When is in-memory caching actually worth the effort?

hey everyone, i've been thinking a lot about caching lately, specifically for application-level data that's read frequently but updated less often. we all know Redis and Memcached exist, but spinning those up, managing them, and ensuring cache invalidation adds a good chunk of complexity. i'm curious when you all find that the performance gains from in-memory caching truly outweigh the operational overhead. are there specific metrics or scenarios where you'd say 'yes, it's time to add a cache' versus just optimizing database queries or relying on client-side caching? what's your go-to strategy for cache invalidation in these cases