When is event sourcing more trouble than its worth for system design?

I've been checking out event sourcing for a new project. And while I like how it helps with audits and rebuilding stuff, the operations side looks like a lot of work. I mean, managing event stores, moving data models around, and that whole eventual consistency thing. For projects that aren't super complicated, or where just a good old audit log with CRUD works, does event sourcing just make things harder than it helps? How do you know when to just stick with the usual way of saving data?