Planning a glass cut is less about placing parts on a sheet than about building the sequence of cuts. The guillotine constraint requires every cut to run edge to edge, and that single rule eliminates most of the layouts that look possible on paper before the work even begins. A layout is only genuinely cuttable if it can be divided end to end; if it cannot, what you are holding is not a plan but a drawing, and it is of no use at the bench. The first sign of a good plan is therefore not how tightly packed it looks, but whether the order of cuts is obvious.
What piles up under the bench at the end of the day rarely comes from one large mistake. It is the sum of dozens of small decisions made since morning: which sheet was opened first, which orders were cut together, whether the leftover piece was set aside or thrown away. None of these stands out on its own and most are not wrong; but when they all lean the same way, the evening pile stops being a coincidence. Talking about the decisions that produce the pile, rather than the pile itself, is how the problem actually gets addressed.
Opening a sheet commits the rest of it to that order. Committed area is unused area, and it stays that way until the sheet is closed. Considering orders of the same glass type together removes part of the edge waste, because two separate cuts produce two separate edges while a single cut produces one. The person making this call by eye at the bench is doing the same thing; the difference is not in method but in consistency. What software brings is not a new idea, but the same judgement applied to every order, including the rushed ones at the end of the day.
Large pieces left over from a cut are stock, not scrap. Recorded with their dimensions, they are tried first on the next order, and a new sheet is opened only if they do not suffice. A piece that is never recorded becomes glass nobody remembers a week later; while it sits on the rack it does not appear in inventory, and freshly ordered material piles on top of it. Remnant tracking looks like a matter of tidiness, but it directly determines how much material gets consumed and quietly shapes the next purchase.
This is exactly where automatic nesting earns its place. It tries thousands of orderings and picks, among those that respect the guillotine constraint, the one that minimises empty area. It is the same work a person does by eye, but done with the same care every time rather than in a hurry. The decision still belongs to the person: the software lays out the options and shows how much glass each one spends, without saying which to choose. Printing the plan is no small detail either, because when the operator can see where each part comes from, wrong cuts become less likely.
The answer to where to start has to do with measurement. Without recording the utilisation of every cut there is no way to say which jobs went well and which did not; what is not measured cannot be improved. Optimising orders of the same glass type together, and taking usable remnants into stock with their dimensions, are both things that can be done by hand. The only thing a system adds is that they happen for every order and are never forgotten. Once that becomes routine, the difference shows up not in a single cut but in the number of sheets opened by month end.