One diagonal proves nothing
The standard advice is to measure the diagonals and make them match. The part that gets dropped in the retelling is that there are two of them and both matter. Take a rectangle and push one corner sideways, keeping all four side lengths the same. You now have a parallelogram: every side is still the length it should be, but one diagonal has grown and the other has shrunk. If you measure only the one that grew, and it happens to land near the number you calculated, the shape passes a check it should have failed spectacularly.
Two equal diagonals, with both pairs of opposite sides equal, is the condition that actually forces a rectangle. It costs one additional tape pull. Everything downstream — foundation walls, framing, floor squareness, whether a sheet of plywood lands on a joist forty feet away — inherits whatever this layout got wrong, and layout errors do not shrink as the building goes up.
Where the diagonal number comes from
For sides a and b, the diagonal is the square root of a squared plus b squared. A 24 by 32 rectangle gives 576 plus 1,024, which is 1,600, whose square root is exactly 40 feet. That is a pleasant coincidence of a 3-4-5 family shape scaled by eight, and it is why those dimensions turn up so often in examples. Most real buildings give an ugly decimal, which is why the result here is shown in feet and inches to a sixteenth as well as in decimal feet.
The 3-4-5 method is the same identity at a smaller scale. Mark three units along one line, four units along the line that should be perpendicular, and when the distance between the marks is exactly five units the corner is a right angle. The units are arbitrary, which is the useful part: scale it to whatever your tape and the ground allow. On anything larger than a shed, 6-8-10 or 9-12-15 will beat 3-4-5, because a marking error at the triangle gets multiplied as the line is extended out to the real corner.
Correcting a rectangle that will not close
| Symptom | What it means | What to do |
|---|---|---|
| Diagonals differ, sides correct | Racked — a parallelogram | Shift a corner sideways by roughly half the difference, toward the short diagonal, holding both side lengths |
| Diagonals equal, both off the target | Square, but the wrong size | Re-check the side measurements; the shape is fine, the dimensions are not |
| Diagonals differ and neither matches | Both size and square are out | Fix the two sides against a reference line first, then square it |
| It closes on flat ground and not on slope | Tape is measuring slope distance | Hold the tape level and plumb down, or all your horizontal dimensions are long |
Half the difference is a starting move, not a formula that closes in one go. Move, re-measure both diagonals, move again. It converges quickly if you resist the urge to make one large correction, and it converges not at all if you let the side lengths drift while you are adjusting.
The errors that squaring will not catch
Tape sag on a long pull reads long, and the effect is worse on a hot day with a heavy tape held loosely. Measuring along sloping ground rather than horizontally reads long as well, and on a site with real fall that error can be substantial — the tape has to be level and the point plumbed down, every time. Hooking the tape on a nail set slightly off the mark repeats the same small error into every measurement taken from that corner. And a layout that is perfectly square can still be in the wrong place entirely, which no diagonal will ever reveal.
Set batter boards back from the corners, far enough that the excavator can work without touching them, and square the offset rectangle as well as the building rectangle — the strings between the boards are only as square as the boards are. Mark string positions with a saw kerf rather than a pencil. Once the layout closes, the excavation volume calculator gives the dirt that comes out of it, the concrete footing calculator handles what goes back in, and for corners that are not right angles at all the triangle calculator solves the general case.
Questions people ask
Why does the 3-4-5 method work?
Because 3 squared plus 4 squared equals 5 squared: 9 plus 16 is 25. The converse of the Pythagorean theorem says that if the three sides of a triangle satisfy that relationship, the angle opposite the longest side is exactly 90 degrees. So marking 3 and 4 along two lines and forcing the distance between the marks to 5 forces the corner square. The numbers can be any unit and any scale — 6-8-10, 9-12-15 and 15-20-25 are the same triangle enlarged, and larger is better because it reduces the effect of a small marking error when the line is extended.
How close do the two diagonals need to be?
It depends entirely on what is being built on the layout, which is why the tolerance is an input here rather than a fixed number. A garden shed and a foundation that will receive engineered floor trusses are not held to the same standard, and a slab that will take a tiled finish is fussier again. What is worth knowing is that layout error does not stay put: it propagates into wall lines, into the way sheet goods land, and into every measurement taken from a corner. It is far cheaper to spend twenty minutes closing the diagonals than to discover the consequence at the roof.
Which corner do I move when the diagonals do not match?
Take the corner at the end of the long diagonal and shift it sideways toward the side that would shorten it, by roughly half the difference between the two diagonals, holding both adjoining side lengths with tapes while you move it. Then re-measure both diagonals. Half the difference is a first move rather than an exact solution, and two or three small iterations close it faster than one confident large correction. The critical discipline is holding the side lengths, because it is easy to square the shape while quietly making it the wrong size.
Does this work for shapes that are not rectangles?
Only in pieces. An L-shaped or T-shaped footprint is normally laid out as a large rectangle that is then squared, with the notches measured off established square lines rather than laid out independently. Squaring each piece separately tends to accumulate error at the junctions. For a genuinely non-rectangular corner, or when you need to lay out a specific angle rather than a right angle, the general triangle solution is the tool. Nothing here handles curves, which are laid out from a centre point and a radius or from offsets to a chord.
Where should batter boards go?
Far enough back from each corner that the excavator can dig the foundation without touching them, which is why the offset is an input here — typically several feet, depending on how much room the machine and the sloped or shored sides of the excavation take. Set them level with each other where practical, so the strings also carry the elevation reference, and mark the string positions with a saw kerf or a nail rather than a pencil line, because pencil vanishes under rain and traffic. Square the offset rectangle as carefully as the building rectangle: the strings can only be as square as the boards they are tied to.