Here's a version of the story as it was told to me. I am still trying to find out when this occurred. I'm guessing its been several years since the $20K price tag seems a bit high, at least today.
I, personally, never got around to learning to drive it. I looked over several shoulders with interest, particularly as the situation deteriorated. I've used AutoCAD for about ten years on a steady basis. I've driven Vellum, Pro/E, MiniCad, and I'm getting better at Rhino. I programmed embedded computers professionally for ten years. I've got a pretty good understanding of what goes on inside a computer, and how most software works.
Our chief designer at the time was real fast. We bought one seat for $20k, and put him to work learning it by using it.
Our environment is atypical. We design and produce exhaust systems for big yachts, mostly full custom, and we quote a turnaround time of ten working days. We send a survey sheet comprising an outline of the customer's engine, and let him fill in the critical dimensions we need, locating stuff like the waterline, the holes where the exhaust leaves the boat, and whatever decks, bulkheads, generators, and structure happen to be in the way. We run centerlines from the turbo outlets to the hull exits, size the pipes, and produce 3D surfaced models of the system from one end to the other in two days or less. From those, we produce fax approval sketches and 2D manufacturing drawings. Then we cut sheet alloy into rectangles and preforms, press elbows, roll our own tubes, and weld all the stuff together real nice, pressure test, polish or powder coat, and add insulation. A typical CAD shop might do ten major 3D designs a year. We do hundreds.
Solidworks _should_ have been a good investment. First, of course, we had to build 3D models of a bunch of our standard parts, then import them into an assembly drawing, then, well, you know how to drive it better than I do.
The designer would go through the agony of building the components and structuring them and fitting them into an assembly model, and then there would be a crash, and all of his work would basically evaporate. The assembly model, the product structure, pretty much everything, would become just a big collection of lines, circles, and arcs. No parametric or relational information survived. He went through this several times before we called for help.
The factory sent in four different teams of experts. The first one was not real helpful, but the next three were progressively better. The last one actually worked for the factory and helped write the software. Each would come in all full of sunshine promises, fully expecting to leave the next day with our problem solved. Every one stayed at least a week, and left disappointed and disillusioned. They tried to build the same assemblies as the designer, and got the same results. Great looking assembly, zoomy interface, cool capabilities, and then one random keystroke caused a crash, and all but a pile of ones and zeroes remained.