It's pretty boring to be honest, which is probably why it's so reliable. I sync media using rsync, the movie and stream player is omxplayer, the image and news feeds use Lua and Love2D, and the scheduler is a simple algorithm which was written first in LuaJIT, then Python and soon in Go because of speed and concurrency. The players push their status to a central server for monitoring purposes but I'm replacing that with pull and Ansible. The customer interface is web2py and the schedule creator is desktop wxpython running on Linux, the Mac and Windows.
An old friend approached me with the idea and we have developed it together. He used to work for a media company which had these sort of clients already so we got a head start when they folded and we stepped in to pick up the slack. Subsequent sales have all been word-of-mouth. Product development has been entirely driven by customers: they need something, I put it in and then every player gets the new functionality once it's battle-tested.
When we first looked at the market, we realized that no-one really knew how it worked because it's still very young. So we decided to compete on price and reliability. Instead of lock-in contracts and up-front hardware purchases, we charge a small month-to-month rental and the hardware remains ours.
