On a major contract, we'd had to purchase and install two transmitters of a particular type. I won't give too many details away about the type of transmitters or the manufacturer's name, to protect the innocent and the not-so-innocent, because this is only a story about how to fix a systems engineering problem when project manager solutions don't work.
The problems started in the way that many do. The manufacturer had convinced us that they had built many of these transmitters, so that we were buying an established, reliable product - but this turned out to be debatable. They had built quite a few, but all different to each other, and all different to what we were buying. So it wasn't really a standard model, just something built partly from more or less standard components, with plenty of changes thrown in.
It all started going wrong when the customer rejected the instruction manuals, on the not unreasonable grounds that they were wrong; he sent a two page letter listing problems that he'd found.
My project manager followed the dumbed-down standard solution - send the letter on to the manufacturer. So he corrected the mistakes described in the letter, and issued new manuals - which the customer rejected again.
This time, the customer explained that the letter contained only what he'd found in an hour of checking the transmiiter against the book - it was examples of what the problems were, not a comprehensive list, which was our problem to solve. So my project manager acted again as postman, sending the letter on - and this time, the manufacturer contacted me, to say that he had no idea what to do next.
Having no idea what to do next actually meant that he didn't know what he had built. I soon found out that one transmitter was built by trying things out until they worked, and the second by "copy whatever he does" - there were no more records left of what they had built, which is a big problem when the tramsitters are now located half way across the planet from you.
Normal project management solutions are to withhold the remaining money until it's solved (which didn't work because we'd given them all the money) or to threaten them with no future work (which wouldn't work because everyone knew we'd lost so much money that there would be no more orders). My project manager had no answer to this, apart from lawyers at ten paces.
I realised that any way of getting the manufacturer to work out what was really in the transmitters, and update the manuals, would be expensive for us, or very slow (hence becoming expensive for us anyway) - or both - and decided that there was a less awful solution. At this stage, minimising our further losses was the most important thing to do.
So, what I did was to trace through the wiring and the circuit boards of the entire transmitter - all two metres high, two metres wide, and one metre deep of it. Fortunately, there was no software. It took me three days. Then I re-drew most of the circuit diagrams, added quite a few that had been completely missing, updated the words in the manual where needed (which wasn't many places), and rearranged the chapters so that the book made sense, then printed the new version with a more professional cover design.
The customer stopped complaining and paid us. I sent a copy back to the manufacturer "for information" and, just as I'd expected, never got a reply - presumably it was too embarrassing.