The edge is surface too
A 48 square inch panel with a 30 inch perimeter in 8 oz leather has 3.75 square inches of edge band — 30 inches of perimeter times 0.125 inches of thickness. Against a single grain face that is 7.2 percent of the wetted area. Dye both faces and it drops to 3.8 percent. Neither figure is enormous, but neither is zero, and a batch of two dozen pieces at two coats turns it into real liquid. Thicker leather makes it worse: the same panel in 12 oz has 5.6 square inches of edge.
Where it stops being a rounding error is on small parts. A keyfob 1 inch by 4 inches has 8 square inches of face over two sides and 1.25 square inches of edge, so the edge is 13.5 percent of everything the dye touches.
Counting from the shelf, not from the job
The default batch — 24 pieces, two coats, 8 sq ft per fluid ounce — wants 4.16 fluid ounces. The shelf holds two full 4 oz bottles and one 40 percent bottle, so 9.6 fluid ounces, and the batch goes through with 5.44 left, which is another 31 pieces. Change the batch to 120 and the shelf runs dry at piece 56, 11.18 fluid ounces short of the whole run. Knowing which piece runs dry is more useful than knowing how many fluid ounces are missing, because it tells you whether to start at all.
That matters more with dye than with most materials, because opening a fresh bottle mid-batch is not neutral. Two bottles of the same colour from different lots are not reliably the same colour, and a piece that got half its second coat from one and half from another shows it.
Coverage is not a property of dye
The rate is the input that swings everything and it is not knowable in advance. Vegetable tanned leather drinks dye; a chrome tanned or previously finished surface sits and lets it pool. The flesh side takes noticeably more than the grain side. The first coat takes more than the second because the first one has already filled the surface. A label figure has to cover all of that, so it is a careful figure rather than a predictive one.
The way to replace it is arithmetic on a bottle you have already used: count the pieces it did, work their wetted area the way this page does, multiply by coats, and divide. After that the field holds your number for your leather.
What is not in here
Shade, evenness, whether a colour needs a second coat to look like itself, whether a finish should go over the top, what a resist does to the area that takes dye. None of that is arithmetic and none of it is on this page. It works out volume from an area, a coat count and a rate, and stops.
Questions people ask
How far does a 4 oz bottle of leather dye go?
At 8 square feet per fluid ounce, a 4 oz bottle covers 32 square feet of one coat. On the panel in the form — 48 sq in of face dyed both sides plus a 30 inch perimeter in 8 oz leather, two coats — that is 23 pieces from one bottle. Change the coverage figure and the answer moves proportionally, which is why the field asks for your own count rather than assuming.
Does the edge of the leather need counting?
It is real area, and the page counts it as perimeter times thickness. On a 48 square inch panel with a 30 inch perimeter in 8 oz leather that is 3.75 square inches, about 4 percent of a two-sided piece. On small parts it matters much more: a 1 by 4 inch keyfob is 13.5 percent edge. Whether 4 percent changes your order is another question, but the page shows it rather than dropping it.
How do I work out my own coverage figure?
Use a bottle you have already finished. Count the pieces it dyed, work out the wetted area of one of them — both faces if you dyed both, plus perimeter times thickness — multiply by the number of coats and by the piece count, and divide by the bottle size in fluid ounces. That gives square feet per fluid ounce for your leather, your dye and your hand.
Why does the page tell me which piece runs dry?
Because a shortfall in fluid ounces is abstract and a piece number is not. If the dye runs out at piece 56 of 120, you know before starting that you need another bottle, and you know it is not a problem you can defer to the end. Opening a fresh bottle mid-batch risks a lot difference in colour that will show on whichever piece straddles the change.
Is coverage different for the flesh side and the grain side?
In practice yes, and noticeably so, which is one reason the page treats coverage as a single field you supply rather than trying to model it. The flesh side is open and absorbent and the grain side is not. If the difference matters for what you are making, run the page twice — once for the grain area at one rate and once for the flesh area at another — and add the two volumes.