Caching stale data: How aggressive should we be Tho

We've been working on improving our services that deal with a lot of reading, and caching is a huge piece of that puzzle. We're actually using Redis for some of our most important services. My main question is about balancing stale data versus how often we actually hit the cache. How do you all figure out how long data should hang out in the cache before it's cleared out or updated? Do you have any go-to rules or methods, especially for data that changes a lot but doesn't need to be absolutely up-to-the-second accurate?