My mental model is : each carrier is a sound (a harmonic). I can stack the sounds by chosing an algorithm with multiple carriers. So I start with one carrier and sculpt it by chosing frequency and envelope. This is the time I also activate it's modulators and try to tweak them to make the sound better. Modulators have very complex effects on the sound and I don't try to have a mental model of what they do, just empirical outlines. The next step is to add another carrier to the sound and continue tweaking the current and all the previous carriers.