In the past I’ve written quite a bit about high-level level design, specifically, getting a good overall layout for multiplayer maps that isn’t too convoluted nor over simplistic.
What I haven’t touched on is the more low-level part of this process, where an area from this overview is taken and changed into something much more interesting and enjoyable for the player. An area may look simple from the high-level view, but when you actually have to make it, it requires a whole different way of thinking.
For the Dust bomb spots, I stuck to simplicity: as soon as you enter those areas (which were fairly enclosed), it had to be immediately obvious where the bomb spot was and that the area was significant (read: important). In the case of Dust 2, this also meant making the areas very different to the rest of the map, with more height-changes and cover than anywhere else. Part of the aim was to ensure both a CT and a Terrorist could be in the same area without knowing it - a quick glance would not be enough.
The same is true for any good bomb spot, or any area that is the focus of a CS map. There’s little point trying to do this to the conduits between areas since much of the time they’ll go unused; and the times they are used the chances are they won’t be used in a way that appeals to the gameplay. The problem is that they detract from everything else.
As always, coming up with these designs was ad-hoc. It was a case of knowing the conditions that had to be met: simplicity, significance and depth. The difficult bit is working out exactly what to try - things like a raised platform, second entrance, sunken area, a one-way drop and most likely a combination of those and others. I find these easier to create in the spur-of-the-moment rather than plan them in advance, unless of course they are very significant and affect the rest of the map. In the case of my CS maps, I could afford to effectively make it up as I went along, and when time came to change, I could pick a different spot with similar characteristics (as happened with both Dust and Dust 2.)
The point I’m labouring is that initial designs are great for getting high-level flow right. However, when it gets to the nitty-gritty low-level stuff, nothing really beats going straight in and trying things out.