A curious story came my way, which is a piece of optimisation which has never been (and probably never will be) written up as an academic paper. But that's true of much applied O.R. all the time. Even academics who need to publish find that some of their studies are simply unpublishable. The story came from Dustin Curtis' blog.
He described meeting a sound engineer who set up ambient sound systems for Walt Disney World. Here's an extract:
In the mid 1990’s, the park started researching the problem. It would eventually find no existing solution, so the engineers had to design and construct, on their own, one of the most complex and advanced audio systems ever built. The work paid off: today, as you walk through Disney World, the volume of the ambient music does not change. Ever. More than 15,000 speakers have been positioned using complex algorithms to ensure that the sound plays within a range of just a couple decibels throughout the entire park. It is quite a technical feat acoustically, electrically, and mathematically.
Just think what O.R. tools would be needed for that sort of optimisation and design!