When to use a database read replica?

hey all, ive been thinkin about scalin our db reads. were hittin some perf bottlenecks durin peak traffic. i know read replicas r a common solution, but im curious about the tipping point. at what point does the complexity and cost of settin up and managin read replicas start to outweigh the benefits for read-heavy workloads? r there specific metrics or traffic patterns u look at b4 decidin to go this route