
Finite element method analysis divides a structure into thousands of small elements and calculates how each one responds to load. The result shows where stress concentrates, how much the structure deflects, and where it would fail first. It is a powerful tool and it is not free, so the honest question is when it earns its cost.
It is usually not worth it when you are repeating a design that has already been built and has performed well in service, or when the structure is conservatively sized against a well-established standard. Analysing something you already have years of evidence for buys you a number you could have assumed.
It is worth it whenever failure would be expensive or dangerous. Anything lifting or supporting a load above people, anything carrying rotating equipment, anything whose collapse would stop a production line: the cost of the analysis is trivial against the cost of being wrong.
It is also worth it when you are trying to take weight or material out. If you want to know how light a structure can safely go, guessing means either overbuilding and paying for steel you did not need, or underbuilding and finding out in service. Analysis is what makes the saving safe to take.
The third case is dynamic or cyclic loading. A structure that is loaded and unloaded thousands of times, or that is driven near a resonant frequency, can fail at stresses far below what a static calculation would call safe. Fatigue is not intuitive, and hand calculations tend to miss it.
The most useful thing about doing the analysis before fabrication is simply that changing a model costs an afternoon. Changing fabricated steel costs the material, the labour, the coating and the schedule.