when dont you really need read replicas for databases?

hey everyone, i've been thinking a lot about scaling read-heavy applications. we're seeing a lot of traffic hitting our primary database, and it's starting to cause some slowdowns. i've read up on read replicas and they seem like a good solution for distributing the load. but how do you know when it's *actually* the right time to implement them? are there specific metrics you look at, or is it more about anticipating future growth? curious to hear your experiences