When should you add read replicas to your database?

hey all, starting a discussion about database scaling. we're running a pretty standard relational db for our main application, and writes are generally fine. but we're starting to see some performance degradation on read-heavy queries, especially during peak hours. it feels like a good candidate for introducing read replicas. what are your experiences. at what point did you decide to add read replicas. was it purely performance metrics, or were there other factors like complexity of managing them, cost, or application architecture considerations. curious to hear your thoughts and war stories