The main thing is that 'tolerance' is what one can 'tolerate' but doing this requires more analysis than people are willing to put up with.

Eventually, what a design can tolerate is a money problem, though it may detour through a stress problem on its way to being a money problem.

Another distraction is that of fit problems, but that's a combination of an apportionment problem with a cost problem on top of it. If car parts were made to nanometer level accuracy for cheap, most all parts could have nominal fits and they would go together.

Ridiculous unnecessary close tolerances are costly, like for bolt hole locations when the better thing to do is to open up the hole to accommodate an increased tolerance range. Some things need to be close, but having a cad system default on some stupid small tolerance is senseless. Any engineers that deal with assignment of tolerances on parts should first have to deal with the machining of the parts himself before the parts go into production. And then re evaluate the need for the tolerance. Yes any part that deals with fluids needs good tolerance regulation, I have dealt with so many parts that have been way over toleranced, and the reason for it was so the outfit could up the cost of the product, and in most cases there was no profit on it, they paid for it in extra time and headaches.

From a design stand point consider a simple idea, When you apply a tolerance of +- .X, the cost to fabricate will be $, go one step farter to +-.0X cost goes to $$, go to +-.00X cost goes to $$$ and on and on. Again from a design standpoint you can specify whatever tolerance you want, but what can the manufacturing processes deliver? That is the problem for the Manufacturing Engineer, to try and fit your design into the available resources in your factory. If you have equipment that can only hold +- .0X0 and the design calls for a tighter tolerance, they will be forced to outsource it, or expect extremely low yield rates. Another consideration is the raw material available for the product. Will it be made from bar stock, cast, powdered metal, plastic . . . All these things need to be considered. Parts made from sand castings will have far greater variability than those from investment casting.

As was stated above, in general tighter tolerances really mean a lower level of engineering. The real idea for any engineer should be to design parts that will perform their intended function for the lowest possible cost.

It is also useful to understand that tighter tolerances can mean lower assembly costs.

Another thing is that cost is frequently stair-stepped relative to the tolerance; wide bands of tolerance can have exactly the same cost. On any of these stairs, the smaller value will result in less variation in the final product and less variation in performance.

Some of the most expensive parts to use I've come across are the ones that are the lowest cost to make.

IME, reasonable design tolerances are an ongoing refinement. My industrial experience is with small-run machinery with custom or semi-custom designs. It's practically impossible to calculate a tolerance for every dimension on every drawing and do it correctly. No tolerance table in a book on webpage has much relevance. "CNC" manufacturing is not the panacea it's sold for either. If you ask the machine operator or vendor what tolerance they can hit, they usually give you a proudly optimistic number - but they don't mention the numerous extenuating circumstances that cause it to be worse. If you ask the CMM/inspection department you'll see much more realistic responses.

Engineers who say "make it to the print" and lazily walk off don't last, because if we need 8 pieces, only 8 get produced. There usually isn't time or money for scrap and starting over or changing manufacturing processes. This forces the conversation of what process was used, what tolerance was achieved, and is compared to what tolerance will really actually succeed. The drawing tolerance in most cases was a reasonable guess on all of those things up to that point.

We try to stick to common manufacturing methods: milling, drilling, water/plasma/laser jet cut profiles, turning. Key features may have an analytical tolerance based on engineering reasoning, others have tolerance based on reasonable judgment. Most all tolerances could be better - but most of those also have very little return on investment. The key is to get out there in the assembly area and with the component manufacturing centers and find the pain points. Tighten up tolerances if assembly is not going cleanly, open up tolerances if component manufacturing is having trouble, learn how the part is being made and used and adapt the design where possible. If the part is made by outside vendors, now the conversations and discovery must occur several times and the feedback must be aggregated.

David