 2017 measurement's own definition of the apex angle uses. Call the length L and the bend A. Because the geometry scales, the rise is a fixed fraction of L for each A, and the fraction is the part that transfers to a brow that is not 60 mm. Multiply your own L by the fraction in the column heading.

Rise of the peak above the line joining the two ends, in millimetres and inches, at the four angles that matter and six brow lengths. The fraction beside each angle is rise ÷ brow length; multiply it by any length not in the table.
Brow length L 144°
× 0.136542
135°
× 0.168483
127.9°
× 0.192640
126°
× 0.198925
45 mm = 1.77 in6.14 / 0.247.58 / 0.308.67 / 0.348.95 / 0.35
50 mm = 1.97 in6.83 / 0.278.42 / 0.339.63 / 0.389.95 / 0.39
55 mm = 2.17 in7.51 / 0.309.27 / 0.3610.60 / 0.4210.94 / 0.43
60 mm = 2.36 in8.19 / 0.3210.11 / 0.4011.56 / 0.4611.94 / 0.47
65 mm = 2.56 in8.88 / 0.3510.95 / 0.4312.52 / 0.4912.93 / 0.51
70 mm = 2.76 in9.56 / 0.3811.79 / 0.4613.48 / 0.5313.92 / 0.55
mm — yours    

A length that is not in the table, worked through. On a 52 mm brow at the measured mean: 0.168483 × 52 = 8.7611, so 8.76 mm. Then 8.76 ÷ 25.4 = 0.3449, so 0.34 in. Every conversion on this sheet is exact division by 25.4, not an approximation — the international inch has been defined as exactly 25.4 mm since 1959.

Going the other way, from a rise you have measured to the bend you actually have, there is a trap. The run you take with a tape is along the brow, which makes it the hypotenuse, so the slope of each run is arcsin(rise ÷ run) and not arctan(rise ÷ run). The bend is then A = 180° − arcsin(rise ÷ inner run) − arcsin(rise ÷ tail run). Check it against the table at L = 60 and the mean: rise 10.11, runs 40 and 20, so 180 − 14.64 − 30.36 = 135. Using arctan on the same three figures gives 139.0 — four degrees out, which on an 18-degree band is nearly a quarter of the whole acceptable range. At the sharp edge the arctan form is wrong by 6.6 degrees.

Near the mean the sensitivity is about 0.21 mm per degree on a 60 mm brow: nine degrees either side of 135 buys 1.83 mm going sharper and 1.91 mm going flatter. That is the practical size of the whole argument, and it is also why one degree is not worth chasing. No pencil, no stencil and no mirror resolves one degree, and 0.21 mm is under the width of a hair.

3. The band, marked

Four published sources forbid the two ends of this band without ever putting a number on either. Too rounded. Too flat. Overly sharp. A full semi-circle. Thin strings. Wide horizontals. Six prohibitions and no thresholds, which is why they cannot be checked against each other. The band below is numbered from the only measurement in the record — 135 as the centre because it is the one brow angle anyone has measured, ±9 as the width because that is the measured spread and there is no other candidate. The edges are then 144 and 126. The jaw's own corner sits inside it, near one end, and its position is the useful part.

11.94 mm 11.56 10.11 8.19 mm 126° 127.9° 135° 144° sharp edge the jaw's corner measured mean flat edge 3.74 mm / 0.15 in — the whole band, 18 degrees 1.45 mm / 0.06 in — jaw to mean 0.38 mm — the shaded tenth, ruled out
The four positions in the band, with each one's share of it counted from the sharp end in degrees. Rise figures are for a 60 mm brow; use the fractions in section 2 for any other length.
Position Angle Rise at L = 60 Share of the band What it means
Flat edge 144° 8.19 mm / 0.32 in 100% The bend is very nearly gone — 36 degrees short of straight. Two sources ban this. It is still not horizontal, which is a different thing.
Measured mean 135° 10.11 mm / 0.40 in 50% Dead centre. Seven point one degrees more open than the jaw underneath it, which is 0.79 of the study's own standard deviation.
The jaw's corner 127.9° 11.56 mm / 0.46 in 10.6% The threshold. Above it you are echoing the corner; below it — more open than it — you are doing what every source tells this face to do.
Sharp edge 126° 11.94 mm / 0.47 in 0% Two other sources ban this end. Between here and the jaw's corner is the tenth of the band that is ruled out twice over.

Read the shading and the two prohibitions turn out to be one. Do not echo the jaw and do not go too angular are the same instruction in two wordings, and between them they rule out 1.9 degrees of the 18 — a tenth of the band. Everything the record tells this face to avoid on the sharp side is avoided by a single move: open the bend until it is more obtuse than the corner underneath it. On a 60 mm brow that move is worth at most 0.38 mm of peak height. It is the smallest instruction on this sheet and the only one that is load-bearing.

The flat edge is where the record contradicts itself, and the contradiction resolves. A 2023 guide offers a flat but upward tilted brow as one of its four styles for this face, on the condition that the overall brow angle still tilts upward, and notes it can suit a fairly low forehead. The largest body of advice in the pool says the same thing from the other direction: even a straight brow should run upward rather than horizontal, because a horizontal line reads as sad on a resting face. Two pools, no contact, one instruction. Flat is available to this face. Horizontal is not. What separates them is not the bend at the peak but whether the tail sits above, level with, or below the inner start.

What the band does not describe is thickness, and the record does not settle it either. Strong and thick, from 2014. Thick, angular and well defined, from 2020. Medium or slightly larger in width, from 2021. A fuller brow without straight lines, from 2023. Kept looking full, from 2026. And in the pool, one person stating the actual trade instead of a recommendation: when the brows are thinner the eyes stand out, when they are thicker the two balance each other, so the question is whether you want your eyes to stand out or the face to look harmonious. All of it is compatible with the band, because the band measures the bend and not the mass inside it.

4. Rounding the front corner

A brow has two corners and this face's advice never separates them. There is the corner at the peak, which is what section 3 measures, and the corner at the front, where the brow turns up out of the nose. The squareness people complain about lives in the second one. A 2023 guide warns against a blocky shape at the front of the brows, especially when filling them in, saying it reads harsh on anyone but especially on this face. A person asking for help in 2016 said the same thing in their own words: they did not like how their brows were squared off at the beginning, and had stopped filling them in because the result looked too harsh.

The two faults that both get called square, and the fix for each. They are different corners and the fixes are not interchangeable.
What people type Which corner What the fault actually is What fixes it
Square shaped eyebrows The front A fault, not a goal. The brow turns a right angle where it rises out of the nose, so the front reads as a block — worse once it is filled in. Round the top edge of that corner and leave the bottom edge square. Section 4's steps.
A square arch The peak A different fault. The bend is a corner rather than a curve, usually because the brow thickens abruptly at the arch until the whole shape reads as a semicircle. Hold the thickness even through the bend, and open the angle past the jaw's corner. Sections 3 and 5.
  1. Run the resting test first. Leave the brows unbrushed and unset and look. If the arch is still visible at rest, the bend is in the hair rather than in the styling, so the peak corner is already a curve and the front corner is the one to work on. If it is not visible, the reverse is true and section 5 applies first. This is the only criterion anywhere in the record for the thing section 3 measures, and it takes two seconds in a mirror.
  2. Keep the bottom inner corner square. The vertical line at the front is what stops the brow looking drawn. Removing it is the common mistake and it is the one that cannot be undone.
  3. Round the top corner only. One of the corner's two edges loses its right angle; the other stays. This is the whole of the only two-corner instruction in twelve years of material on this face, and it is one sentence long.
  4. Train the hairs up and do not trim them. Let the wispy ends extend past wherever you fill in. It softens the outline by blurring it rather than by removing anything, which makes it the second reversible move in this section.
  5. Do not fill the front into a block. Filling is what turns a slightly square front into a visibly square one. If the front needs colour, put it above the existing outline rather than square into the corner — which is move three in section 5 as well.
  6. Taper the tail rather than stopping it. The tip can run a little longer than usual and short lines should be avoided; a tail that ends abruptly puts a hard line back into a brow whose other two corners you have just softened. Length here means the length of the taper, not the length of the brow across the face.
  7. Look at it for a week before anything comes out. Steps two to five are all colour, direction or nothing. Step six on the next section is the first one that is not.

One shape the record cannot agree on, and the disagreement is resolvable. A 2021 academy guide prescribes a semicircular soft arch for this face. A 2023 brand guide forbids a full semi-circle, on the grounds that it is liable to backfire and clash with your features, and recommends a fuller brow without straight lines with just a hint of rounding at the arch instead. The largest body of advice in the pool uses the semicircle as the name of a mistake — the brow that thickens at the arch until it looks like one. Three positions, one shape. The difference between them is thickness rather than curvature: a semicircular outline drawn at an even thickness is what the 2023 guide means by a hint of rounding, and the same outline drawn with the mass piled at the peak is what the other two are describing. None of the three says so.

5. Four ways to lower an arch that remove nothing

Opening a bend has a limit that closing one does not, which is why the reversible moves come first and why there is a ceiling on them. These four remove no hair, so each one costs an evening and each one can be run again with a different setting. Do them in this order, because they are listed from least to most committed.

The four no-removal moves, what each one does to the bend, and how to undo it. All four wash off.
# Move What you do Why it lowers the bend
1 Set direction Set the brow with a wax, then comb the front section upward rather than to the side. Hairs standing up fill in the space above the front run, which takes the angle out of it. On its own this reduces the arch angle, without a colour product involved.
2 Colour placement More emphasis on the top front portion of the brow, less on the bottom. Moves the visible upper edge up at the front, so the bend reads gentler than the hair underneath it actually is.
3 Draw above the outline Take the colour onto the skin above the brow, past where the hair stops. The only one of the four that can go beyond the natural limit, because it is not bounded by where the hairs grow. It is also the easiest to overdo.
4 Fill below the brow Add colour under the existing lower edge of the brow. Raises the line the peak is measured from, so the same peak reads lower. In the record this move appears paired with removal at the top; the filling half works on its own and is the half to test first.

The ceiling. Once those four are exhausted there is nothing left that removes nothing, and the record says so plainly — that there is no way to make a strong natural arch straighter other than to draw it in at the arch, and that the alternative would be a procedure along the lines of implanting hair into a beard or a hairline, which the person describing it calls a bit extreme. Anyone promising to flatten a strong arch into a straight brow by removal alone is promising something that cannot be done.

The cost rule, stated twice by two people in two different years who agree. Shaving is safer than plucking, because shaved hairs grow back and plucked ones can stop. One of the two wrote it from experience — they had plucked their own ends, were growing them out, and reported one brow looking a bit sad as a result. The other gives the same rule for a different reason: styles and faces change over time. Neither of the nine published sources on this face mentions reversibility once. The only maintenance figure any of them prints is touch-ups every few weeks, which tells you how often to go back and not what happens if the shape you went back for was wrong.

The order is not caution. Opening a bend means either taking height off the peak or adding height below it, and only one of those two is undoable in an evening. Work from the top with colour first, since colour is the reversible half of the same move. If the result is right, the removal version of it is a small bounded amount of hair in a place you have already confirmed. If it is wrong, nothing has come out and the next attempt costs the same as the last. That is the only sequence in which the third attempt is as informative as the first.

6. Log to fill in before anything comes out

One row per candidate shape, and the two columns that matter are the fourth and the fifth. The fourth is the resting test from section 4, run before and after rather than once. The fifth is a verdict written a week later, not on the day — the reason is a named failure of judgement that two people in the pool use the same term for independently, eyebrow blindness, which means you cannot see your own brows accurately because you have seen them every day for years. On a band 3.74 mm wide, judged by an instrument documented to misread its own result, a same-day verdict is not a verdict.

Eight candidate shapes, in the order to try them — reversible first. Write the date, run the resting test, and leave the verdict for a week.
Candidate shape How you saw it Date Visible at rest? Verdict, a week later
Combed up, no colourSection 5, move 1   
Colour on the top frontSection 5, moves 2 and 3   
Filled below the lower edgeSection 5, move 4   
Softly archedOverlay or stencil   
Arched closer to the tailOverlay or stencil   
Rounded, even thicknessOverlay or stencil   
Angled, not harshOverlay or stencil   
Flat but tilted upwardOverlay or stencil   

Three ways to see a shape before committing to it, and their limits. An overlay: there are filters that lay a guide over the face, and one person reports that it almost always looks too far, too thick or too arched on screen until you actually do it, at which point it looks right — so the overlay is calibrated for the wrong thing, but it is still the only way to see four shapes in a minute. A stencil: the same method with plastic in it, and a 2024 product guide recommends a kit of ten shaped stencils, which is the only countable number in twelve years of published advice on this face. A trace: two of the nine give a method for finding the outline of the face itself — hair pulled back and the outline read in a mirror, or the face traced with a lipstick or a liner to measure the width between the widest points. Neither trace is for the brow. Both are for the thing under it, and both belong in section 1's fields, not here.

One more reason the log has a week in it rather than a mirror. The person who asked about a blocky face in 2025 said, in passing, that they expected their own features to change over the following months. That is the situation the shape-name advice never accounts for: a decision taken against a moving target. And the published advice makes it worse without meaning to, because one source conditions the brow on the hair — you might look better with sharper brows if the hairstyle is doing the softening — which means a haircut can invalidate a brow you chose on the strength of a guide. Write the date down so you can tell which of the two changed.