The test we used for every track was the same. Put it on, work on something else, and see how long it takes to notice it is playing. If the answer was under twenty minutes, the track was wrong.
That is an unusual brief. Most game music is written to be noticed: a theme you can hum, a swell when something happens, a motif attached to a character. A business simulation has none of those moments. It has a person looking at a balance sheet at half past eleven at night, deciding whether to open a second site, and they have been doing it for four hours.
There are 17 tracks in Business Baron. They are all written against that test.
Music for a screen you stare at, not a scene you move through
There is no action to score. Nothing chases you, nothing explodes, and the dramatic events of the game are a letter from the bank and a line on a cash flow statement.
So the music is not reacting to events. It is establishing a room to sit in. That pushes it towards a narrow set of decisions and away from most of what makes music interesting in isolation.
Tempo stays in a band that is neither restful nor urgent, because slow music makes a long session feel like a chore and fast music makes an accounting screen feel like a crisis.
Dynamic range stays compressed within a track. Not for loudness, but because a swell is an event, and an event asks for attention we do not want to take. A track that is quiet for ninety seconds and then blooms will pull a player out of a spreadsheet every time it comes round.
The register avoids the middle where speech sits, which keeps it from competing with the part of the brain that is reading words, and it avoids the very top, where repeated transients become irritating after an hour.
And there are no strong melodic hooks. This is the hardest discipline, because a good hook is the thing a composer reaches for, and a good hook heard for the fortieth time in one evening is torture. What is left is harmony, texture and slow movement, which is a smaller toolkit and a more demanding one.
A loop should not have a seam, or an announcement
Every track loops, and the loop has two failure modes.
The obvious one is the seam: an audible join where the end does not lead into the beginning. That is a technical problem and it is solvable with care about tails, reverb that is allowed to run past the loop point and fold back, and an end that resolves into the start rather than at it.
The second failure mode is worse and less discussed. A loop can be technically seamless and still announce itself, because the listener learns the shape. If a track has a distinctive four-note figure at 0:00, then at minute three the player knows exactly where they are, and after an hour they are counting. The music has become a clock.
The way out is to make the beginning unremarkable and to put whatever variation exists somewhere in the middle, so there is no landmark at the boundary. Longer tracks help, and so does layering: several stems that enter and leave on different cycles, so that the combination does not repeat as often as any individual part.
None of this is unique to us. It is how ambient music has always worked. What is specific to a game is that you cannot control how long someone listens, so the design target is not a three minute experience, it is an indefinite one.
The encoding budget is part of the composition
Seventeen tracks of long-form ambient music is, if handled carelessly, a larger download than the entire rest of the game.
So the audio budget was set before the music was written, and it influenced what got written. Sparse, sustained material with limited high-frequency content encodes efficiently at a modest bitrate. Dense percussive material with a lot of transient detail does not, and it also fails the six hour test, so the constraint and the brief happened to point the same way.
Every track is encoded at a bitrate chosen by listening rather than by rule, in a modern lossy format, and checked on the equipment people actually use: laptop speakers, cheap earbuds, and one decent pair of headphones. Laptop speakers are the honest test, because they remove the low end entirely and reveal a track that was relying on it.
The mix is quieter than a music release would be. The game sits alongside whatever else a person has running, and a soundtrack that is mastered to compete is a soundtrack that gets turned off. There is a volume control, and the number of people who turn game music off entirely is high enough that we would rather not give them a reason.
They are named after places because the game is about places
Each track is named after somewhere in the game's world. Ninety-one towns and cities, nearly a century, and a soundtrack that is a set of rooms in those places rather than a set of moods.
There are two reasons for it, one honest and one practical.
The honest one is that it kept the writing specific. "Calm track three" produces generic music. Writing something for a wet Tuesday in a northern market town in 1979 produces choices: what the harmony does, how much space there is, whether there is anything mechanical in the texture. The name came first and the track followed it, which is the opposite of the usual order and worked better.
The practical one is that the names are memorable, and a player who wants to find the track that was playing when they bought their second warehouse has something to search for.
The music is not tied to location in play. We considered it, and a system where the soundtrack changes with the town would mean a player who stays in one place hears three tracks and a player who moves hears everything. Instead the tracks are shuffled with a long memory, so the same one does not come round twice in an evening.
Why we care about this
Business Baron is a game people play for long sessions, and long sessions are where sound design either disappears or becomes a reason to quit. Seventeen tracks that nobody mentions in a review is, for this kind of game, the correct outcome. We would rather the music be the thing you notice you have been listening to at the end of the evening than the thing you notice in the first ten minutes.