When should we think about using a distributed cache?

hey everyone, ive been thinking about performance bottlenecks lately. weve hit a few spots where repeated computations or expensive data lookups are slowing things down. its tempting to just slap a cache on everything, but i know that can introduce its own set of complexities, especially around invalidation and consistency. what are your thoughts on the tipping point? when does the complexity of managing a distributed cache outweigh the benefits for a typical web service? are there specific patterns or metrics you look at before making that jump?