We're storing complex, loosely-structured data, which perfectly fits MongoDB's schemaless document model. The fact that it will easily scale up with us is just icing on the cake.
Did you look into the schemata of other tracking tools for inspiration, or was this an ab initio analysis?
Mixpanel does deal with funnels differently, however. With their API, you explicitly define events as being some numbered step in a funnel; in ours, the events are agnostic about what funnels they participate in, and we piece them together in a post-processing step.